Aulvem Blog

What Size Tablet for Reading? 8, 10 or 11 Inch

Thu, 15 Oct 2026 00:00:00 GMT

Calculated for tablets on Amazon.com, showing one whole page in portrait, Japanese bunko paperbacks (4.1 x 5.8 in) fit on an 8-inch tablet, though 8.0 inches is borderline. B6 manga (5.0 x 7.2 in) needs 10 inches or more; none of the tablets checked shows an A4 PDF at full paper size. US paperback sizes weren’t calculated.

The models covered are tablets listed on Amazon.com as of October 2026 with official screen sizes and resolutions. No tablet was tested in person.

Quick reference by what you read

Find the column for what you read and the row for the screen size. Each percentage is how big one whole page shows in portrait compared with the printed page. 100% means the same size as paper.

The columns are Japanese book sizes plus A4: bunko is the Japanese paperback size, and shinsho and B6 are Japanese manga volume sizes. Measure your own book and compare.

Novels: bunko paperback, A6, 4.1 x 5.8 in	Manga: shinsho volume, 4.4 x 6.9 in	Manga: B6 volume, 5.0 x 7.2 in	Magazines: B5, 7.2 x 10.1 in	PDFs: A4, 8.3 x 11.7 in
8-inch class (8-8.7 in)	✓ Paper size or larger, smallest model just barely (about 102-110%)	△ Depends on the model (about 93-101%)	Smaller than paper (about 84-90%)	Smaller than paper (about 59-63%)
10-11-inch class (10.1-11.5 in)	✓ Paper size or larger (about 129-151%)	✓ Paper size or larger (about 119-134%)	✓ Paper size or larger, smallest model just barely (about 106-123%)	Smaller than paper (about 74-87%)
12-13-inch class (11.95-13.1 in)	✓ Paper size or larger (about 155-178%)	✓ Paper size or larger (about 139-157%)	✓ Paper size or larger (about 127-145%)	△ Depends on the model (about 89-102%)

A check mark means every model in that class shows the page at paper size or larger. A triangle means some do and some don’t. “Smaller than paper” means every model shows it smaller.

Each range runs from the smallest to the largest model in the class, bars included.

See how big a book page shows on a tablet

Pick what you read (paperback, shinsho or B6 manga, B5 magazine, A4 PDF) and a screen size class. It calculates what percentage of paper size one page shows at in portrait on current models from major makers, and draws the screen and the page to the same scale.

Screen size for novels and manga

Showing one page in portrait, a bunko page fits at paper size or larger on the 8-inch class, and a B6 manga page on the 10-11-inch class. On the smallest model in each, it is borderline.

Here, sizes are for showing a printed page as it is, one whole page. How much text fits when you change an e-book’s font size is not tested.

On the 8-inch class (8-8.7 in), a bunko page shows at about 102-110% of paper size. The smallest, 8.0 inches, shows it at about 102%, which is borderline. The 10-11-inch class shows it at about 129-151%, larger than the printed book.

A shinsho manga page shows at about 93-101% of paper size on the 8-inch class. That is close to paper size but borderline: an 8.0-inch screen, or a view that keeps the top and bottom bars, shows it slightly smaller.

A B6 manga page shows at about 84-90% on the 8-inch class, smaller than paper. The 10-11-inch class shows it at about 106-123%, at paper size or larger. The smallest, 10.1 inches, is only slightly larger than paper, so borderline.

Reading manga as two-page spreads is covered in what size tablet for manga spreads.

How big a screen PDFs and magazines need

Showing one page in portrait, a whole A4 page shows at about 51-55% of paper size on the 8-inch class and about 64-75% on the 10-11-inch class. Models from 12.9 inches up reach 80% or more; the 13-inch class shows it at about 83-88%. Among the models checked, none shows it at full paper size.

Going by assumed paper text sizes, a whole A4 page on the 8-inch class shows its text much smaller than a bunko paperback’s. On a small screen you zoom in as you read PDFs.

The PDF figures are for A4. US Letter wasn’t calculated.

A B5 magazine page (Japanese JIS B5 size) shows at about 74-87% of paper size on the 10-11-inch class. A 13-inch 4:3 screen shows it at about 98-102%, right at paper size but borderline. A 13-inch 16:10 screen shows about 96%, slightly smaller than paper.

For an A4-size magazine, use the PDF (A4) column of the table.

Portrait or landscape: which is better for reading?

Books that fill one page, such as bunko paperbacks and manga, read best in portrait. For PDFs, turning the tablet sideways and fitting the page width makes the text bigger.

Held sideways, on the 10-11-inch class, fitting the A4 width shows text at about 103-119% of paper size. One screen then holds only about 42-50% of the page, so you scroll as you read.

On an 11-inch screen with a changed aspect ratio, an A4 PDF shows at about 75% of paper size on a 10:7 screen versus about 70% on 16:10. A4, B5, bunko and B6 pages get bigger on screens closer to their own shape, 10:7 to 4:3, while shinsho manga is bigger on 3:2 to 16:10.

Weight difference and eye comfort

Per the makers’ specs as of October 2026, weights run about 293-337 g, or 10.3-11.9 oz, for the 8-inch class. The 10-11-inch class is 433-650 g, or 15.3-22.9 oz, and the 12-13-inch class 530-775 g, or 18.7-27.3 oz. Every 8-inch model is lighter than the lightest 10-11-inch model.

An official comparison of eye strain by screen size alone wasn’t found. Some tablets can warm the screen color at night (Apple’s Night Shift, Samsung’s Eye comfort shield).

For holding one in one hand or reading lying down, see how to choose a tablet for one-handed reading in bed. For e-readers, see tablet or Kindle if you also watch video.

Picks by what you read

Here, the PDF pick comes from 10-11-inch-class models that show a whole A4 page at 72% of paper size or more, whether the bars are 0 or 8 mm. The 72% line leaves out the cheaper 16:10 models, which fall short of it.

The picks are new, with 100+ Amazon.com ratings, 4.0+ stars and Amazon or the maker as the seller. Each is the cheapest that meets its condition, shipping included. Check the price on Amazon.com before you buy.

Novels and shinsho manga on an 8-inch tablet

The Fire HD 8 (8.0 in) shows a bunko page at about 102% of paper size, which is borderline. Shinsho manga shows at about 93-96% and B6 manga at about 84%. It weighs 337 g, or 11.9 oz. Per Amazon’s developer docs, Fire tablets can’t use Google services such as Google Play. Reading apps come from the Amazon Appstore.

Whether the rating count is shared with other models on the listing wasn't checked.

B6 manga at paper size on a budget

The Fire HD 10 (10.1 in) shows a B6 manga page at about 106% of paper size. It is the smallest screen in the 10-11-inch class, so the page is only slightly larger than paper, which is borderline. Like the Fire HD 8, it can’t use Google Play.

A4 PDFs on one portrait page, on a 10-11-inch tablet

Its screen is a bit squarer than most tablets (about 10:7), and it shows A4 at about 74% of paper size.

The rating count is shared with other models on the same listing.

A4 PDFs and magazines near paper size

The Galaxy Tab S10 FE+ (13.1 in) shows a whole A4 page at about 83% of paper size and a B5 magazine page at about 96%. It is the cheapest 12-13-inch tablet that meets the rating and seller conditions. It weighs 664 g, or 23.4 oz.

FAQ

What size tablet is best for reading?

Is an 8-inch tablet big enough for novels?

Showing one page in portrait, on the 8-inch class, a fixed Japanese bunko page shows at about 102-110% of paper size, only slightly larger (borderline) on the smallest 8.0-inch screen. Changing the font size isn’t tested.

How big does a tablet need to be for manga?

Showing manga one page at a time, aim for 10 inches or more for B6 manga. The 10-11-inch class shows a page at paper size or larger, with 10.1 inches borderline.

Shinsho manga on the 8-inch class comes out about paper size or slightly smaller, depending on the model.

What size tablet can read PDF textbooks and papers?

Showing a whole A4 page in portrait, among the models checked, none shows it at full paper size, and even the 12.9-13.1-inch models in the table reach only about 83-88%. Held sideways and fit to width, the 10-11-inch class shows it larger than paper.

Is a cheap Android tablet fine for reading?

For page size, yes: size depends only on the screen’s dimensions, and the table includes budget Android models. But most budget Android models on Amazon.com have too few reviews or are sold by third-party sellers, so none is a pick.

Sharpness, brightness and glare aren’t compared here.

Data and formulas used (for verification)

How this was checked: Screen sizes are worked out from the makers’ official diagonal and resolution as of October 2026. The formulas, assumptions and page-size sources are in the data and formulas at the end of the article. Sources for the night-screen features:. Apple’s explanation. Samsung’s explanation.

Method

Covers the official specifications of tablets listed on Amazon.com (Apple, Samsung, Amazon, Lenovo, Xiaomi and others) as of October 14, 2026. Nothing was tested on real devices. Only models with both an official screen size and resolution were calculated.
Screen width and height come from the official diagonal (inches x 25.4 mm) and the aspect ratio of the resolution: short side = diagonal / sqrt(1 + ratio^2), long side = short side x ratio. Rounded corners are ignored.
Scale of one whole page in portrait = min(screen short side / page width, (screen long side - bars) / page height). 100% means the same size as paper. Bars are assumed 0-8 mm; each range covers every model of that size class across this assumption.
Page sizes: bunko (the Japanese paperback size, A6) = 4.1 x 5.8 in, or 105 x 148 mm; shinsho manga volume (the size of Japanese Jump Comics, two Shueisha listings) = 4.4 x 6.9 in, or 112 x 176 mm; B6 manga volume (one Japanese B6 book listing, same as JIS B6) = 5.0 x 7.2 in, or 128 x 182 mm; magazine = JIS B5 = 7.2 x 10.1 in, or 182 x 257 mm; PDF = A4 = 8.3 x 11.7 in, or 210 x 297 mm. bunko, shinsho, B6 and JIS B5 are Japanese book and paper sizes. Only these listings were checked for shinsho and B6; other publishers were not. A4-size magazines are read from the PDF (A4) column. Letter-size PDFs are not calculated.
Landscape fit-to-width scale = screen long side / page width. Share of the page on one screen = screen short side / (page height x scale).
Text size on screen = text size on paper (assumed: A4 body 10-11 pt, paperback body 8-9.5 pt; 1 pt is about 0.35 mm) x scale. A whole A4 page on the 8-inch class puts A4 body text at about 5.1-6.2 pt, smaller than paperback body text at 8-9.5 pt (estimate).
Size classes: 8-inch class = 8-8.7 in, 10-11-inch class = 10.1-11.5 in, 12-13-inch class = 11.95-13.1 in (range of the Amazon.com models checked). The 14.6-inch Galaxy Tab S Ultra models are sold on Amazon.com but are not put in any class or compared for the picks; for reference, A4 would show at about 93% of paper size and B5 at about 107%. Weights are official Wi-Fi model figures. Two 13.4-inch models are not calculated: the TECLAST T65 has no official maker spec page (only Amazon listings), and the DOOGEE Tab G6 Max has no official resolution.
That Fire tablets can’t use Google Play or other Google services comes from Amazon’s developer documentation; no consumer help page stating it was found.
PDF pick (10-11-inch class, A4 shown at 72% or more of paper on one portrait page), excluded models: every 10-11-inch model listed cheaper than the iPad (A16) 11-inch on Amazon.com, and meeting the rating and seller conditions, falls short. Formula: min(short screen side / 210, (long side - bar) / 297). Fire HD 10 (10.1-inch 16:10) about 64.7%; Galaxy Tab S10 Lite (10.9-inch 16:10) about 69.9%; Galaxy Tab A11+ (11-inch 16:10) about 70.5%. Checked 2026-10-15.

Assumptions

Reading apps hide the top and bottom bars in full screen, but some views keep them. The article subtracts 0-8 mm from the long side in portrait (no official figure; roughly one thin bar at the top and bottom).: 0 (0–8)
Body text size of A4 documents and papers. There is no fixed standard; the article assumes the common 10-11 pt.: 10.5 (10–11)
Body text size of a Japanese paperback (bunko). It varies by publisher and no official standard was found, so the article holds 8-9.5 pt.: 8.75 (8–9.5)

Official values used

Item	Value	Source	Checked
A4 page size	width: 210 / height: 297 mm	kikakurui.com	2026-10-14
JIS B5 page size (Japanese)	width: 182 / height: 257 mm	nipponpapergroup.com ‘B5 182mm×257mm’	2026-10-14
Bunko paperback size (Japanese, A6)	width: 105 / height: 148 mm	shinchosha.co.jp	2026-10-14
JIS B6 page size (Japanese)	width: 128 / height: 182 mm	kaunet.com ‘A5 148 x 210mm B5 182 x 257mm A6 105 x 148mm B6 128 x 182mm’	2026-10-14
B6 manga volume size (Japanese)	height: 182 / width: 128 mm	hanmoto.com	2026-10-14
Shinsho manga volume size (Japanese Jump Comics)	height: 176 / width: 112 mm	hanmoto.com	2026-10-14
Fire tablets and Google services	can’t use Google services on Amazon Fire devices statement	developer.amazon.com ‘you can’t use Google services on Amazon Fire devices’	2026-10-14
Apple Night Shift	Warmer display colors after sunset; scheduled or manual feature	support.apple.com ‘Night Shift automatically adjusts the colors of your display to the warmer end	2026-10-14
Samsung Eye comfort shield and Extra dim	Eye comfort shield (limit blue light) and Extra dim (dim the screen) feature	samsung.com ‘You can also use Eye comfort shield to limit the screen’s blue light, and Extra	2026-10-14

See how big a book page shows on a tablet: quick reference

The calculator in this article shows the result and a matching product for your own settings. Settings can be set in the URL; each link below opens the calculator with those settings.

Settings	Result	Calculator
Novels (paperback, A6), 8-inch class (8-8.7 in)	✓ 8-inch class: a Japanese paperback (bunko, A6) page at paper size or larger. One Japanese paperback (bunko, A6) page (105 x 148 mm) in portrait shows at about 102-110% of paper size on the 8-inch class. Calculated for every model in this class, allowing 0-8 mm for top and bottom bars	Open with these settings
Manga (shinsho), 8-inch class (8-8.7 in)	△ 8-inch class: a Shinsho manga volume page reaches paper size on some models and is smaller on others. One Shinsho manga volume page (112 x 176 mm) in portrait shows at about 93-101% of paper size on the 8-inch class. Calculated for every model in this class, allowing 0-8 mm for top and bottom bars	Open with these settings
Manga (B6), 8-inch class (8-8.7 in)	✕ 8-inch class: a B6 manga volume page smaller than paper. One B6 manga volume page (128 x 182 mm) in portrait shows at about 84-90% of paper size on the 8-inch class. Calculated for every model in this class, allowing 0-8 mm for top and bottom bars	Open with these settings
Magazines (B5), 8-inch class (8-8.7 in)	✕ 8-inch class: a Magazine (B5) page smaller than paper. One Magazine (B5) page (182 x 257 mm) in portrait shows at about 59-63% of paper size on the 8-inch class. Sideways and fit to width, it shows at about 94-104% of paper size, but one screen holds only about 42-47% of the page. Calculated for every model in this class, allowing 0-8 mm for top and bottom bars	Open with these settings
PDFs (A4), 8-inch class (8-8.7 in)	✕ 8-inch class: a PDF (A4) page smaller than paper. One PDF (A4) page (210 x 297 mm) in portrait shows at about 51-55% of paper size on the 8-inch class. Sideways and fit to width, it shows at about 82-90% of paper size, but one screen holds only about 42-47% of the page. A4 body text (assumed 10-11 pt) shows at about 5.1-6 pt (estimate). Calculated for every model in this class, allowing 0-8 mm for top and bottom bars	Open with these settings
Novels (paperback, A6), 10-11-inch class (10.1-11.5 in)	✓ 10-11-inch class: a Japanese paperback (bunko, A6) page at paper size or larger. One Japanese paperback (bunko, A6) page (105 x 148 mm) in portrait shows at about 129-151% of paper size on the 10-11-inch class. Calculated for every model in this class, allowing 0-8 mm for top and bottom bars	Open with these settings
Manga (shinsho), 10-11-inch class (10.1-11.5 in)	✓ 10-11-inch class: a Shinsho manga volume page at paper size or larger. One Shinsho manga volume page (112 x 176 mm) in portrait shows at about 119-134% of paper size on the 10-11-inch class. Calculated for every model in this class, allowing 0-8 mm for top and bottom bars	Open with these settings
Manga (B6), 10-11-inch class (10.1-11.5 in)	✓ 10-11-inch class: a B6 manga volume page at paper size or larger. One B6 manga volume page (128 x 182 mm) in portrait shows at about 106-123% of paper size on the 10-11-inch class. Calculated for every model in this class, allowing 0-8 mm for top and bottom bars	Open with these settings
Magazines (B5), 10-11-inch class (10.1-11.5 in)	✕ 10-11-inch class: a Magazine (B5) page smaller than paper. One Magazine (B5) page (182 x 257 mm) in portrait shows at about 74-87% of paper size on the 10-11-inch class. Sideways and fit to width, it shows at about 119-137% of paper size, but one screen holds only about 42-50% of the page. Calculated for every model in this class, allowing 0-8 mm for top and bottom bars	Open with these settings
PDFs (A4), 10-11-inch class (10.1-11.5 in)	✕ 10-11-inch class: a PDF (A4) page smaller than paper. One PDF (A4) page (210 x 297 mm) in portrait shows at about 64-75% of paper size on the 10-11-inch class. Sideways and fit to width, it shows at about 103-119% of paper size, but one screen holds only about 42-50% of the page. A4 body text (assumed 10-11 pt) shows at about 6.4-8.3 pt (estimate). Calculated for every model in this class, allowing 0-8 mm for top and bottom bars	Open with these settings
Novels (paperback, A6), 12-13-inch class (11.95-13.1 in)	✓ 12-13-inch class: a Japanese paperback (bunko, A6) page at paper size or larger. One Japanese paperback (bunko, A6) page (105 x 148 mm) in portrait shows at about 155-178% of paper size on the 12-13-inch class. Calculated for every model in this class, allowing 0-8 mm for top and bottom bars	Open with these settings
Manga (shinsho), 12-13-inch class (11.95-13.1 in)	✓ 12-13-inch class: a Shinsho manga volume page at paper size or larger. One Shinsho manga volume page (112 x 176 mm) in portrait shows at about 139-157% of paper size on the 12-13-inch class. Calculated for every model in this class, allowing 0-8 mm for top and bottom bars	Open with these settings
Manga (B6), 12-13-inch class (11.95-13.1 in)	✓ 12-13-inch class: a B6 manga volume page at paper size or larger. One B6 manga volume page (128 x 182 mm) in portrait shows at about 127-145% of paper size on the 12-13-inch class. Calculated for every model in this class, allowing 0-8 mm for top and bottom bars	Open with these settings
Magazines (B5), 12-13-inch class (11.95-13.1 in)	△ 12-13-inch class: a Magazine (B5) page reaches paper size on some models and is smaller on others. One Magazine (B5) page (182 x 257 mm) in portrait shows at about 89-102% of paper size on the 12-13-inch class. Sideways and fit to width, it shows at about 139-155% of paper size, but one screen holds only about 44-54% of the page. Only 4:3 screens of about 13 inches come close to paper size (about 98-102%). Every other model shows it smaller than paper. Calculated for every model in this class, allowing 0-8 mm for top and bottom bars	Open with these settings
PDFs (A4), 12-13-inch class (11.95-13.1 in)	✕ 12-13-inch class: a PDF (A4) page smaller than paper. One PDF (A4) page (210 x 297 mm) in portrait shows at about 77-88% of paper size on the 12-13-inch class. Sideways and fit to width, it shows at about 120-134% of paper size, but one screen holds only about 44-54% of the page. A4 body text (assumed 10-11 pt) shows at about 7.7-9.7 pt (estimate). Calculated for every model in this class, allowing 0-8 mm for top and bottom bars	Open with these settings

Best Earbuds Mic for a Noisy Office: Look Past ANC

Wed, 14 Oct 2026 00:00:00 GMT

For calls in a noisy office, this guide picks the Bose QuietComfort Ultra Earbuds (2nd Gen), which Bose says reduces office chatter and wind on calls with AI. It lists 8 mics, but Bose doesn’t say how many are for calls.

This guide is based on the makers’ official pages. No earbuds were tested in person, so how your voice sounds to the other person isn’t judged.

ANC (active noise cancellation): reduces the noise you hear
Beamforming: using several mics to pick up sound from your mouth more strongly
Bone-conduction sensor: tells your voice apart from other sound by the vibration when you speak (also called voice pickup)
AI noise removal: processing that reduces background noise in the voice it sends
HFP: the Bluetooth call profile, which is mono
LE Audio: the newer Bluetooth standard, which can keep stereo during calls when the conditions are met
Teams certification: Certified for Microsoft Teams, which checks audio quality and Teams call controls

ANC and call noise reduction are separate

Per the makers’ pages as of October 2026, ANC (active noise cancellation) cuts the noise you hear; call noise reduction cuts the noise the other person hears. Sony, Bose and Jabra describe them as separate features, and none says strong ANC keeps background noise out of your calls.

Among the models checked, none says more mics mean better calls. The Sony WF-1000XM6’s 8 mics are for ANC; calls use 2 mics plus a bone-conduction sensor. Bose doesn’t say whether its 8 mics include the ANC ones.

Will it remove chatter, keyboards, air conditioning, street noise and wind?

Among the earbuds compared here, only the Bose QuietComfort Ultra Earbuds (2nd Gen) says it reduces office chatter on calls. Bose describes AI suppression of chatter and wind.

No page names keyboard clatter or air-conditioning hum as something it removes. Bose names voices and wind, the Technics EAH-AZ100 wind and sudden noises, and no page gives figures.

Among the models checked, whether train or street noise is removed like office chatter isn’t clear, and this guide hasn’t verified it.

Every model in the table lists wind measures, and the methods differ. See the Wind measures column.

Call features and wind measures the makers publish

The table comes from the makers’ official pages. Read the “Call features” and “Wind measures” columns. “Not confirmed” means this guide didn’t check, not that the feature is missing, and the table shows only what is published, not how good it sounds.

Model	Call features	Wind measures	ANC
Sony WF-1000XM6	2 call mics aimed at your mouth, bone-conduction sensor and AI	Mic-area structure (micro-perforations, streamlined shape) reduces wind noise	Yes
Sony WF-1000XM5	Voice pickup and bone-conduction sensor (call mic count not listed)	Mic design that reduces wind noise	Yes
Bose QuietComfort Ultra Earbuds (2nd Gen)	8 mics and AI noise suppression (unclear whether the 8 include the ANC mics)	AI suppression said to reduce wind	Yes
Jabra Elite 10 (1st Gen)	6 mics (one per side picks up your voice by bone conduction)	Wind mode (switches to the in-ear mic)	Not confirmed
Google Pixel Buds Pro 2	3 mics per side, voice accelerometer and Clear Calling (how it works isn’t explained)	Wind-blocking mesh covers	Not confirmed
Apple AirPods Pro 3	Dual beamforming mics, an inward mic and a speech-detecting accelerometer (Voice Isolation)	Voice Isolation removes wind and other noise (calls on iPhone, iPad and Mac)	Not confirmed
Technics EAH-AZ100	3 mics per side (6 total) and AI noise reduction	AI said to suppress wind and sudden noises	Yes
Poly Voyager Free 60	3 mics per side and AI noise removal (NoiseBlockAI)	WindSmart	Yes

Among the models checked, the Call features column shows which models publish a voice-pickup method or AI noise removal.

The Jabra Elite 10 found on Amazon is the first generation. The newer Gen 2 lists 6 mics, but neither their layout nor a Teams certification was found.

Does Teams certification tell you how well noise is removed?

Per Microsoft, no standard for how strongly noise is removed is listed on its pages. Teams certification (Certified for Microsoft Teams) checks audio quality and Teams call controls such as mute sync. Nothing found says earbuds need the certification to work in Teams.

For earbuds paired straight to the PC, Teams certification covers classic Bluetooth only; LE Audio isn’t certified, and a PC with a separate USB Bluetooth adapter is outside the certification.

The only Teams certification found here is for the Teams version of the Poly Voyager Free 60. The regular UC version may not be certified, and the Amazon listing is the UC version from a third-party seller, so this guide doesn’t recommend it.

Why call audio sounds muffled on Windows

Per Microsoft, with classic Bluetooth, Windows switches to the call profile (HFP) when a call app uses the earbuds’ mic, and audio becomes mono. HFP also limits the audio band, which is why call audio sounds muffled. LE Audio is the exception.

The exception is LE Audio: Windows 11 24H2 or later on a PC with factory LE Audio support, and earbuds that keep stereo during calls. None of the models here officially confirms stereo during calls.

Per Apple, AirPods Voice Isolation, which removes background noise such as wind during calls, works in calls on an iPhone, iPad or Mac. Apple doesn’t say whether the same processing applies to Windows calls.

How to check for your own office

Decide which noise you want gone, and which call app and PC you use
In the tool below, see which models publish a feature for that noise, then check the ones you like on the maker’s page
After buying, ask a coworker how you sound on a call

Check which earbuds publish call features for a noisy office

Pick the noise around you (voices, typing, air conditioning, wind), your call app (Teams, Zoom or other) and your PC (Windows or Mac). It shows the specs to check, the caveats and the models that publish a call feature for that noise. It doesn't rate how you sound.

In the tool, pick the noise around you, your call app and your PC to see what to check and which models publish a feature for that noise. This guide doesn’t judge how you sound to the other person, because no earbuds were recorded and compared. It lists only the call features the makers publish.

Picks by situation

Per the makers’ pages, candidates here list call features, wind measures and ANC, plus AI noise reduction or reduced chatter. Of those, only listings with 100+ Amazon.com ratings, 4.0+ stars and a seller of Amazon or the maker are considered.

For calls in a noisy office

The Bose QuietComfort Ultra Earbuds (2nd Gen) is the only candidate that meets those rating and seller rules, and it says it reduces office chatter and wind.

Whether the rating count is shared with other models on the listing wasn't checked.

For calls on an iPhone and a Mac

This guide picks the AirPods Pro 3 for iPhone and Mac calls. Per Apple, it has Voice Isolation and dual beamforming mics. Apple doesn’t say whether Voice Isolation works in Windows calls.

FAQ

Does strong ANC mean the earbuds’ mic also cuts the noise around me on calls?

No. ANC cuts the noise you hear, while call noise reduction cuts the noise the other person hears. Sony, Bose and Jabra describe them as separate features.

Do more mics mean cleaner calls?

No official page among the major makers’ earbuds compared here says that more mics mean better calls. The Sony WF-1000XM6 lists 8 mics for ANC, plus 2 call mics and a bone-conduction sensor.

Will earbuds cut keyboard clatter or air-conditioning hum on calls?

No official page among the earbuds compared here names keyboard or air-conditioning noise as something it removes.

Do Teams-certified earbuds also remove more noise?

Microsoft’s pages don’t list how strongly noise is removed as a Teams certification criterion.

Data and formulas used (for verification)

Method

Scope: current earbuds from Sony, Bose, Jabra, Apple, Google, Technics and Poly as of 2026-10-10, plus official Microsoft and Apple documentation. No recordings were made.
This guide does not judge how a mic sounds. Mic counts, methods and certifications are the makers’ published specs.

Official values used

Item	Value	Source	Checked
windows_hfp_mono	When a mic is opened, Windows switches from A2DP (stereo) to HFP (mono, 8 kHz or 16 kHz); only a Bluetooth LE Audio PC plus earbuds can keep stereo during calls	learn.microsoft.com Microsoft Learn: ‘Bluetooth Classic audio supports stereo audio playback over th	2026-10-10
windows11_hfp_trigger	Windows 11 picks HFP when an app opens the mic or plays a Communications-category stream	learn.microsoft.com Microsoft Learn: ‘A2DP only supports audio output from the host to the audio dev	2026-10-10
windows_le_audio_requirements	Stereo during calls needs Windows 11 24H2+, a PC with factory-integrated Bluetooth LE support, and earbuds that support stereo while the mic is in use	support.microsoft.com	2026-10-10
sony_video_call_note	Sony tells users to check the PC link is HFP or HSP for video calls	helpguide.sony.net Sony: ‘When using a video calling application on a computer, check that the head	2026-10-10
jabra_mics_elite10	6 microphones	jabra.com	2026-10-10
xm6_anc	Yes	helpguide.sony.net	2026-10-10
xm5_anc	Yes	helpguide.sony.net	2026-10-10
bose_anc	Yes	bose.com	2026-10-10
az100_anc	Yes	jp.technics.com	2026-10-10
anc_vs_call_nr_separate	Makers describe ANC (what you hear) and call microphone noise reduction (what the other person hears) as separate features	sony.jp	2026-10-10
bose_anc_vs_call	Bose lists ANC changes and AI call noise suppression as separate items	bose.com	2026-10-10
xm6_call_mics	Sony WF-1000XM6: 2 mics for calls (directional toward the mouth) + bone conduction sensor + AI noise reduction; ANC uses a separate 4 mics per earbud (8 total)	sony.jp	2026-10-10
xm5_call_mics	Sony WF-1000XM5: wind-reducing mic structure + precision voice pickup technology + bone conduction sensor; 3 ANC mics per earbud (6 total)	sony.jp	2026-10-10
bose_qc2_mics	Bose QuietComfort Ultra Earbuds (2nd Gen): 8 microphones; AI noise suppression for calls	bose.com	2026-10-10
jabra_elite10g2_mics	Jabra Elite 10 Gen 2: 6 MEMS microphones, noise reduction on calls, HFP v1.8	jabra.com	2026-10-10
jabra_elite10_wind	Jabra Elite 10 (1st gen): a wind mode switches to the in-ear microphone; Jabra says one mic per earbud picks up voice through bone conduction	jabra.com	2026-10-10
pixel_buds_pro2_mics	Google Pixel Buds Pro 2: three microphones per earbud, voice accelerometer, wind-blocking mesh covers, Clear Calling	store.google.com	2026-10-10
airpods_pro3_mics	Apple AirPods Pro 3: dual beamforming microphones, an inward-facing microphone, a speech-detecting accelerometer, Voice Isolation	apple.com	2026-10-10
airpods_voice_isolation	AirPods Voice Isolation removes background noise such as wind for the listener; AirPods 4, Pro 2 and Pro 3	support.apple.com	2026-10-10
az100_mics	Technics EAH-AZ100: three microphones per earbud plus AI noise reduction (trained on 500 million audio samples); suppresses wind and sudden noises	us.technics.com Technics: ‘The AZ100 is equipped with AI trained on 500 million pieces of audio	2026-10-10
poly_free60_mics	Poly Voyager Free 60: three microphones per earbud, WindSmart wind protection, NoiseBlockAI	hp.com	2026-10-10
poly_free60_teams	Poly Voyager Free 60 is Microsoft Teams Certified (Teams version only), so the plain UC model is not necessarily certified	hp.com	2026-10-10
teams_cert_meaning	Teams certification checks audio quality and Teams call control (e.g. mute sync); Microsoft’s two pages state no noise-removal strength criterion	learn.microsoft.com	2026-10-10
teams_native_bt_limits	Native Bluetooth Teams certification covers Classic Bluetooth only (LE Audio not certified); aftermarket USB Bluetooth adapters are not certified	learn.microsoft.com Microsoft Learn: ‘This certification is currently for Classic Bluetooth devices.	2026-10-10
xm6_wind_mic	Streamlined body, micro-perforated mic area and an airflow structure to reduce wind noise, which also helps what the caller hears	sony.jp	2026-10-10

Check which earbuds publish call features for a noisy office: quick reference

The calculator in this article shows the result and a matching product for your own settings. Settings can be set in the URL; each link below opens the calculator with those settings.

Settings	Result	Calculator
Coworkers talking, Teams, Windows	✓ 6 models publish a feature for coworkers talking. Sony WF-1000XM6, Sony WF-1000XM5, Bose QuietComfort Ultra Earbuds (2nd Gen), Jabra Elite 10 (1st Gen), Google Pixel Buds Pro 2, Apple AirPods Pro 3. For voices around you, look for a way of picking out your own voice (bone-conduction or motion sensors, mics aimed at your mouth) or a stated reduction of nearby speech. The Bose QuietComfort Ultra Earbuds (2nd Gen) are the ones that clearly say they cut coworkers’ voices in calls. ANC cuts the noise you hear; the noise the other person hears is handled by a separate call feature. More mics don’t mean clearer calls (the Sony WF-1000XM6’s 8 mics are for ANC; calls use 2). Bose doesn’t say whether its 8 mics include the ANC ones. Teams certification covers audio quality and call controls such as mute sync; Microsoft’s pages don’t list a standard for noise removal. For earbuds paired straight to the PC, certification covers classic Bluetooth only; LE Audio and setups with a separate USB Bluetooth adapter are outside it. The only certification found is for the Teams version of the Poly Voyager Free 60. The Amazon listing is the UC version from a third-party seller, so it isn’t recommended. On Windows, once a call app uses the mic, audio switches to the call profile (HFP) and turns mono. The exception is LE Audio (an LE Audio PC on Windows 11 24H2 or later and earbuds that keep stereo in calls); no model here officially confirms stereo during calls. AirPods Pro 3 Voice Isolation works in calls on an iPhone, iPad or Mac; Apple doesn’t say whether it works on Windows	Open with these settings
Keyboard typing, Teams, Windows	△ None names keyboard typing. 4 models publish AI noise removal. Sony WF-1000XM6, Bose QuietComfort Ultra Earbuds (2nd Gen), Technics EAH-AZ100, Poly Voyager Free 60. No official page checked names keyboard clatter or air-conditioning hum as something it removes. Even models with AI noise removal don’t say which sounds it handles or how well. ANC cuts the noise you hear; the noise the other person hears is handled by a separate call feature. More mics don’t mean clearer calls (the Sony WF-1000XM6’s 8 mics are for ANC; calls use 2). Bose doesn’t say whether its 8 mics include the ANC ones. Teams certification covers audio quality and call controls such as mute sync; Microsoft’s pages don’t list a standard for noise removal. For earbuds paired straight to the PC, certification covers classic Bluetooth only; LE Audio and setups with a separate USB Bluetooth adapter are outside it. The only certification found is for the Teams version of the Poly Voyager Free 60. The Amazon listing is the UC version from a third-party seller, so it isn’t recommended. On Windows, once a call app uses the mic, audio switches to the call profile (HFP) and turns mono. The exception is LE Audio (an LE Audio PC on Windows 11 24H2 or later and earbuds that keep stereo in calls); no model here officially confirms stereo during calls. AirPods Pro 3 Voice Isolation works in calls on an iPhone, iPad or Mac; Apple doesn’t say whether it works on Windows	Open with these settings
Air conditioning, Teams, Windows	△ None names air conditioning. 4 models publish AI noise removal. Sony WF-1000XM6, Bose QuietComfort Ultra Earbuds (2nd Gen), Technics EAH-AZ100, Poly Voyager Free 60. No official page checked names keyboard clatter or air-conditioning hum as something it removes. Even models with AI noise removal don’t say which sounds it handles or how well. ANC cuts the noise you hear; the noise the other person hears is handled by a separate call feature. More mics don’t mean clearer calls (the Sony WF-1000XM6’s 8 mics are for ANC; calls use 2). Bose doesn’t say whether its 8 mics include the ANC ones. Teams certification covers audio quality and call controls such as mute sync; Microsoft’s pages don’t list a standard for noise removal. For earbuds paired straight to the PC, certification covers classic Bluetooth only; LE Audio and setups with a separate USB Bluetooth adapter are outside it. The only certification found is for the Teams version of the Poly Voyager Free 60. The Amazon listing is the UC version from a third-party seller, so it isn’t recommended. On Windows, once a call app uses the mic, audio switches to the call profile (HFP) and turns mono. The exception is LE Audio (an LE Audio PC on Windows 11 24H2 or later and earbuds that keep stereo in calls); no model here officially confirms stereo during calls. AirPods Pro 3 Voice Isolation works in calls on an iPhone, iPad or Mac; Apple doesn’t say whether it works on Windows	Open with these settings
Wind (window or outdoors), Teams, Windows	✓ 7 models publish a feature for wind (window or outdoors). Sony WF-1000XM6, Sony WF-1000XM5, Bose QuietComfort Ultra Earbuds (2nd Gen), Jabra Elite 10 (1st Gen), Google Pixel Buds Pro 2, Technics EAH-AZ100, Poly Voyager Free 60. For wind, look for a stated wind measure (mic design, mesh, wind mode, AI). ANC cuts the noise you hear; the noise the other person hears is handled by a separate call feature. More mics don’t mean clearer calls (the Sony WF-1000XM6’s 8 mics are for ANC; calls use 2). Bose doesn’t say whether its 8 mics include the ANC ones. Teams certification covers audio quality and call controls such as mute sync; Microsoft’s pages don’t list a standard for noise removal. For earbuds paired straight to the PC, certification covers classic Bluetooth only; LE Audio and setups with a separate USB Bluetooth adapter are outside it. The only certification found is for the Teams version of the Poly Voyager Free 60. The Amazon listing is the UC version from a third-party seller, so it isn’t recommended. On Windows, once a call app uses the mic, audio switches to the call profile (HFP) and turns mono. The exception is LE Audio (an LE Audio PC on Windows 11 24H2 or later and earbuds that keep stereo in calls); no model here officially confirms stereo during calls. AirPods Pro 3 Voice Isolation works in calls on an iPhone, iPad or Mac; Apple doesn’t say whether it works on Windows	Open with these settings
Wind (window or outdoors), Teams, Mac	✓ 8 models publish a feature for wind (window or outdoors). Sony WF-1000XM6, Sony WF-1000XM5, Bose QuietComfort Ultra Earbuds (2nd Gen), Jabra Elite 10 (1st Gen), Google Pixel Buds Pro 2, Apple AirPods Pro 3, Technics EAH-AZ100, Poly Voyager Free 60. For wind, look for a stated wind measure (mic design, mesh, wind mode, AI). ANC cuts the noise you hear; the noise the other person hears is handled by a separate call feature. More mics don’t mean clearer calls (the Sony WF-1000XM6’s 8 mics are for ANC; calls use 2). Bose doesn’t say whether its 8 mics include the ANC ones. Teams certification covers audio quality and call controls such as mute sync; Microsoft’s pages don’t list a standard for noise removal. For earbuds paired straight to the PC, certification covers classic Bluetooth only; LE Audio and setups with a separate USB Bluetooth adapter are outside it. The only certification found is for the Teams version of the Poly Voyager Free 60. The Amazon listing is the UC version from a third-party seller, so it isn’t recommended. How calls work on a Mac wasn’t checked for this guide. AirPods Pro 3 Voice Isolation works in Mac calls	Open with these settings
Coworkers talking, Zoom, Windows	✓ 6 models publish a feature for coworkers talking. Sony WF-1000XM6, Sony WF-1000XM5, Bose QuietComfort Ultra Earbuds (2nd Gen), Jabra Elite 10 (1st Gen), Google Pixel Buds Pro 2, Apple AirPods Pro 3. For voices around you, look for a way of picking out your own voice (bone-conduction or motion sensors, mics aimed at your mouth) or a stated reduction of nearby speech. The Bose QuietComfort Ultra Earbuds (2nd Gen) are the ones that clearly say they cut coworkers’ voices in calls. ANC cuts the noise you hear; the noise the other person hears is handled by a separate call feature. More mics don’t mean clearer calls (the Sony WF-1000XM6’s 8 mics are for ANC; calls use 2). Bose doesn’t say whether its 8 mics include the ANC ones. Teams certification is for Teams; it isn’t a standard for calls in Zoom or other apps. On Windows, once a call app uses the mic, audio switches to the call profile (HFP) and turns mono. The exception is LE Audio (an LE Audio PC on Windows 11 24H2 or later and earbuds that keep stereo in calls); no model here officially confirms stereo during calls. AirPods Pro 3 Voice Isolation works in calls on an iPhone, iPad or Mac; Apple doesn’t say whether it works on Windows	Open with these settings

What Size Power Bank Do You Need for a 3-Day Trip?

Mon, 12 Oct 2026 00:00:00 GMT

On a 3-day trip at 60% efficiency, if you can charge at the hotel each night, a phone alone just barely gets by on 10,000mAh. A tablet or handheld console needs 20,000mAh, and a laptop 25,000mAh. With no outlet for 3 days and a phone alone, carry 20,000mAh and 10,000mAh banks.

Power bank size for a 3-day trip at a glance

Assuming 1 charge a day (half for tablets and laptops) at 60% efficiency, the size you need comes out like this.

What you carry	Can charge at the hotel each night	No outlet for 3 days
Phone only	10,000mAh (just barely)	Two banks: 20,000mAh and 10,000mAh
Phone + tablet (iPad Air 11-inch)	20,000mAh	Two banks: 25,000mAh and 25,000mAh
Phone + handheld console (Nintendo Switch 2)	20,000mAh	Some models need more than two banks. Charge fewer times or recharge on the way
Phone + laptop (MacBook Air 13-inch)	25,000mAh	Some models need more than two banks. Charge fewer times or recharge on the way

5,000mAh may not fill even the iPhone 18 Pro, the smallest battery here, from 0% to 100% once. Use it as a half-day spare, not as your main bank for 3 days.

Why a power bank charges less than its label says

The label shows the battery inside the bank, and some power is lost while charging. Wh (watt-hours) is capacity times voltage, so it compares different devices. Conversion efficiency is the share you can actually use.

A 10,000mAh power bank holds 37Wh, and about 22-26Wh (60-70%) reaches your device. Power is lost as heat when it is drawn out over USB and again when the device battery charges. The makers’ figure of about 6,000-7,000mAh counts at USB’s 5V (30-35Wh), while this article assumes more loss.

10,000mAh charges an iPhone 18 Pro about 1.3 to 1.8 times. Dividing 10,000mAh by its 4,056mAh battery gives 2.5, but power is lost along the way, so the real number is lower. The 4,056mAh is an EU-label figure from press reports; Apple does not publish it.

Per the makers’ figures as of October 2026, wireless charging is less efficient than a cable, so expect fewer charges than the wired figures here..

Charging at the hotel, and not charging

If you can charge each night and carry only a phone, 10,000mAh just barely covers a day. The margin is only about 0.8Wh. The Pro Max’s 5,391mAh is an EU-label figure from press reports (Apple does not publish it), and the US models are unchecked, so 20,000mAh is safer for a Pro Max.

With no outlet for 3 days, a phone alone needs 20,000mAh and 10,000mAh banks. A tablet or console raises the need.

An iPad Air 11-inch with a Pro Max needs about 105.7Wh; two 25,000mAh banks (about 108Wh) cover it. With a Switch 2, an iPhone 18 Pro needs about 107.2Wh and just barely fits, but a Pro Max needs about 122.6Wh and does not. A Steam Deck or MacBook Air does not fit, so charge fewer times or recharge on the way.

Per Amtrak’s website, every seat in Amtrak unreserved coach has a 120V outlet, so charging on board lets you carry less. It varies by train and service.

How to check your own devices

Find each device’s battery capacity. If it’s listed in Wh, use that. If it’s only in mAh, convert it to Wh in the next step. The table below shows which unit is published.
Multiply mAh × voltage ÷ 1000 to get Wh. The voltage wasn’t found in the official specs, so use 3.85 V to avoid underestimating.
Multiply by your charges from the bank: one day’s worth if you can charge each night, all the days if you can’t. A phone is 1 charge a day.
Multiply the power bank’s Wh by the conversion efficiency from the section above to get what you can actually use. Compare that with your total.

Device	What’s published
iPad, MacBook	Wh
Galaxy, Pixel, Switch 2	mAh (voltage not published)

To share one bank, add each person’s Wh and pick a size from the table. This assumes you can charge each night and each person charges a phone once a day. If both carry only a phone, two of the largest phone here need about 41.6Wh, so 20,000mAh (at least 43.2Wh usable) is the guide.

With a camera, add its battery Wh: Sony’s NP-FZ100 is 16.4Wh, and the GoPro Enduro is about 7.3Wh at an assumed 3.85V. A phone of about 20.8Wh plus the NP-FZ100 comes to about 37.2Wh, or 28.1Wh with the GoPro. Both fit within 20,000mAh if you charge one camera battery a day; a battery listed only in mAh needs its voltage to get Wh.

Apple Watch and AirPods don’t publish their Wh, so they aren’t counted.

Calculate the power bank size for your trip

Pick the devices you bring (phone, tablet, handheld console, laptop), the number of days and whether you can charge at the hotel each night. It shows the Wh you need, which of 5,000, 10,000, 20,000 and 25,000mAh is enough, and how many charges each device gets.

In the calculator, pick your devices, the number of days and whether you can charge each night. It shows the Wh you need, which of 5,000, 10,000, 20,000 and 25,000mAh is enough, and how many charges each device gets.

Taking a power bank on a plane

Under FAA rules, power banks of 100Wh or less can be carried on without airline approval. Those of 101 to 160Wh need the airline’s approval, and none can go in checked baggage. 25,000mAh is about 96Wh even at 3.85V, inside the 100Wh limit.

Some airlines, such as Southwest, allow only 1 power bank per person, so check your airline’s rules.

Under Japan’s rules since April 24, 2026, if you fly a Japanese airline, you may carry up to 2 power banks of 160Wh or less in the cabin. None can go in checked baggage. You may not charge a bank, or charge a device from a bank, on board.

With no outlet for 3 days, two banks fit Japanese airlines’ limit of 2. An airline that allows only 1 won’t let you carry both.

Beyond the FAA, Southwest and Japanese airlines, other rules aren’t covered here.

Picks for each situation

Candidates have 100 or more ratings, 4.0 or more stars, a seller that is Amazon or the maker, and no parallel imports.

Of the products checked here, these are the cheapest that meet the conditions (as of October 2026). If you carry only a phone and can’t charge for 3 days, combine a 20,000mAh pick with a 10,000mAh pick.

Phone only, and you can charge at the hotel each night

10,000mAh just barely covers a day of phone charging alone, with about 0.8Wh to spare.

A tablet or handheld console besides your phone, and you can charge at the hotel each night

20,000mAh with two or more output ports, so it charges two devices at once.

A laptop too, and you can charge at the hotel each night

25,000mAh with a maximum output of 65 W or more, so it can charge a laptop too.

Sold by AnkerDirect, the maker's official store.

FAQ

Is 10,000mAh enough for a 3-day trip?

Assuming 60-70% efficiency and a 0% to 100% charge, if you can charge each night and carry only a phone, 10,000mAh just barely covers a day. The margin is only about 0.8Wh.

The Pro Max’s 5,391mAh is an EU-label figure from press reports (Apple does not publish it), and the US models are unchecked, so 20,000mAh is safer for a Pro Max.

Do I need 20,000mAh for a 3-day trip?

Assuming 1 charge a day (half for tablets and laptops) at 60% efficiency, with an outlet each night, 20,000mAh covers a phone plus a tablet or console. A phone plus a laptop needs 25,000mAh.

What size power bank do you need if you can’t charge for 3 days?

Assuming 1 charge a day (half for tablets and laptops) at 60% efficiency, with no outlet for 3 days, a phone alone needs 20,000mAh and 10,000mAh banks. A tablet or console raises the need.

How many times can a 10,000mAh power bank charge my phone?

Assuming 60-70% efficiency and a 0% to 100% charge, 10,000mAh charges an iPhone 18 Pro about 1.3 to 1.8 times. Dividing 10,000mAh by its 4,056mAh battery gives 2.5, but power is lost along the way, so the real number is lower.

The 4,056mAh is an EU-label figure from press reports; Apple does not publish it.

To share one bank, add each person’s Wh and pick a size from the table. This assumes you can charge each night and each person charges a phone once a day.

If both carry only a phone, two of the largest phone here need about 41.6Wh, so 20,000mAh (at least 43.2Wh usable) is the guide.

Can I bring a power bank on a plane?

Under FAA rules as of October 2026, power banks of 100Wh or less can be carried on without airline approval. Those of 101 to 160Wh need the airline’s approval, and none can go in checked baggage.

25,000mAh is about 96Wh even at 3.85V, inside the 100Wh limit. Some airlines, such as Southwest, allow only 1 power bank per person, so check your airline’s rules.

Under Japan’s rules since April 24, 2026, if you fly a Japanese airline, you may carry up to 2 power banks of 160Wh or less in the cabin. None can go in checked baggage.

You may not charge a bank, or charge a device from a bank, on board.

Data and formulas used (for verification)

Method

Scope: current power banks from major makers and official specs of phones, tablets, handheld consoles and laptops as of 2026-10-10. No hands-on tests.
Energy needed = device battery mAh x voltage / 1000 (Wh). Usable bank energy (Wh) = rated mAh x cell voltage x conversion efficiency / 1000.
Verdict uses the unfavorable end (3.6V cell, 60% efficiency, 3.85V device battery, 0% to 100%): OK if enough there, partial if enough only at the favorable end, not enough otherwise. The iPhone 18 Pro and Pro Max battery sizes are figures from EU energy labels as reported by the press (Apple does not publish them); the US models are unchecked.
Daily charges from the bank are assumptions: 1 for phones and consoles, half for tablets and laptops. The initial full charge is not counted.
No published wireless-charging efficiency was found, so counts are calculated for wired charging.
Needed amounts are rounded up and usable amounts rounded down to one decimal, so the result never errs toward “enough”.

Assumptions

60-70% of the energy in Wh (MagGo FAQ cites 30-45% loss). The makers’ 6,000-7,000mAh counts at USB 5V (30-35Wh), a different basis, so this calculation is stricter; it varies with the model, temperature and cable.: 0.65 (0.6–0.7)
Nominal cell voltage, held between 3.7V (Anker) and 3.85V (Anker MagGo label); the lower 3.6V side is the conservative end.: 3.7 (3.6–3.85)
No primary source gives a number for wireless vs wired loss, so the article only says wireless is less efficient.: “not found” (0–0)
Nominal voltage of the device battery, used to turn mAh into Wh for devices that publish only mAh (iPhone, Galaxy, Switch 2). Same 3.6-3.85V span as the bank cell; the high side (more Wh) is the conservative end.: 3.7 (3.6–3.85)
How much of the battery one charge fills: 20% to 100% (0.8) up to 0% to 100% (1.0). 1.0 is the conservative end.: 0.8 (0.8–1)
Phone charges per day from the bank. Maps, photos and payments drain more on a trip, so the article assumes the higher side, 1 charge; lighter use is half.: 1 (0.5–1)
Tablet charges per day. Half a charge for videos on the move is the premise; all-day use is 1.: 0.5 (0.5–1)
Handheld console charges per day. Consoles drain in a few hours, so 1 charge is the premise; light play is half.: 1 (0.5–1)
Laptop charges per day. Half a charge for some work on the trip is the premise; all-day use is 1.: 0.5 (0.5–1)

Official values used

Item	Value	Source	Checked
cell_nominal_voltage_v	3.7 V	service.anker.com ‘which typically operate at a voltage of around 3.7V’	2026-10-10
usb_output_voltage_v	5 V	service.anker.com ‘the rated output capacity of the power bank is often calculated at a standard o	2026-10-10
wh_formula	Wh = mAh x V / 1000 formula	service.anker.com ‘Energy in Watt-hours = Capacity in Ampere-hours × Voltage in Volts’	2026-10-10
wh_10000mah_at_3_7v	37 Wh	service.anker.com ‘10Ah × 3.7V = 37Wh’	2026-10-10
usb_ceiling_10000mah_mah	7400 mAh	service.anker.com ‘37Wh / 5V = 7.4Ah = 7,400mAh’	2026-10-10
circuit_efficiency_anker	0.82 ratio	service.anker.com ‘typically around 82% efficiency’	2026-10-10
circuit_efficiency_anker_uk_range	0.8–0.9 ratio	anker.com ‘operate at around 80–90% efficiency’	2026-10-10
usable_of_10000mah_wired_mah	6000–7000 mAh	anker.com ‘a fully charged Anker MagGo Power Bank (10K) offers an estimated 6,000mAh to 7,	2026-10-10
anker_uk_charge_example_iphone14pro	2 times	anker.com ‘iPhone 14 Pro (3,200mAh) → ~ 2 full charges.‘	2026-10-10
wireless_5w_when_usbc_in_use	5 W	anker.com ‘If the USB-C port is charging a device, wireless charging output will be reduce	2026-10-10
wireless_less_efficient_elecom	Wireless charging is less efficient than wired (it generates power through a magnetic field)	elecom.co.jp	2026-10-10
qi2_alignment_efficiency_wpc	Qi2 magnet alignment improves efficiency (no numeric loss given) text	wirelesspowerconsortium.com ‘Qi2 offers greater charging efficiency, faster charging, and greater convenienc	2026-10-10
iphone18pro_wireless_max_w	25 W	apple.com ‘MagSafe wireless charging up to 25W’	2026-10-10
iphone18pro_battery_mah_eu	4056 mAh	tomsguide.com ‘iPhone 18 Pro will have a 4,056 mAh battery’	2026-10-10
iphone18promax_battery_mah_eu	5391 mAh	tomsguide.com ‘5,391 mAh capacity in the EU’	2026-10-10
iphone17promax_battery_mah_eu	4823 mAh	macrumors.com ‘iPhone 17 Pro Max’s 4,823 mAh’	2026-10-10
iphone18pro_video_hours	36 hours	apple.com ‘up to 36 hours of video playback on iPhone 18 Pro’	2026-10-10
iphone18promax_video_hours	45 hours	apple.com ‘up to 45 hours of video playback on iPhone 18 Pro Max’	2026-10-10
iphone18pro_release_us	2026-09-18 date	apple.com ‘Pre-orders begin Saturday, September 12, with availability beginning Friday, Se	2026-10-10
macbook_air_13_wh	53.8 Wh	apple.com ‘Built-in 53.8‑watt‑hour lithium‑polymer battery’	2026-10-10
macbook_air_15_wh	66.5 Wh	apple.com ‘Built-in 66.5‑watt‑hour lithium‑polymer battery’	2026-10-10
ipad_air_11_wh	28.93 Wh	apple.com ‘Built‐in 28.93‐watt‐hour rechargeable lithium‑polymer battery’	2026-10-10
ipad_air_13_wh	36.59 Wh	apple.com ‘Built‐in 36.59‐watt‐hour rechargeable lithium‑polymer battery’	2026-10-10
switch2_battery_mah	5220 mAh	nintendo.com ‘Lithium-ion battery/5220mAh’	2026-10-10
steamdeck_oled_whr	50 Wh	steamdeck.com ‘50Whr battery’	2026-10-10
galaxy_s26_ultra_typical_mah	5000 mAh	samsung.com ‘5000 mAh ultra-long lasting battery for all-day performance’	2026-10-10
faa_li_ion_limit_wh	100 Wh	faa.gov ‘Lithium ion (rechargeable) batteries are limited to a rating of 100 watt hours	2026-10-10
faa_101_160_wh_approval	101–160 Wh	faa.gov ‘With airline approval, passengers may also carry up to two spare larger lithium	2026-10-10
faa_carry_on_only	carry-on only text	faa.gov ‘Spare (uninstalled) lithium ion and lithium metal batteries, including power ba	2026-10-10
faa_airline_stricter	airlines may set stricter limits text	faa.gov ‘Many airlines, both domestic and international, may have stricter quantity and	2026-10-10
southwest_power_bank_limit	1 units	support.southwest.com ‘each Customer is limited to one power bank (effective April 20, 2026)‘	2026-10-10
mlit_carry_on_limit	2 units, 160 Wh or less each (from 2026-04-24)	mlit.go.jp	2026-10-10
mlit_in_cabin_charging_ban	No charging the power bank or charging other devices from it in flight	mlit.go.jp	2026-10-10
mlit_2025_overhead_bin	From 2025-07-08, do not stow power banks in overhead bins; keep them where you can see them	mlit.go.jp	2026-10-10
mlit_checked_baggage_ban	Checked baggage is prohibited	ana.co.jp	2026-10-10
ana_rule_2026	ANA: 2 per person, 160 Wh or less, no checked baggage, no in-flight charging	ana.co.jp	2026-10-10
jal_rule_2026	JAL: 2 per person, 160 Wh or less	ontrip.jal.co.jp	2026-10-10
jr_n700s_outlets	On the N700S every seat has an outlet; on other cars only window seats and the front/rear seats	railway.jr-central.co.jp	2026-10-10
amtrak_coach_outlet	Each unreserved coach seat has a 120V outlet	amtrak.com ‘There is also a fold-down tray, individual reading light, and a 120v electric o	2026-10-10
sony_np_fz100_wh	16.4 Wh	sony.jp	2026-10-10
sony_np_fz100_mah	2280 mAh	sony.jp	2026-10-10
gopro_enduro_hero13_mah	1900 mAh	gopro.com ‘Advanced 1900 mAh battery delivers longer runtimes in a wide variety of tempera	2026-10-10
canon_lp_e6p_mah	2130 mAh	faq.canon.jp	2026-10-10
apple_watch_series_11_battery_hours	24 hours	apple.com Apple Watch Series 11: ‘Up to 24 hours’	2026-10-10
apple_watch_ultra_3_battery_hours	42 hours	apple.com Apple Watch Ultra 3: ‘Up to 42 hours’	2026-10-10
airpods_pro_3_case_listening_hours	24 hours	apple.com ‘Up to 24 hours of listening time with Active Noise Cancellation enabled’	2026-10-10
pixel_10_pro_battery_mah	4870 mAh	store.google.com	2026-10-10

Calculate the power bank size for your trip: quick reference

The calculator in this article shows the result and a matching product for your own settings. Settings can be set in the URL; each link below opens the calculator with those settings.

Settings	Result	Calculator
iPhone 18 Pro, No tablet, No console, No laptop, Can charge at the hotel	✓ 10,000mAh is enough. You need 11.6–15.7Wh (1 day, energy into the device batteries). With hotel charging each night, it counts 1 day whatever the trip length. Charges per day from the bank: phone 1. Wireless charging is less efficient than a cable, so you get fewer charges than shown	Open with these settings
iPhone 18 Pro Max, No tablet, No console, No laptop, Can charge at the hotel	✓ 10,000mAh is enough. You need 15.5–20.8Wh (1 day, energy into the device batteries). With hotel charging each night, it counts 1 day whatever the trip length. Charges per day from the bank: phone 1. Wireless charging is less efficient than a cable, so you get fewer charges than shown	Open with these settings
iPhone 18 Pro, No tablet, No console, No laptop, 3 days, No outlet	✓ 25,000mAh is enough. You need 35–46.9Wh (3 days, energy into the device batteries). Charges per day from the bank: phone 1. 25,000mAh stays under 100Wh even at 3.85V, so no airline approval is needed to carry it on. Wireless charging is less efficient than a cable, so you get fewer charges than shown	Open with these settings
iPhone 18 Pro, iPad Air 11-inch, No console, No laptop, Can charge at the hotel	✓ 20,000mAh is enough. You need 23.2–30.1Wh (1 day, energy into the device batteries). With hotel charging each night, it counts 1 day whatever the trip length. Charges per day from the bank: phone 1, tablet 0.5. Wireless charging is less efficient than a cable, so you get fewer charges than shown	Open with these settings
iPhone 18 Pro, iPad Air 11-inch, No console, No laptop, 3 days, No outlet	✕ Even 25,000mAh isn’t enough. You need 69.7–90.3Wh (3 days, energy into the device batteries). Charges per day from the bank: phone 1, tablet 0.5. If 1 bank isn’t enough, charge on the move or carry 2 banks (Japanese airlines allow 2 in the cabin, each 160Wh or less). Wireless charging is less efficient than a cable, so you get fewer charges than shown	Open with these settings
iPhone 18 Pro, No tablet, Nintendo Switch 2, No laptop, Can charge at the hotel	✓ 20,000mAh is enough. You need 26.7–35.8Wh (1 day, energy into the device batteries). With hotel charging each night, it counts 1 day whatever the trip length. Charges per day from the bank: phone 1, console 1. Wireless charging is less efficient than a cable, so you get fewer charges than shown	Open with these settings
iPhone 18 Pro, No tablet, Nintendo Switch 2, No laptop, 3 days, No outlet	✕ Even 25,000mAh isn’t enough. You need 80.1–107.2Wh (3 days, energy into the device batteries). Charges per day from the bank: phone 1, console 1. If 1 bank isn’t enough, charge on the move or carry 2 banks (Japanese airlines allow 2 in the cabin, each 160Wh or less). Wireless charging is less efficient than a cable, so you get fewer charges than shown	Open with these settings
iPhone 18 Pro, No tablet, No console, MacBook Air 13-inch, Can charge at the hotel	✓ 20,000mAh is enough. You need 33.2–42.6Wh (1 day, energy into the device batteries). With hotel charging each night, it counts 1 day whatever the trip length. Charges per day from the bank: phone 1, laptop 0.5. To charge a laptop, pick a bank with a high maximum output. Wireless charging is less efficient than a cable, so you get fewer charges than shown	Open with these settings
iPhone 18 Pro, No tablet, No console, MacBook Air 13-inch, 3 days, No outlet	✕ Even 25,000mAh isn’t enough. You need 99.6–127.6Wh (3 days, energy into the device batteries). Charges per day from the bank: phone 1, laptop 0.5. If 1 bank isn’t enough, charge on the move or carry 2 banks (Japanese airlines allow 2 in the cabin, each 160Wh or less). To charge a laptop, pick a bank with a high maximum output. Wireless charging is less efficient than a cable, so you get fewer charges than shown	Open with these settings
Galaxy S26 Ultra, No tablet, No console, No laptop, Can charge at the hotel	✓ 10,000mAh is enough. You need 14.4–19.3Wh (1 day, energy into the device batteries). With hotel charging each night, it counts 1 day whatever the trip length. Charges per day from the bank: phone 1. Wireless charging is less efficient than a cable, so you get fewer charges than shown	Open with these settings
iPhone 18 Pro, iPad Air 11-inch, Nintendo Switch 2, No laptop, Can charge at the hotel	✓ 25,000mAh is enough. You need 38.2–50.2Wh (1 day, energy into the device batteries). With hotel charging each night, it counts 1 day whatever the trip length. Charges per day from the bank: phone 1, tablet 0.5, console 1. 25,000mAh stays under 100Wh even at 3.85V, so no airline approval is needed to carry it on. Wireless charging is less efficient than a cable, so you get fewer charges than shown	Open with these settings
iPhone 18 Pro, No tablet, No console, No laptop, 1 day, No outlet	✓ 10,000mAh is enough. You need 11.6–15.7Wh (1 day, energy into the device batteries). Charges per day from the bank: phone 1. Wireless charging is less efficient than a cable, so you get fewer charges than shown	Open with these settings

What Monitor Does VALORANT Need? 144Hz vs 240Hz vs 360Hz

Sat, 10 Oct 2026 00:00:00 GMT

In this guide’s method, pick a VALORANT monitor by your average in-match fps (frames the game draws per second). For refresh rate (how often the monitor redraws per second), 144Hz is enough at an average of 144 fps or less.

Per Riot’s official pages, the top spec target is 144 fps, and the in-game Stats display shows your fps. A monitor is fully used only when your average fps reaches its refresh rate: 240 fps for 240Hz, 300 fps for 300Hz, 360 fps for 360Hz.

Checked as of October 2026. No monitor was tested in person.

New frames on screen, by average fps and Hz

Assumption: the math assumes a steady fps. Real fps moves up and down, so an average is a guide. Input lag is not included.

“Step” means the common refresh rates 144, 165, 180, 200, 240, 280, 300 and 360Hz.

Average fps	144Hz	240Hz	360Hz	Lowest step that shows the most new frames
60 fps	60	60	60	144Hz
144 fps	144	144	144	144Hz
200 fps	144	200	200	200Hz
240 fps	144	240	240	240Hz
300 fps	144	240	300	300Hz
360 fps	144	240	360	360Hz
400 fps	144	240	360	360Hz (shows 360 of 400)

In this guide’s grouping, casual play is 144–200Hz, ranked play is 240–280Hz, and competitive play is 300Hz and up. This isn’t an official standard.

What the pros use

According to ProSettings, a third-party site, every VALORANT pro it analyzed uses a monitor of at least 240Hz (article updated 2026-01-21). That’s a pro trend, so a normal player should go by their own measured average fps.

In its table updated 2026-10-08, the most-used model among 700 players (24.5-inch, 1080p, 360Hz, TN) has 23.29%. It isn’t a pick: Amazon.com searches on 2026-10-09 found no listing.

How to measure your fps in VALORANT

In this guide’s method, go by your average fps in normal matches. Stats shows the current fps, so watch where it settles. For a number, use FrameView, which shows Average FPS.

Per Riot’s official pages, Settings > Video > Stats shows Client FPS and other stats as numbers or graphs (2021 article). It doesn’t say it shows an average.

Per NVIDIA’s FrameView user guide, FrameView shows Average FPS on screen. Press Scroll Lock to start and stop a capture and it shows a summary, on any GPU brand.

VALORANT’s spec table (dated 2026-05-21) has three targets, 30, 60 and 144 fps, and lists no spec for 240 or 360 fps. For the 144 fps (high-end) tier, Riot lists the GTX 1050 Ti, Radeon R7 370 and Arc A310 as GPUs.

There’s no per-GPU VALORANT fps with test conditions. Fps depends on your PC, so measure it yourself.

In Riot’s crash-troubleshooting article, if the game crashes, set the frame cap (LIMIT FPS ALWAYS) below your average in-match fps.

VALORANT fps-to-monitor calculator

Pick your measured VALORANT fps (or don't know), how seriously you play and your eye-to-screen distance to see the refresh rate that adds frames, the size and resolution, plus a matching pick.

Where 24 inches ends and 27 begins

Eye-to-screen, a 23.8-inch meets the target from 50 cm, a 24.5-inch from 51.59 cm and a 27-inch from 56.73 cm. Within 1 cm past each distance it’s borderline.

In this guide’s grouping, for competitive play, put fps first and use 1080p. A 27-inch at 1080p has pixels 1.13× the size of a 23.8-inch’s, so pick 24 to 25 inches.

How to choose between 1080p and 1440p

In this guide’s grouping, go 1440p only with a 27-inch (56.73 cm or more eye-to-screen, not competitive play). Your average fps at 1440p must still reach the monitor’s refresh rate. Within 1 cm past 56.73 cm, it’s borderline.

1440p has 1.78× the pixels of 1080p to draw. How much fps drops depends on the PC, so measure your average fps at 1440p.

Response time and panel type

You don’t need 1 ms. A rated GtG shorter than one frame (4.17 ms at 240Hz) is enough. GtG is how long a pixel takes to change from one shade to another.

Per the makers’ official specs, MSI’s MP243L lists 1 ms for MPRT and 4 ms for GtG. MPRT is how long a pixel stays visible, so read the GtG line.

In this guide’s grouping, with ASUS’s official specs as the example, panels aren’t ranked here. Choose by refresh rate and rated GtG. Its OLED example, a 27-inch 4K model, rates 0.03 ms GtG, but 24-to-25-inch OLEDs aren’t covered.

Per ASUS’s official specs, among its 280Hz models, the TN and the Fast IPS both rate 1 ms (GTG). Viewing angles are 170°/160° for the TN and 178°/178° for the IPS.

Variable refresh rate and Reflex

In this guide’s grouping, picks must support variable refresh (G-SYNC Compatible, FreeSync or similar). When fps falls below the monitor’s maximum Hz, variable refresh matches the refresh to your fps. It helps in heavy scenes, or when a pick’s Hz is close to your average fps.

In NVIDIA’s own description, G-SYNC Compatible monitors are ones NVIDIA has validated for basic variable refresh rate. They don’t use NVIDIA’s processor.

In AMD’s own description, FreeSync matches the display’s refresh to the GPU’s frame rate to reduce artifacts such as tearing. If your fps regularly exceeds the refresh rate, turn on V-Sync or cap fps.

Reflex cuts PC latency by syncing GPU and CPU work, and VALORANT supports it. It’s a PC-side feature, so it isn’t a monitor condition.

Picks by situation

In this guide’s grouping, picks come from whichever band is lower: the one your fps reaches, or the one for how seriously you play. A 24.5-inch meets the target from 51.59 cm, while at 50 cm a 23.8-inch is the target. Right at 51.59 cm, it’s borderline.

Rules: sold new by Amazon or the maker, not an import, at least 100 ratings, 4 stars or more, and no warning from Sakura Checker. Check prices on the store.

Casual play

It’s a 23.8-inch 1080p monitor, up to 144Hz, Adaptive-Sync. Borderline: its 144Hz maximum is the casual band’s floor. The MSI PRO MP243L E14 is in MSI’s business and productivity line. Its inputs are one HDMI 2.0 (144Hz) and one VGA, with no DisplayPort.

Serious ranked play

It’s a 23.8-inch 1080p IPS monitor, up to 240Hz, FreeSync Premium. Also borderline: its 240Hz maximum is the ranked band’s floor. No Dell US page was found, so these specs, including the 240Hz maximum, come from the UK page.

Competitive play, 300Hz+ over DisplayPort, average 300 fps+, from 51.59 cm (20.4 in)

It’s a 24.5-inch 1080p monitor, up to 310Hz, Adaptive-Sync. Also borderline, just above the 300Hz floor. The VG259QMR5A’s 310Hz is over DisplayPort; over HDMI it tops out at 240Hz.

FAQ

Is 144Hz enough for VALORANT? Does a high-Hz monitor help if my fps is low?

In this calculation, if your fps stays at 144 fps or below, 240Hz or 360Hz adds no new frames, so 144Hz is enough. At 200 fps, a 144Hz monitor shows 144 new frames a second and a 360Hz one shows 200.

You only ever see the smaller of your fps and the refresh rate.

What’s the difference between 240Hz and 360Hz? Do you need 360Hz?

Working from your measured average fps, 360Hz adds frames only when your average fps is above 240 fps, up to 360 fps. Up to 280 fps, a 280Hz monitor adds just as many, and up to 300 fps, a 300Hz one does.

Pros use 360Hz. Should regular players copy them?

Should you get a 24-inch or a 27-inch monitor?

In this calculation, from 50 cm to 56.73 cm eye-to-screen, pick 24 to 25 inches. From 56.73 cm, a 27-inch is also an option.

Within 1 cm past that, it’s borderline. Closer than 51.59 cm, a 23.8-inch rather than a 24.5-inch is the target.

Right at 51.59 cm the 24.5-inch is borderline.

1080p or 1440p?

In this guide’s grouping, go 1440p only with a 27-inch (56.73 cm or more eye-to-screen, not competitive play). Your average fps at 1440p must still reach the monitor’s refresh rate.

Within 1 cm past 56.73 cm, it’s borderline. 1440p has 1.78× the pixels of 1080p to draw.

How much fps drops depends on the PC, so measure your average fps at 1440p.

Do you need G-SYNC, FreeSync or NVIDIA Reflex? What if fps goes above the Hz?

It helps in heavy scenes, or when a pick’s Hz is close to your average fps. In AMD’s own description, FreeSync matches the display’s refresh to the GPU’s frame rate to reduce artifacts such as tearing.

If your fps regularly exceeds the refresh rate, turn on V-Sync or cap fps. In NVIDIA’s own description, Reflex cuts PC latency by syncing GPU and CPU work, and VALORANT supports it.

It’s a PC-side feature, so it isn’t a monitor condition.

Data and formulas used (for verification)

Method

New frames shown per second = min(your VALORANT fps, the monitor’s refresh rate), the same formula as the sibling 144Hz vs 240Hz guide. Gain from stepping up = min(fps, higher Hz) − min(fps, lower Hz). Going 144Hz → 240Hz adds only the fps above 144Hz, up to 240Hz. Going 240Hz → 360Hz adds only the fps above 240Hz, up to 360Hz. Assumes a steady fps; input lag and feel are not included.
Frame time (ms) = 1000 ÷ Hz: 6.94 ms at 144Hz, 4.17 ms at 240Hz, 2.78 ms at 360Hz.
Riot’s official spec table (dated 2026-05-21) targets 30 fps, 60 fps and 144 fps, and lists no spec for 240 or 360 fps. Riot publishes no per-GPU VALORANT fps with test conditions, so none are used. Readers measure their own (Settings > Video > Stats (shows performance stats or graphs)).
The refresh-rate bands are this guide’s own split, not an official standard: casual = 144–200Hz, ranked = 240–280Hz, competitive = 300Hz and up. The calculator calls the lowest step that shows all your fps (144, 165, 180, 200, 240, 280, 300, 360Hz) the Hz that adds frames, and picks come from the lower of that step’s band and your play style’s band. If you don’t know your fps, it shows how to measure and uses the band for Riot’s high-end target (144 fps).
Size: target distance = 50 cm (the low end of OSHA’s range) × the screen’s viewable width ÷ the 23.8-inch width (52.704 cm). A 24.5-inch (54.3744 cm) meets it from 51.59 cm (20.4 in) and a 27-inch (59.789 cm) from 56.73 cm (22.4 in), rounded up to two decimals, with no room right at that distance. Same formula as the sibling monitor-size guide. Desk distances 50 cm (19.7 in)–90 cm (35.4 in) in 5 cm steps were all checked.
Resolution: pixel size ratio = pixel pitch ÷ the 23.8-inch 1080p pitch (0.2745 mm): 1.13× for 27-inch 1080p (0.3114 mm) and 0.85× for 27-inch 1440p (0.2331 mm). 1440p has 1.78× the pixels of 1080p. How much fps drops depends on the PC, so no number is used: switch resolution and measure with Stats.
Pro usage is from ProSettings (third-party, updated 2026-10-08, 700 players). Only the top model’s share and the page’s own wording are used. Overall shares, such as all 360Hz monitors, aren’t published, so none are given.
Variable refresh (G-SYNC Compatible, FreeSync, LFC) and Reflex are described in NVIDIA’s and AMD’s own terms. Latency numbers, cross-brand response-time or feel comparisons, and prices are not covered.
Picks are new Amazon listings with at least 100 ratings and 4 stars, no sakura warning, not an import, and sold by Amazon or a seller confirmed as the maker’s own store, whose official specs meet each band’s condition (23.8 to 24.5 inch, 1080p, variable refresh, max Hz), cheapest with shipping first. Rating counts pooled with other models are not counted. Dell SE2426HG specs come from Dell’s UK page; no US page was found. No unit was tested in person.

Assumptions

70 cm: 70 (50–90)

Official values used

Item	Value	Source	Checked
Riot minimum-spec target fps	30 fps	support.riotgames.com	2026-10-09
Riot recommended-spec target fps	60 fps	support.riotgames.com Rec Specs (60 FPS)	2026-10-09
Riot high-end-spec target fps	144 fps	support.riotgames.com	2026-10-09
Riot high-end (144 FPS) CPU	Intel i5-9400F 2.90GHz / AMD Ryzen 5 2600X	support.riotgames.com	2026-10-09
Riot high-end (144 FPS) GPU	NVIDIA GTX 1050 Ti / AMD Radeon R7 370 / Intel Arc A310	support.riotgames.com High-end Specs GPU	2026-10-09
Riot recommended (60 FPS) CPU	Intel i3-4150 / AMD Ryzen 3 1200	support.riotgames.com Rec Specs CPU	2026-10-09
Riot recommended (60 FPS) GPU	NVIDIA GeForce GT 730 / AMD Radeon R7 240	support.riotgames.com	2026-10-09
Riot minimum (30 FPS) CPU	Intel i3-540 / AMD Athlon 200GE	support.riotgames.com Min Specs CPU	2026-10-09
Riot minimum (30 FPS) GPU	Intel HD 4000 / AMD Radeon R5 220	support.riotgames.com Min Specs GPU	2026-10-09
Riot RAM requirement	4 GB	support.riotgames.com	2026-10-09
Date shown on Riot spec page	2026-05-21	support.riotgames.com	2026-10-09
Main 16:9 resolutions VALORANT supports	1920x1080 / 2560x1440 / 3840x2160 (16:9, which Riot recommends)	support.riotgames.com	2026-10-09
Where Riot says to set the FPS cap	ESC - VIDEO - LIMIT FPS ALWAYS; Riot suggests a value lower than the average FPS you get while playing	support.riotgames.com Set a framerate cap in settings … ESC - VIDEO - LIMIT FPS ALWAYS. Try setting	2026-10-09
Where Riot says to show performance stats	Settings Video Stats (shows performance stats or graphs)	playvalorant.com	2026-10-09
Riot on what limits FPS	Lower-spec machines tend to be GPU-bound; mid-to-high-spec machines tend to be CPU-bound	playvalorant.com lower spec machines tend to be GPU bound …, whereas mid to high spec machines	2026-10-09
Riot: refresh rate vs FPS	Refresh rate is how many times the display is illuminated per second; FPS is how often images can change. They are different	support.riotgames.com refresh rates determine how many times your display is illuminated per second. T	2026-10-09
Riot Multithreaded Rendering requirements	Multithreaded Rendering is for devices with 8 GB RAM, 2 GB VRAM and a CPU with at least 8 cores (physical and virtual)	support.riotgames.com minimum specs: 8 GB RAM, 2 GB VRAM, CPU with at least 8 cores (physical and virt	2026-10-09
ProSettings sample (VALORANT pros)	700 players	prosettings.net	2026-10-09
ProSettings: most-used monitor	ZOWIE XL2566K (24.5-inch, 1920x1080, 360 Hz, TN) is used by 23.29% of players (163)	prosettings.net The most used monitors by VALORANT players: Size 24.5, 1920x1080, 360, TN, ZOWIE	2026-10-09
ProSettings: pro refresh rates	ProSettings writes that all analyzed VALORANT pros use a monitor capable of at least 240 fps, and that 360 Hz is clearly the current standard in top leagues	prosettings.net All of our analyzed VALORANT professionals use a monitor that’s capable of displ	2026-10-09
ProSettings: pro resolution	ProSettings writes that the vast majority of pro monitors are 1080p and almost no pro is on a resolution above 1920x1080	prosettings.net the vast majority of the monitors in the pro scene are 1080p … Almost no profe	2026-10-09
ProSettings: ZOWIE brand share	74.29 %	prosettings.net Top monitor brands in VALORANT, Based on 700 players, Last updated 8 October, 20	2026-10-09
NVIDIA’s G-SYNC Compatible description	tear-free, stutter-free	nvidia.com	2026-10-09
AMD’s FreeSync description	The display refresh rate is synchronized with the framerate of FreeSync-compatible graphics cards, reducing or eliminating visual artifacts	amd.com	2026-10-09
AMD: when FPS exceeds refresh rate	If game FPS regularly exceeds the display’s refresh rate, turn on V-Sync or cap the framerate for the best tear-free visual experience	amd.com If your game FPS is regularly exceeding your display’s refresh rate, you can tur	2026-10-09
NVIDIA Reflex description	Reflex Low Latency reduces PC latency through precise synchronization of rendering across GPU and CPU	nvidia.com	2026-10-09
NVIDIA Reflex and VALORANT	VALORANT is listed on NVIDIA’s Reflex-compatible products list	nvidia.com Reflex-Compatible Products from Partners: Valorant (games list)	2026-10-09
AMD LFC description	When framerate drops below the display’s minimum refresh rate, frames are duplicated so the display stays in its range (example: 60-144 Hz display, 40 FPS shown at 80 Hz)	amd.com FAQ: What is Low Framerate Compensation? … 40 FPS … doubling them … 80 Hz	2026-10-09
ASUS VG258QM (TN) maker claim	TN panel, 1 ms (GTG), up to 280 Hz, 170/160 degree viewing angle	asus.com Panel Type: TN / Viewing Angle (CR≧10, H/V): 170°/160° / 1ms(GTG) / 280Hz	2026-10-07
ASUS VG259QM (IPS) maker claim	Fast IPS, 24.5-inch, 1920x1080, 1 ms (GTG), up to 280 Hz, 178/178 degree viewing angle	asus.com	2026-10-09
ASUS VG259QMR5A (IPS) maker claim	Fast IPS, 24.5-inch, 1920x1080, 1 ms (GTG) / 0.3 ms (min), up to 310 Hz	asus.com Panel Type: Fast IPS / Response Time: 1ms(GTG), 0.3ms(min.) / Refresh Rate (Max)	2026-10-09
ASUS ROG Swift OLED PG27UCDM GTG claim (OLED example)	0.03 ms	rog.asus.com 0.03ms gray-to-gray (GTG) response time	2026-10-07
BenQ: MPRT floor at 144 Hz	6.9 ms	benq.com	2026-10-07
ViewSonic: frame time at 240 Hz	4.17 ms	viewsonic.com	2026-10-07
fps-to-Hz model (calculation premise)	New frames shown per second = min(fps, refresh rate)	source	2026-10-09
Lower end of OSHA’s preferred range	50 cm	osha.gov 20 to 40 inches (50 to 100 cm)	2026-10-07
Upper end of OSHA’s preferred range	100 cm	osha.gov 20 to 40 inches (50 to 100 cm)	2026-10-07
23.8-inch viewable width	52.704 cm	dl.dell.com S2425H Active Area Horizontal 527.04 mm p.16	2026-10-07
27-inch viewable width	59.789 cm	dl.dell.com S2725H Active Area Horizontal 597.89 mm p.16	2026-10-07
23.8-inch 1080p pixel pitch	0.2745 mm	dl.dell.com Dell S2425H pixel pitch	2026-10-07
27-inch 1080p pixel pitch	0.3114 mm	dl.dell.com Dell S2725H pixel pitch	2026-10-07
27-inch 1440p pixel pitch	0.2331 mm	dl.dell.com Dell S2725DS pixel pitch	2026-10-07
24.5-inch viewable width	54.3744 cm	asus.com Panel Size (inch) : 24.5 / Display Viewing Area (H x V) : 543.744 x 302.616 mm /	2026-10-09
ASUS VG259QM max over DisplayPort	280 Hz	asus.com Digital Signal Frequency : DisplayPort: 255~~255 KHz (H) / 48~~280 Hz (V) / HDMI:	2026-10-09
ASUS VG259QM max over HDMI	240 Hz	asus.com HDMI: 30~~255 KHz (H) / 48~~240 Hz (V)	2026-10-09
ASUS VG259QMR5A max over DisplayPort	310 Hz	asus.com Refresh Rate (Max) : 310Hz / Digital Signal Frequency : DisplayPort: 344~344 KHz	2026-10-09
ASUS VG259QMR5A max over HDMI	240 Hz	asus.com HDMI: 31~~255 KHz (H) / 48~~240 Hz (V)	2026-10-09
MSI PRO MP243L E14 inputs and category	One HDMI 2.0 (FHD@144Hz) and one D-Sub (VGA); no DisplayPort. MSI files it under Business & Productivity monitors	us.msi.com	2026-10-09
MSI PRO MP243L E14 max refresh rate	144 Hz	us.msi.com Refresh Rate 144Hz / HDMI 2.0 (FHD@144Hz)	2026-10-09
Dell SE2426HG max refresh rate (Dell UK page)	240 Hz	dell.com	2026-10-09
ProSettings table update date	2026-10-08	prosettings.net Top monitor brands in VALORANT, Based on 700 players, Last updated 8 October, 20	2026-10-09
ProSettings: share of the most-used monitor (ZOWIE XL2566K)	23.29 %	prosettings.net ZOWIE XL2566K, 23.29% of players (163)	2026-10-09
ProSettings: lowest refresh rate among analyzed pros	240 Hz	prosettings.net All of our analyzed VALORANT professionals use a monitor that’s capable of displ	2026-10-09
ProSettings article last-updated date	2026-01-21	prosettings.net	2026-10-09
MSI PRO MP243L E14 response time as listed	1ms (MPRT) / 4ms (GtG)	us.msi.com	2026-10-09
How NVIDIA FrameView measures average fps	FrameView: overlay shows Average FPS; Scroll Lock / F10 starts and stops a benchmark capture; works across GPU vendors	images.nvidia.com	2026-10-09
NVIDIA app overlay fps display	NVIDIA app overlay: frame rates, clock speeds, GPU temperatures; Alt+R toggles; no average stated	nvidia.custhelp.com	2026-10-09
AMD Software fps metrics	AMD Software Performance Metrics: Frame Rate, Frame Time, 99th% FPS; logging; no average stated	amd.com	2026-10-09
BenQ: what GtG means	GtG: how long a pixel takes to change from one grey value to another	benq.com	2026-10-09
BenQ: what MPRT means	MPRT: how long a pixel remains apparent or visible on the screen	benq.com	2026-10-09

VALORANT fps-to-monitor calculator: quick reference

The calculator in this article shows the result for your own settings. Settings can be set in the URL; each link below opens the calculator with those settings.

Settings	Result	Calculator
Don’t know, Casual, 70 cm (28 in)	△ Measure your fps first. How to measure average fps (Stats and FrameView). Until you measure, start from a 144Hz monitor, matching Riot’s high-end target of 144 fps. Pick band: 144–200Hz. Size: 24–25 inch or 27-inch (27-inch target 56.73 cm or more; 13.27 cm to spare). Resolution: 1080p at 24–25 inch. On a 27-inch, 1440p (pixels 0.85× a 23.8-inch 1080p). 27-inch 1080p pixels are 1.13×, so coarser. 1440p has 1.78× the pixels to draw, so switch resolution, measure with Stats, and use it only if fps still reaches your target. The pick tops out at 144Hz, borderline at the low end of the band	Open with these settings
120 fps, Casual, 70 cm (28 in)	✓ A 144Hz monitor shows all your fps. Refresh rate that adds frames: 144Hz class (at 120 fps, 120 new frames per second). Going above 144Hz adds no new frames. Pick band: 144–200Hz. Size: 24–25 inch or 27-inch (27-inch target 56.73 cm or more; 13.27 cm to spare). Resolution: 1080p at 24–25 inch. On a 27-inch, 1440p (pixels 0.85× a 23.8-inch 1080p). 27-inch 1080p pixels are 1.13×, so coarser. 1440p has 1.78× the pixels to draw, so switch resolution, measure with Stats, and use it only if fps still reaches your target. The pick tops out at 144Hz, borderline at the low end of the band	Open with these settings
144 fps, Casual, 70 cm (28 in)	✓ A 144Hz monitor shows all your fps. Refresh rate that adds frames: 144Hz class (at 144 fps, 144 new frames per second). Going above 144Hz adds no new frames. Pick band: 144–200Hz. Size: 24–25 inch or 27-inch (27-inch target 56.73 cm or more; 13.27 cm to spare). Resolution: 1080p at 24–25 inch. On a 27-inch, 1440p (pixels 0.85× a 23.8-inch 1080p). 27-inch 1080p pixels are 1.13×, so coarser. 1440p has 1.78× the pixels to draw, so switch resolution, measure with Stats, and use it only if fps still reaches your target. The pick tops out at 144Hz, borderline at the low end of the band	Open with these settings
200 fps, Casual, 70 cm (28 in)	✓ A 200Hz monitor shows all your fps. Refresh rate that adds frames: 200Hz class (at 200 fps, 200 new frames per second). Going from 144Hz to 200Hz adds 56 frames per second. Going above 200Hz adds no new frames. Pick band: 144–200Hz. Size: 24–25 inch or 27-inch (27-inch target 56.73 cm or more; 13.27 cm to spare). Resolution: 1080p at 24–25 inch. On a 27-inch, 1440p (pixels 0.85× a 23.8-inch 1080p). 27-inch 1080p pixels are 1.13×, so coarser. 1440p has 1.78× the pixels to draw, so switch resolution, measure with Stats, and use it only if fps still reaches your target. The pick tops out at 144Hz, so it shows up to 144 of your 200 fps	Open with these settings
240 fps, Casual, 70 cm (28 in)	✓ For how you play, 144–200Hz is enough (240Hz would add frames). Refresh rate that adds frames: 240Hz class (at 240 fps, 240 new frames per second). Going from 144Hz to 240Hz adds 96 frames per second. Going above 240Hz adds no new frames. Picks stay within your play style’s band (144–200Hz), this guide’s own split. Pick band: 144–200Hz. Size: 24–25 inch or 27-inch (27-inch target 56.73 cm or more; 13.27 cm to spare). Resolution: 1080p at 24–25 inch. On a 27-inch, 1440p (pixels 0.85× a 23.8-inch 1080p). 27-inch 1080p pixels are 1.13×, so coarser. 1440p has 1.78× the pixels to draw, so switch resolution, measure with Stats, and use it only if fps still reaches your target. The pick tops out at 144Hz, so it shows up to 144 of your 240 fps	Open with these settings
300 fps, Casual, 70 cm (28 in)	✓ For how you play, 144–200Hz is enough (300Hz would add frames). Refresh rate that adds frames: 300Hz class (at 300 fps, 300 new frames per second). Going from 144Hz to 300Hz adds 156 frames per second. Going above 300Hz adds no new frames. Picks stay within your play style’s band (144–200Hz), this guide’s own split. Pick band: 144–200Hz. Size: 24–25 inch or 27-inch (27-inch target 56.73 cm or more; 13.27 cm to spare). Resolution: 1080p at 24–25 inch. On a 27-inch, 1440p (pixels 0.85× a 23.8-inch 1080p). 27-inch 1080p pixels are 1.13×, so coarser. 1440p has 1.78× the pixels to draw, so switch resolution, measure with Stats, and use it only if fps still reaches your target. The pick tops out at 144Hz, so it shows up to 144 of your 300 fps	Open with these settings
360 fps, Casual, 70 cm (28 in)	✓ For how you play, 144–200Hz is enough (360Hz would add frames). Refresh rate that adds frames: 360Hz class (at 360 fps, 360 new frames per second). Going from 144Hz to 360Hz adds 216 frames per second. Picks stay within your play style’s band (144–200Hz), this guide’s own split. Pick band: 144–200Hz. Size: 24–25 inch or 27-inch (27-inch target 56.73 cm or more; 13.27 cm to spare). Resolution: 1080p at 24–25 inch. On a 27-inch, 1440p (pixels 0.85× a 23.8-inch 1080p). 27-inch 1080p pixels are 1.13×, so coarser. 1440p has 1.78× the pixels to draw, so switch resolution, measure with Stats, and use it only if fps still reaches your target. The pick tops out at 144Hz, so it shows up to 144 of your 360 fps	Open with these settings
400 fps or more, Casual, 70 cm (28 in)	✓ For how you play, 144–200Hz is enough (360Hz would add frames). Refresh rate that adds frames: 360Hz class (at 400 fps, 360 new frames per second). Of your 400 fps, anything above 360 isn’t shown. Going from 144Hz to 360Hz adds 216 frames per second. Picks stay within your play style’s band (144–200Hz), this guide’s own split. Pick band: 144–200Hz. Size: 24–25 inch or 27-inch (27-inch target 56.73 cm or more; 13.27 cm to spare). Resolution: 1080p at 24–25 inch. On a 27-inch, 1440p (pixels 0.85× a 23.8-inch 1080p). 27-inch 1080p pixels are 1.13×, so coarser. 1440p has 1.78× the pixels to draw, so switch resolution, measure with Stats, and use it only if fps still reaches your target. The pick tops out at 144Hz, so it shows up to 144 of your 400 fps	Open with these settings
240 fps, Ranked, seriously, 70 cm (28 in)	✓ A 240Hz monitor shows all your fps. Refresh rate that adds frames: 240Hz class (at 240 fps, 240 new frames per second). Going from 144Hz to 240Hz adds 96 frames per second. Going above 240Hz adds no new frames. Pick band: 240–280Hz. Size: 24–25 inch or 27-inch (27-inch target 56.73 cm or more; 13.27 cm to spare). Resolution: 1080p at 24–25 inch. On a 27-inch, 1440p (pixels 0.85× a 23.8-inch 1080p). 27-inch 1080p pixels are 1.13×, so coarser. 1440p has 1.78× the pixels to draw, so switch resolution, measure with Stats, and use it only if fps still reaches your target. The pick tops out at 240Hz, borderline at the low end of the band	Open with these settings
300 fps, Ranked, seriously, 70 cm (28 in)	✓ For how you play, 240–280Hz is enough (300Hz would add frames). Refresh rate that adds frames: 300Hz class (at 300 fps, 300 new frames per second). Going from 144Hz to 300Hz adds 156 frames per second. Going above 300Hz adds no new frames. Picks stay within your play style’s band (240–280Hz), this guide’s own split. Pick band: 240–280Hz. Size: 24–25 inch or 27-inch (27-inch target 56.73 cm or more; 13.27 cm to spare). Resolution: 1080p at 24–25 inch. On a 27-inch, 1440p (pixels 0.85× a 23.8-inch 1080p). 27-inch 1080p pixels are 1.13×, so coarser. 1440p has 1.78× the pixels to draw, so switch resolution, measure with Stats, and use it only if fps still reaches your target. The pick tops out at 240Hz, so it shows up to 240 of your 300 fps	Open with these settings
300 fps, Competitive / tournaments, 70 cm (28 in)	✓ A 300Hz monitor shows all your fps. Refresh rate that adds frames: 300Hz class (at 300 fps, 300 new frames per second). Going from 144Hz to 300Hz adds 156 frames per second. Going above 300Hz adds no new frames. Pick band: 300Hz and up. Size: 24–25 inch (for competitive play at any distance, this guide’s own split). Resolution: 1080p (1920×1080). The pick tops out at 310Hz, borderline at the low end of the band	Open with these settings
360 fps, Competitive / tournaments, 70 cm (28 in)	✓ A 360Hz monitor shows all your fps. Refresh rate that adds frames: 360Hz class (at 360 fps, 360 new frames per second). Going from 144Hz to 360Hz adds 216 frames per second. Pick band: 300Hz and up. Size: 24–25 inch (for competitive play at any distance, this guide’s own split). Resolution: 1080p (1920×1080). The pick tops out at 310Hz, so it shows up to 310 of your 360 fps	Open with these settings
144 fps, Casual, 50 cm (20 in)	✓ A 144Hz monitor shows all your fps. Refresh rate that adds frames: 144Hz class (at 144 fps, 144 new frames per second). Going above 144Hz adds no new frames. Pick band: 144–200Hz. Size: 24–25 inch (closer than the 27-inch target of 56.73 cm). The 24.5-inch target is 51.59 cm, so at this distance a 23.8-inch is the target. Resolution: 1080p (1920×1080). The pick tops out at 144Hz, borderline at the low end of the band	Open with these settings
200 fps, Casual, 60 cm (24 in)	✓ A 200Hz monitor shows all your fps. Refresh rate that adds frames: 200Hz class (at 200 fps, 200 new frames per second). Going from 144Hz to 200Hz adds 56 frames per second. Going above 200Hz adds no new frames. Pick band: 144–200Hz. Size: 24–25 inch or 27-inch (27-inch target 56.73 cm or more; 3.27 cm to spare). Resolution: 1080p at 24–25 inch. On a 27-inch, 1440p (pixels 0.85× a 23.8-inch 1080p). 27-inch 1080p pixels are 1.13×, so coarser. 1440p has 1.78× the pixels to draw, so switch resolution, measure with Stats, and use it only if fps still reaches your target. The pick tops out at 144Hz, so it shows up to 144 of your 200 fps	Open with these settings

My top country in GA4 was a bot — identifying Singapore traffic without ASN data

Tue, 06 Oct 2026 00:00:00 GMT

I was about to add affiliate links to my old gadget reviews. GA4 had them as my most-viewed pages, a mini tripod review on top with a mouse review and a Walkman review close behind, and if anything on this site was going to earn from Amazon, it looked like those.

Then I sorted GA4 by country. Singapore was first by a wide margin, with 210 sessions in 28 days. Across those 210 sessions GA4 counted one engaged session and zero seconds of engagement.

Cloudflare’s free plan does not tell you which network a request came from, so I could not look up the ASN and see a data center. This post is how I settled it anyway, using this site’s own data, and what the page ranking looked like once the bot was out.

What GA4 showed

GA4 recorded the Singapore traffic as ordinary visits, and the pattern did not look like people. Last 28 days, Singapore only:

Metric	Singapore
Sessions	210
Engaged sessions	1
Total engagement time	0 s
Channel	Direct, all of it
Pages per session	1

Every session was a new user who opened one page and left. That shape, session_start ≈ first_visit ≈ page_view, is the one I had already learned to distrust when my AI-crawler numbers turned out to be scanners. Readers in Japan, by contrast, scroll and stay.

Low engagement alone proves nothing, though. Plenty of people open a page and leave. I wanted something a person would not produce.

The screen size gave it away

The strongest single signal was in GA4 itself: 201 of the 210 Singapore sessions reported a screen resolution of 1280x1200.

That is an odd display. It is nearly square, matches no common laptop or phone, and real visitors spread across dozens of resolutions. When almost every session from one country reports the same unusual value, you are looking at a configuration rather than an audience.

Dimension (Singapore, 28 days)	Value	Sessions
Screen resolution	1280x1200	201 of 210
Browser / OS	Chrome / Windows	201 of 210
City	Singapore	210 of 210
Landing page	2021–22 review and how-to posts	almost all

Filtering the whole property on 1280x1200 back to May puts the first session on 2026-06-18. It ramped to 17 a day within four days and has run at 5 to 12 a day since, 556 sessions from Singapore with that screen so far. A scattering of sessions with the same resolution came from about twenty other countries, one or two each. Those could be anything, and I left them alone.

Cloudflare’s free plan without ASN

The usual next step is to look up the network. If the requests come from a cloud provider rather than a consumer ISP, you are done. On a free zone that is not available. Asking the GraphQL Analytics API for clientAsn returns:

zone '...' does not have access to the field 'clientasndescription'

What the free plan does give you is the user agent, path, status and hour of each request group. And if Cloudflare Web Analytics is on, there is one path that only a JavaScript-executing client hits: the beacon at /cdn-cgi/rum. Grouping that path by user agent shows which clients actually ran the page, which are the same ones that fire a GA4 tag.

QUERY = """
query($zone: String!, $since: Time!, $until: Time!, $country: String!) {
  viewer { zones(filter: { zoneTag: $zone }) {
    httpRequestsAdaptiveGroups(
      limit: 5000,
      filter: {
        datetime_geq: $since, datetime_lt: $until,
        clientCountryName: $country,
        clientRequestPath: "/cdn-cgi/rum"
      }
    ) { count dimensions { userAgent datetimeHour } }
  } }
}
"""

The free plan caps each query at one day, so I ran it for seven one-day windows and summed. Then I ran the same query for Japan as a control group.

The beacon’s Chrome versions against real readers

The control group is what settled it. Seven days of beacons at /cdn-cgi/rum:

	Singapore	Japan (control)
Beacons	147	—
OS	Windows, all 147	Windows
Chrome versions	103 to 133, more than a dozen different majors	153 and 154
Hours (UTC) with a beacon	23 of 24	a few, clustered

Chrome updates itself, and my readers in Japan are on 153 and 154, the current releases as of October 2026. One browser stuck on 103 would be unusual. A stream of visitors from one city spread evenly across 103, 105, 106, 107, 109, 110, 111, 112, 116, 117, 131 and 133 is a tool picking a user agent from a list.

The hours point the same way. Singapore’s beacons arrive in 23 of 24 UTC hours with no night-time dip, and a city does not read like that.

The rest of Singapore’s 845 requests in those seven days fit too. 277 came from Linux x86_64 Chrome 131, a common headless default. 154 were Huawei’s PetalBot. 273 were 403s from my path rule on things like /admin/.env and /phpinfo.php, sent under the same rotated Windows Chrome strings.

None of this directly proves a data center, and I would rather say so than overclaim. But screen size, browser version, hour of day and engagement all point the same way, and no single human explanation covers all four.

Which posts the bot was reading

The bot did not spread evenly over the site. It went almost entirely to old review and buying-guide posts. Its pageviews by page over the same 28 days:

Page	Bot pageviews
Manfrotto PIXI EVO review	15
Monitor arm buying guide	15
Logitech MX Master 3 review	14
Sony NW-A105 review	11
microSD buying guide	10
HDMI cable buying guide	9

Those were exactly the pages that looked like my most-read. With the profile filtered out, the top post on the site is the Cloudflare Workers cron post, which the bot never visited. The reviews still have real readers, but they were never the top of the site.

The affiliate links stay on those reviews, with smaller expectations.

The bot is not only in Singapore, either. In the same week, 47 beacons from the United States were Chrome 122, all inside a single UTC hour. A country filter takes out the biggest block, but it does not make the rest clean.

What I changed, and what I left

I did not block Singapore. The traffic costs nothing beyond inflated counts, its scanning is already stopped by the path rule, and a country block would also turn away real readers there. Blocking is the right tool for protecting the origin and the wrong one for cleaning a metric, the same split I ended up with after a month of WAF rules.

GA4 cannot filter this for me. It drops bots on its known list automatically, and that list cannot be seen or extended. Its own data filters cover internal and developer traffic only.

So the fix lives in how I read the numbers:

Read engagedSessions instead of sessions. The bot produced 1 engaged session in 210.
Exclude screenResolution = 1280x1200 from Singapore when ranking pages, and recheck the profile now and then. The resolution is the bot’s choice, so it can change.
Before acting on a “top page,” check where its views come from.

The last one is the step I almost skipped. A page that looked like my best earner was mostly one machine’s routine, and ten minutes in the country report was enough to see it.

FAQ

Why does GA4 show so many direct visits from Singapore?

On my site it was an automated browser, not readers. 210 sessions in 28 days produced one engaged session and zero seconds of engagement, 201 of them reported the same 1280x1200 screen, and each landed on a single page and left. Singapore hosts a lot of cloud data centers, so a crawler that runs JavaScript there fires your GA4 tag the way a visitor would.

Doesn’t GA4 exclude bots automatically?

Only known ones. GA4 drops traffic from bots and spiders on its known list, and you cannot see or extend that list. A headless Chrome that sends a normal browser user agent is not on it, so it is recorded as a session. GA4’s own data filters only cover internal and developer traffic, which you define by IP or parameter, so they do not help with someone else’s crawler.

Can I see the ASN of a visitor on Cloudflare’s free plan?

Not through the GraphQL Analytics API. Asking httpRequestsAdaptiveGroups for clientAsn on a free zone returns an authz error saying the zone does not have access to the field. The user agent, path, status and hour are available, and the Web Analytics beacon at /cdn-cgi/rum tells you which of those user agents actually executed JavaScript.

Should I block Singapore at the WAF?

I didn’t. The traffic does no damage beyond inflating counts, its credential-scanning requests were already 403s from an existing path rule, and a country block would also turn away the occasional real reader there. I exclude it when reading the numbers instead: look at engaged sessions, and drop the 1280x1200 profile when ranking pages.

Blocking spoofed AI user agents did not clean my analytics — a month of numbers

Sun, 06 Sep 2026 00:00:00 GMT

On 8/3 I put two Cloudflare WAF rules in front of this site. One blocks credential-scanning paths like /.env, the other blocks five user agents that had fetched zero articles and only scanned. The scanners were claiming AI crawler names, which meant my AI traffic numbers were partly a record of someone else’s port sweep. Stopping them at the edge would fix the damage and the measurement at the same time.

The second half of that turned out to be wrong. Over the following month, scans claiming AI user agents went from 145 to 991 in an eight-day window. Perplexity-User, which the rules were blocking the whole time, still showed up as 595 requests. And the scanning had spread onto Amazonbot and ChatGPT-User — names that real crawlers also use.

A blocked request is still a counted request

Cloudflare’s httpRequestsAdaptiveGroups aggregates what the edge responded, so a rule turns a request into a 403 rather than removing it from the data.

Perplexity-User shows this cleanly. Before the rule it arrived and got 404s. After the rule it got 403s, and the volume went up.

Snapshot	403	404	State
08-03 (rules added that day)	1	4	window is almost entirely pre-rule
08-08	2	4	rules live, scanning still small
08-16	297	0	all blocked, all counted
08-22	311	0	same
08-28	595	0	same
09-05	37	0	same

The rule works. The 404s are gone, so nothing reaches the origin. But as long as “AI crawler traffic” is counted by user agent, those 595 requests sit inside the number. A WAF protects the origin; it does not correct a metric. Those are two jobs and I had merged them.

The setup and the two rules

A static Astro site on Cloudflare Pages, free plan zone. The rules go in through the API from a script that is idempotent — it matches on description, updating the rule if it exists and creating it if not.

The first rule matches on path and ignores the user agent.

SCAN_PATH_TOKENS = [
    "/.env", "/.git", "/.aws", "/.svn", "/.ssh",
    "/wp-", "wp-admin", "wp-login", "wp-includes", "wp-config",
    "wlwmanifest", "xmlrpc.php",
    "secrets.json", "service_account", "/actuator/", "/api/auth/session",
    "phpinfo", "/.npmrc", "/.htpasswd", "id_rsa", "config.json",
]

scan_expr = " or ".join(
    f'(http.request.uri.path contains "{t}")' for t in SCAN_PATH_TOKENS
)

contains matters here. The zone already had a hand-written rule using starts_with(http.request.uri.path, "/.env"), which only matches at the root. The requests actually arriving were /server/.env (18 of them), /admin/.env and /app/.git/HEAD, so a rule existed and caught none of them.

The second rule matches on user agent, gated on Cloudflare not having verified the client.

SPOOFED_ONLY_UAS = [
    "Bytespider", "cohere-ai", "anthropic-ai", "Perplexity-User", "Google-Extended",
]

ua_expr = "(not cf.client.bot) and (" + " or ".join(
    f'http.user_agent contains "{u}"' for u in SPOOFED_ONLY_UAS
) + ")"

Those five were the ones with zero article fetches and a 0% verification rate at the time. CCBot and PerplexityBot stayed off the list: both verified at 0% too, but both were fetching real pages with 200s, and cutting off whoever collects training data only reduces the site’s reach.

The blocked name was kept, and real names were added

The thing that changed over the month was not the rules but what the scanner called itself. Grouping scan-classified requests by claimed user agent:

Claimed UA	08-08	08-16	08-22	08-28	09-05
Perplexity-User (blocked by rule 2)	4	147	173	387	27
Amazonbot	0	67	284	119	0
ChatGPT-User	21	51	181	67	0
OAI-SearchBot	9	19	74	25	0
GPTBot	16	18	72	26	0
ClaudeBot	13	16	70	27	0
PerplexityBot	18	21	67	24	0
Google-Extended (blocked by rule 2)	5	18	70	24	0
Others (CCBot, cohere-ai, …)	28	0	0	0	0
Total	114	357	991	699	27

The blocked name stayed in use, and unblocked names were layered on top of it: Amazonbot from 67 to 284, ChatGPT-User from 51 to 181. The paths did not diversify — the same scanning simply arrived under more names.

Those two names are the problem, because real crawlers send them. In the same eight days Amazonbot split 135 verified against 489 unverified, and ChatGPT-User split 211 against 336. Matching the UA string would drop the genuine ChatGPT-User that fetched 216 articles along with the fake ones. A blocklist keyed on user agent expires the moment the impersonation moves to a name you need.

Names that cannot exist are a different case. Google’s crawler documentation states that Google-Extended has no HTTP request user agent — it is a robots.txt control token — so every request claiming it is fake and blocking costs nothing. That is the only shape where a UA rule holds up.

A path blocklist is maintained after the fact

Rule one ignores the user agent, so it stops a match no matter who claims to be sending it. It stops exactly the tokens I wrote down, though.

The 08-28 snapshot still had 300 requests from AI-claiming UAs ending in 404. They were hitting /secrets.yml, /.zshrc, /id_ecdsa and /@fs/proc/self/environ, none of which are in the token list. I wrote secrets.json but not .yml; I wrote id_rsa but not id_ecdsa.

The list grows by reading what arrived. config.json started as an exact eq "/config.json" match and was walked around with /runtime-config.json and /api/runtime-config.json. It only became a contains after I checked that none of the 164 real paths in the sitemap contain that string.

Every token you add also widens the chance of catching something you actually serve. Checking new tokens against the sitemap before adding them keeps that from turning into an outage.

The drop was not the rules

In the 09-05 snapshot, scans claiming AI user agents fell to 27. Total requests from those UAs went from 2,024 to 688, and ChatGPT-User, Amazonbot, GPTBot and OAI-SearchBot all came back 100% verified. Read on its own, that looks like the mitigation finally landing.

I changed nothing in those eight days, and these rules do not have a delayed effect.

Meanwhile zone-wide 403s went from 2,754 to 4,137. The scanning did not stop; it stopped using AI names. I cannot prove that from one snapshot, but writing “the rules brought it down” would be attributing someone else’s decision to my own change.

When judging a mitigation, put the date you changed something next to the date the number moved. If they do not line up, the cause is outside your change. Here they did not line up.

What I run now

Blocking and measurement live in separate layers.

flowchart TD
  Req["Request claiming an AI user agent"] --> WAF{"Matches a WAF rule?"}
  WAF -->|"match"| B["403 at the edge, origin untouched"]
  WAF -->|"no match"| O["Origin responds 200 / 404"]
  B --> A["Both are counted in analytics"]
  O --> A
  A --> C{"Classify at collection time"}
  C --> Y["Verified × real path = the number to read"]
  C --> N["Unverified / scan path = discard"]

The WAF is there to keep traffic off the origin and 404s out of the logs. Cleaning the numbers is not its job.

Collection handles that instead, splitting on two axes: whether verifiedBotCategory is non-empty, and whether the path exists in the sitemap. The previous post has that classifier in full. The scan series in this post is just its scan column lined up across snapshots.

Keeping the layers apart means editing a WAF rule does not move the measurement series, and changing the classifier does not change what gets blocked. After a month, that independence is the part I would keep.

Wrapping up

The WAF earns its place: nothing reaches the origin and the 404 pile shrank. The half I got wrong was expecting the analytics to get cleaner. Blocked requests stay in the data as 403s, so contamination has to come out at collection time.

The other lesson is that user-agent blocking has a short shelf life. Names that cannot exist — Google-Extended, anthropic-ai — stay blockable indefinitely. Once the impersonation moves to a name real crawlers share, matching the string is no longer available to you, and judging by the verification result instead is about the only way out.

For the other side of this — making the site readable to the crawlers I do want — there is the multilingual llms.txt post.

The same contamination showed up on the GA4 side as phantom sessions from Singapore; telling that bot apart without ASN data is its own post.

FAQ

Do WAF-blocked requests disappear from Cloudflare analytics?

No. The GraphQL Analytics API (httpRequestsAdaptiveGroups) aggregates the response the edge returned, so a blocked request stays in the data as a 403. In my logs, a user agent I was actively blocking accounted for 595 requests over eight days, every one of them a 403. Blocking keeps traffic off your origin; it does not remove it from your counts.

Is it safe to block a spoofed AI crawler by matching the User-Agent string?

Only for names that cannot exist. Google-Extended has no HTTP user agent of its own, so every request claiming it is fake and a UA match costs nothing. ChatGPT-User and Amazonbot are different: real crawlers send those exact strings, so a UA match would also drop genuine article fetches — and those are precisely the names the scanning moved to once I blocked the others.

How far does a path blocklist get you?

As far as the tokens you wrote. Common ones like /.env and /wp- stop, but paths outside the list — /secrets.yml, /.zshrc, /id_ecdsa, /@fs/proc/self/environ — pass through and become 404s; 300 of them remained in one eight-day window. A path blocklist is maintained after the fact, by reading what actually arrived.

If scans drop after I add a rule, did the rule work?

Not necessarily. Scans claiming AI user agents fell from 699 to 27 in a window where I changed no rules at all, while zone-wide 403s rose from 2,754 to 4,137. The scanning did not stop — it stopped using AI names. Line up when you changed something against when the number moved, and look outside your own changes when the two do not match.

AI crawler user agents are self-reported: 468 real fetches, 991 fake ones

Sat, 22 Aug 2026 00:00:00 GMT

Every eight days I pull the AI crawler numbers for this site: how many requests from ChatGPT-User, how many from GPTBot, one row per User-Agent. Rising numbers meant the AI systems were picking the site up.

That reading is gone. In this week’s run, requests that actually fetched an article came to 468, while credential scans came to 991 — more than double. Of the requests calling themselves GPTBot, Cloudflare could verify only 13% as OpenAI. The series I had been watching was tracking attacker volume, not AI interest.

Check the claim against a verification result

A User-Agent is a header. Writing GPTBot/1.2 into it costs nothing, so a table grouped by UA is a table of claims.

Behind Cloudflare you get two things to check those claims against. The first is verifiedBotCategory, which carries Cloudflare’s own reverse-DNS verification result and is an empty string when the request could not be verified. The second is whether the requested path exists in your sitemap. Split on both and what remains is “verified bots fetching real pages” — the only figure worth putting in a report.

The setup

A static Astro site on Cloudflare Pages, free plan zone, queried from Python through the GraphQL Analytics API.

Piece	Purpose	Free plan
`httpRequestsAdaptiveGroups`	request aggregation	yes (max one day per request)
`verifiedBotCategory`	bot verification result	yes
`clientAsn` / `botClass`	source ASN, bot classification	not permitted
`requests` + `pandas`	fetch and aggregate	—

Retention on the free plan is short — daily granularity reaches back about a week. You cannot re-derive the past later, so the script writes a snapshot to disk on every run.

One UA string, two rows

Adding verifiedBotCategory to the dimensions is the whole trick. Eight days of my logs:

Claimed UA	Verified	Unverified	Verified share
ChatGPT-User	211 (AI Assistant)	336	39%
Amazonbot	135 (AI Crawler)	489	22%
ClaudeBot	133 (AI Crawler)	126	51%
OAI-SearchBot	61 (Search Engine Crawler)	132	32%
GPTBot	19 (AI Crawler)	123	13%
meta-externalagent	209 (AI Crawler)	0	100%
Applebot	56 (AI Search)	0	100%
PerplexityBot	0	144	0%
Perplexity-User	0	311	0%

The same string lands in two rows. Of 547 requests presenting as ChatGPT-User, 211 came from an address Cloudflare traced back to OpenAI.

Two agents came through clean: meta-externalagent and Applebot, both 100% verified with no impersonation mixed in. Those are the only two rows where the claimed number is usable as-is. At the other end, 455 Perplexity-branded requests were all unverified — not a single confirmed one in the window.

The query:

QUERY_BY_UA_VERIFIED = """
query ($zoneTag: String!, $since: Time!, $until: Time!) {
  viewer {
    zones(filter: { zoneTag: $zoneTag }) {
      httpRequestsAdaptiveGroups(
        limit: 5000
        filter: { datetime_geq: $since, datetime_leq: $until }
        orderBy: [count_DESC]
      ) {
        count
        dimensions { userAgent verifiedBotCategory }
      }
    }
  }
}
"""

A free zone returns at most one day per request, so the range gets chunked by the caller. Days past the retention window error out, so a failed day is skipped rather than fatal.

def fetch_rows(token, query, zone_tag, since, until):
    # split the range into one-day windows and concatenate the rows
    all_rows, cursor = [], since
    one_day = dt.timedelta(days=1)
    while cursor < until:
        win_end = min(cursor + one_day, until)
        variables = {
            "zoneTag": zone_tag,
            "since": cursor.isoformat() + "Z",
            "until": win_end.isoformat() + "Z",
        }
        try:
            all_rows.extend(rows_from(gql(token, query, variables, exit_on_error=False)))
        except RuntimeError as e:
            print(f"    [skipped] {cursor.date()}-{win_end.date()}: {e}")
        cursor = win_end
    return all_rows

One of the agents in my logs cannot exist

123 requests claimed Google-Extended. Verified share 0%, and 70 of them hit paths like .env or wp-login.

No inference is needed on this one. Google’s crawler documentation states that Google-Extended has no separate HTTP request user agent string: crawling happens under the existing Google user agents, and the token exists only to be addressed in robots.txt when you want to control AI training use.

So every request presenting that UA is, by definition, not Google. The same logic extends to any bot whose operator publishes IP ranges — OpenAI ships gptbot.json, and a reverse lookup settles the question.

On my zone all 123 were already stopped with a 403.

Classify the path, not just the client

Verification alone is not enough. A verified bot fetching robots.txt has read nothing. Five buckets:

SCAN_PATTERNS = (
    "wp-", ".env", ".git", ".aws", ".svn", ".ssh", "secrets", "credentials",
    "config.json", "service_account", "actuator", "api/auth", "phpinfo",
    ".bak", ".yml", ".yaml", ".php", ".sql", "id_rsa", ".npmrc", ".htpasswd",
)
OPS_PREFIXES = ("/robots.txt", "/sitemap", "/llms.txt", "/favicon", "/rss", "/feed", "/.well-known/")
ASSET_PREFIXES = ("/_astro/", "/images/", "/assets/", "/fonts/", "/cdn-cgi/", "/_image")


def classify_path(path, sitemap_paths):
    # content = a real page was consumed / ops = crawl bookkeeping
    # asset = static file / scan = credential probing / other = path does not exist
    if not path:
        return "other"
    low = path.lower()
    if any(k in low for k in SCAN_PATTERNS):
        return "scan"
    if low.startswith(OPS_PREFIXES):
        return "ops"
    if low.startswith(ASSET_PREFIXES):
        return "asset"
    if not sitemap_paths:
        return "unknown"   # cannot assert existence, so cannot call it content
    return "content" if (path.rstrip("/") or "/") in sitemap_paths else "other"

Using the sitemap as the source of truth for existence is what makes the classification hold up. Deciding from the response status is the obvious alternative, but redirects and paths that answer 200 without being real pages both leak into it. The set of paths you have declared public is a cleaner definition of “a page of mine”.

The unknown branch earns its place too. Fold it into content and the number spikes on any day the sitemap fetch fails, with nothing in the output to explain why.

flowchart TD
  Req["Request claiming an AI bot UA"] --> V{"verifiedBotCategory empty?"}
  V -->|"empty = unverified"| Fake["impersonation, drop from the count"]
  V -->|"verified"| P{"path in sitemap?"}
  P -->|"exists"| C["content = the number to report"]
  P -->|"robots.txt / sitemap"| O["ops = crawl bookkeeping"]
  P -->|".env / wp-login"| S["scan = credential probing"]

The result: 468 article fetches against 991 scans

Eight days, both axes applied. The content column is the one worth reading.

Claimed UA	content	ops	scan	total	verified
ChatGPT-User	216	0	181	547	39%
meta-externalagent	93	9	0	209	100%
Amazonbot	85	1	284	624	22%
Applebot	27	9	0	56	100%
OAI-SearchBot	17	43	74	193	32%
PerplexityBot	13	8	67	144	0%
GPTBot	7	9	72	142	13%
ClaudeBot	6	129	70	259	51%
Google-Extended	0	0	70	123	0%
Perplexity-User	0	0	173	311	0%

content totals 468 and scan totals 991. By status code, 403s came to 957 against 705 requests answered with 200 — the WAF is now blocking more than the site serves to these clients.

Six days earlier the same site read 610 content against 357 scan, so the ratio inverted inside a week. Grouped by UA alone, the story would have been “AI crawler traffic is up”.

The ClaudeBot row is worth pulling apart as well. Of 133 verified requests, 6 fetched articles and 129 fetched robots.txt and similar. “ClaudeBot sent 259 requests” and “six articles were read” are the same data.

What the split still cannot tell me

Both axes remove impersonation. Neither resolves this: content fell from 610 to 468, and I cannot say whether AI-side interest dropped or whether impersonation that used to land in content got reclassified as scan. GPTBot went from 85 content fetches to 7 while its verified share fell from 77% to 13%, so “fetched less” and “buried under fakes” are happening at once.

The aggregate output therefore keeps the verified share as a column, and any series whose share collapses gets marked unreadable for that snapshot. Two numbers side by side — the denominator and the verified share — beat one clean-looking figure.

Wrapping up

If you are reporting AI crawler numbers, distrust the UA first. One extra dimension splits claim from reality, and the shape of the split also tells you how far to trust that series this week.

What surprised me is which agents came through clean: meta-externalagent and Applebot, and nothing else. The better-known the bot, the more it gets impersonated, and the less usable its row becomes. Reputation of the operator turned out to say nothing about the quality of the data.

The same measure-first habit shows up in the post about measuring cross-post overlap. The other side of this work — making the site legible to AI clients in the first place — is in generating a bilingual llms.txt.

What came of blocking the impersonators at the edge is in a month of blocking spoofed AI user agents: the blocked requests stayed in the analytics, and the scanning moved onto names real crawlers use.

FAQ

Why can’t I count AI crawler traffic by User-Agent?

The User-Agent is a request header, so anyone can send GPTBot or ChatGPT-User in it. In eight days of my own logs, only 211 of the 547 requests calling themselves ChatGPT-User came from an IP Cloudflare could verify as OpenAI. Amazonbot was 135 verified out of 624. Count by UA alone and those unverified requests become your evidence that AI interest is growing.

Can I separate impersonators on Cloudflare’s free plan?

Yes. In the GraphQL Analytics API, add userAgent and verifiedBotCategory to the dimensions of httpRequestsAdaptiveGroups and you get Cloudflare’s reverse-DNS verification result per row, with an empty string meaning unverified. clientAsn and botClass need a paid plan, but verifiedBotCategory works on a free zone. The one constraint is that a single request covers at most one day, so loop over the range.

What should I do with requests claiming to be Google-Extended?

Treat every one of them as fake. Google’s crawler documentation states that Google-Extended has no separate HTTP request user agent string — it is a robots.txt control token, and crawling happens under the regular Google user agents. My logs had 123 requests claiming it, 0% verified, with 70 of them hitting credential-scanning paths.

Is a verified request the same as someone reading an article?

No, and separating the two matters. A verified bot fetching robots.txt, sitemap.xml or llms.txt is doing crawl bookkeeping, not reading. Of ClaudeBot’s 133 verified requests, 6 fetched articles and 129 fetched files like robots.txt. Classifying by whether the path exists in your sitemap splits content from ops and leaves only the number worth reading.

Is cross-posting duplicate content? I measured it before deleting anything

Sun, 16 Aug 2026 00:00:00 GMT

An AdSense review came back with the “low value content” label and no indication of which pages triggered it. Looking for a plausible cause in my own setup, I landed on the fact that the same articles go out to Zenn, Qiita and dev.to. I wrote a remediation doc: unpublish the five live cross-posts, set canonical everywhere, resubmit.

Before running any of it, I counted how much text actually overlaps. Across 11 articles and 22 files, three prose lines matched — and two of those were horizontal rules. The remediation doc went in the bin, and the work turned into adding canonical URLs on the platforms that accept them.

Measure the overlap before you delete a channel

If you write each syndicated version separately rather than pasting the original, the lines that match will be code, tables and headings. Prose barely survives the rewrite. A single overlap percentage hides that completely, so split the match by line type before drawing any conclusion.

The fix, when there is one, is a canonical URL rather than a deletion. dev.to, Hashnode and Medium all accept one through their APIs. Qiita, Zenn and note do not, which means the option of pointing at an original was never available there in the first place.

What the setup looks like

Aulvem runs a hub-and-spoke layout. The site is the hub; the platforms are spokes that stand on their own and link back. Same subject, different reader and different depth, written from scratch on each side.

Layer	Where	Body file
Hub	aulvem.com	`src/content/blog/{en,ja}/<slug>.mdx`
Spoke	dev.to / Hashnode	`body.en.md`
Spoke	Zenn / Qiita	`body.ja.md`

The measured set covers everything published between 2026-05-17 and 2026-08-08: 11 articles, counted as 22 pairs because English and Japanese are separate files. The syndicated bodies carry no frontmatter and the hub files are MDX, so both sides get their frontmatter stripped before comparison.

The script is about thirty lines

Normalize whitespace, drop blank lines, put the hub’s lines in a Set, and walk the syndicated file. The only part that needs care is the classification.

import fs from "node:fs";

const norm = (s) => s.replace(/\s+/g, " ").trim();

function bodyOf(file) {
  let t = fs.readFileSync(file, "utf8");
  if (t.startsWith("---")) t = t.slice(t.indexOf("\n---", 3) + 4); // strip frontmatter
  return t;
}

// Split into [normalized line, kind]. Everything inside a fence is code.
function classify(text) {
  const out = [];
  let inFence = false;
  for (const raw of text.split("\n")) {
    const line = norm(raw);
    if (/^```/.test(line)) { inFence = !inFence; out.push([line, "code"]); continue; }
    if (!line) continue;
    if (inFence) { out.push([line, "code"]); continue; }
    if (/^#{1,6}\s/.test(line)) { out.push([line, "heading"]); continue; }
    if (/^\|/.test(line)) { out.push([line, "table"]); continue; }
    out.push([line, "prose"]);
  }
  return out;
}

const hub = new Set(classify(bodyOf(hubFile)).map(([l]) => l));
const ext = classify(bodyOf(syndicatedFile));
const matched = ext.filter(([l]) => hub.has(l));

Toggling inFence matters more than it looks. Shell snippets start lines with #, SQL and YAML samples start lines with |, and mermaid blocks are full of both. Classify by leading character alone and a large share of the code lands in the prose bucket, which is the exact number you were hoping to trust.

Dropping blank lines matters too. They always match, so leaving them in inflates both sides of the ratio.

47.9% overlap, and 762 lines of it were code

Metric	Value
Lines in syndicated bodies	1,658
Lines matching the hub	795 (47.9%)
— inside code fences	762
— table rows	13
— headings	17
— prose	3
Prose lines in syndicated bodies	393
Prose overlap	0.8%

Printing those three prose lines took a second: two are --- rules, and the third is a single English bullet from the Zod schema article. Nothing else in 393 lines of prose survived from one version to the other.

Per article the pattern holds. High overlap tracks code density, not copying.

Article	Overlap	Code
sitemap-lastmod (en)	64.5%	71
idempotent-notification-outbox (en)	65.6%	40
sheets-xlsx-excel-silent-breakage (en)	67.8%	53
aulvem-blog-architecture (en)	6.5%	3

Code repeats because both versions describe the same implementation. Rewriting CREATE UNIQUE INDEX uq_outbox_dedup for the syndicated copy would make one of the two articles wrong. Shared snippets are a weak basis for calling your own site a duplicate.

Where a canonical URL fits, and where it does not

flowchart LR
  Devto["dev.to<br/>canonical_url"] --> HP["aulvem.com<br/>(original)"]
  Hashnode["Hashnode<br/>originalArticleURL"] --> HP
  Medium["Medium<br/>canonicalUrl"] --> HP
  Qiita["Qiita<br/>not supported"] -.-> HP
  Zenn["Zenn<br/>not supported"] -.-> HP
  Note["note<br/>not supported"] -.-> HP

Platform	Canonical	Where it goes
dev.to	Yes	`article.canonical_url` on the articles endpoint
Hashnode	Yes	`originalArticleURL` on the `publishPost` mutation
Medium	Yes	`canonicalUrl` on the posts endpoint, or Import from URL
Qiita	No	no equivalent parameter on the items API
Zenn	No	no such key in article frontmatter
note	No	no field in the editor

The Forem API exposes canonical_url on the article object, so syndicating with an original declared costs one line in the payload.

body: JSON.stringify({
  article: {
    title: fm.title,
    body_markdown: body,
    published: fm.published ?? false,
    tags,
    canonical_url: fm.canonical_url,   // here
  },
}),

Hashnode has a trap worth knowing about. originalArticleURL exists on the publishPost input but not on createDraft. If your pipeline creates a draft and you press publish in the dashboard, the canonical never gets attached.

const input = publish
  ? { title, contentMarkdown, publicationId, tags, originalArticleURL: fm.canonical_url }
  : { title, contentMarkdown, publicationId, tags };   // no place for it on a draft

Qiita’s API v2 docs list title, body, tags, private, tweet and organization_url_name for item creation, with nothing that declares an external original. Zenn’s article frontmatter is title, emoji, type, topics, published and publication_name — same story.

I added canonical_url: https://aulvem.com/blog/<slug>/ to all 11 dev.yaml drafts. Because the value is fixed at post time, the five already-published articles do not pick it up; those need PUT /articles/{id} or a manual edit. That part is still open.

”Temporarily unpublish” is not a reversible operation

On a platform with no canonical field, the way back from unpublishing is to post again — new URL, views reset, inbound links broken. What reads as a temporary measure is a deletion with extra steps.

The traffic makes the trade obvious.

Channel	Result
Qiita, 11 articles, cumulative views	2,240
dev.to, cumulative views	309
Zenn, cumulative impressions	182
aulvem.com search clicks (28 days)	7, from 1,335 impressions

The article I published on 8 August took 196 views on Qiita in its first eight days, with a 221-second average read time. The same article on my own site: 10 page views over the same window. Removing the syndicated copies would have shut down the side where people read, to protect the side where they do not, in exchange for an AdSense placement worth a few hundred yen a month.

The two things I had conflated

The first was treating an AdSense decision as evidence about search-side duplicate handling. AdSense reviews a site against its program policies; search-side duplicate URL consolidation is a normalization step driven by canonical and hreflang. One does not explain the other.

The second was reading “Qiita’s canonical points at Qiita” as a symptom. Every platform without canonical support emits a self-referencing canonical. It is the default state, not a signal about my site.

While those two ideas were in play I had also drafted 17 gadget reviews from 2021 out of the site. Those are back. Hiding writing I produced myself is a strange move when the concern on the table is originality.

Wrapping up

The useful step was counting, not deleting. Thirty lines of script, one run, and the answer — three matching prose lines out of 393 — removed every task that came after it.

For canonical, the rule I settled on is simple: set it where the platform accepts it, accept its absence where it does not. Qiita, Zenn and note give me no way to point at an original, and that is not a reason to stop syndicating to them.

What stayed with me is how close I came to running the remediation first. The same instinct — make breakage visible before acting on a hunch — shows up in the Zod schema enforcement post and in the .xlsx silent breakage post.

FAQ

How do I check whether my cross-posts count as duplicate content?

Normalize both bodies line by line, put the canonical article’s lines in a set, and count how many of the syndicated article’s lines are in it. The part that decides the answer is tracking code fences, because a fenced block is full of lines starting with # or | that a naive classifier reads as prose. Across my 11 articles and 22 files, 762 of the 795 matching lines sat inside fences and only 3 of 393 prose lines matched.

Which platforms let me declare a canonical URL?

dev.to takes article.canonical_url on the article endpoint, Hashnode takes originalArticleURL on the publishPost mutation, and Medium takes canonicalUrl on the post endpoint. Qiita’s items API has no equivalent parameter, Zenn’s article frontmatter has no such key, and note’s editor has no field for it. On those three the option to point at an original simply does not exist.

Is unpublishing the syndicated copies a safe temporary fix?

Not on platforms without canonical support, because the only way back is to post again under a new URL, which drops the views and any inbound links. My syndicated totals are 2,240 views on Qiita, 309 on dev.to and 182 impressions on Zenn, against 7 search clicks to aulvem.com over 28 days. Deleting the copies removes the channel where the readers actually are.

No. AdSense reviews a site against its program policies, while search-side duplicate handling is URL consolidation driven by canonical and hreflang signals. Reasoning from one to the other is how I ended up drafting a fix for a problem I had never measured.

A Google Sheets .xlsx breaks silently in Excel — rewriting 203 cells without dynamic arrays

Sat, 08 Aug 2026 00:00:00 GMT

Build something in Google Sheets, export it as .xlsx, hand it to someone, and they open a different file than the one you tested. The failure is quiet enough that neither of you finds out.

After putting a password manager sheet on Booth, I wanted to check that the “works in Excel and Google Sheets” line on the product page was true, so I opened the exported .xlsx with openpyxl. Every dashboard formula had turned into a function I’d never seen called __xludf.DUMMYFUNCTION. There were 203 of them. All 203 were wrapped in IFERROR. This is how I found them and what I rewrote them into.

No error appears, so nothing looks broken

Google-only functions don’t get stripped on export. Excel can’t parse them, so they’re stored as __xludf.DUMMYFUNCTION("original formula") with the original kept as a string argument. Opening that in Excel should give you #NAME?.

It didn’t, because the original formula was already inside an IFERROR.

=IFERROR(__xludf.DUMMYFUNCTION("IFERROR(SORTN(FILTER({'パスワード一覧'!A2:A80, …

That wrapper was there to print an em dash when a query returned nothing. In Excel it also catches “no such function.” What comes back is an empty string or the fallback text, and the cell reads ”— no matches —”. The display for zero results and the display for a formula that never ran are the same.

Five places in the shipped file were affected.

Location	Functions used	Result in Excel
Password list, column G (strength)	`ARRAYFORMULA` + `REGEXMATCH`	Every row blank
Dashboard B10 (top 5 to rotate)	`SORTN` + `FILTER`	”No matches”
Dashboard B19 (candidates to drop)	`FILTER`	”No matches”
Dashboard B28 (reused-password warning)	`SORTN` + `UNIQUE` + `FILTER`	”No matches”
Subscriptions B10 (listing)	`QUERY`	Blank

flowchart TD
  A["Sheets formula<br/>ARRAYFORMULA / FILTER / SORTN / QUERY"] --> B["Opened in Excel"]
  B --> C["Function does not exist"]
  C --> D["The outer IFERROR catches it"]
  D --> E["Empty string returned"]
  E --> F["Screen shows '— no matches —'"]

The reused-password warning read “no matches” not because there were no duplicates, but because the check never ran. That’s a safety-adjacent readout showing a clean bill of health it hadn’t earned.

Count the broken cells instead of hunting for them

__xludf.DUMMYFUNCTION persists as text, so walking every sheet with openpyxl gives you an exact count. It beats squinting at the dashboard, and the same code proves the count went to zero after the rewrite.

KEYS = ("DUMMYFUNCTION", "REGEXMATCH", "SORTN",
        "QUERY(", "ARRAYFORMULA", "UNIQUE(", "FILTER(")

def check_no_google_funcs(path):
    wb = openpyxl.load_workbook(path)
    bad = []
    for sh in wb.worksheets:
        for row in sh.iter_rows():
            for c in row:
                v = c.value
                txt = getattr(v, "text", None) or (v if isinstance(v, str) else "")
                # formula cells only — the FAQ sheet mentions these names in prose
                if not txt.startswith("="):
                    continue
                if any(k in txt for k in KEYS):
                    bad.append(f"{sh.title}!{c.coordinate}")
    return bad

That txt.startswith("=") line earns its place. The workbook ships with a how-to sheet whose FAQ answers use the word ARRAYFORMULA in ordinary prose. Without the filter, prose you can’t fix shows up in your list of things to fix.

The count came to 203: 199 in the password list, 3 on the dashboard, 1 in subscriptions. Nearly all of the 199 were column G, where only G2 held a real formula. G3 downward had become =IFERROR(__xludf.DUMMYFUNCTION("""COMPUTED_VALUE"""),"") — 198 placeholder husks where the array formula used to spill.

Skip dynamic arrays — Excel 2016 doesn’t have them

Excel 365 ships FILTER, SORT, UNIQUE, and TAKE. They map closely onto what the Google formulas were doing, which makes them the obvious target. They also don’t exist in Excel 2019 or 2016. I can’t know which version a buyer runs, and quietly amending the product page to say “requires 365” isn’t a fix, it’s a retreat.

So the rewrite avoids dynamic arrays entirely.

Approach	Sheets	Excel 365	Excel 2016
`ARRAYFORMULA` / `QUERY` / `SORTN`	works	no	no
`FILTER` / `SORT` / `UNIQUE` (dynamic arrays)	works	works	no
Per-row helper columns + `INDEX`/`MATCH`	works	works	works

The idea is to split “sort and take the top N” into two steps. Each row computes its own rank into a helper column; the display side just asks for the row whose rank is n. Delete the sorting operation and the need for dynamic arrays goes with it.

I generated the replacements row by row from openpyxl. Pasting 200 rows × 4 helper columns by hand isn’t realistic, and when the range needs to change later it’s one constant in the script.

Write the rank into a helper column with SUMPRODUCT

“Everything past 90 days since its last update, ordered by days elapsed, descending” becomes a per-row formula. RANK doesn’t fit because the ranking applies to a filtered subset.

def rank_renew_formula(row):
    L, A = f"$L{row}", f"$A{row}"
    la, lb = f"$L$2:$L${LAST_ROW}", f"$A$2:$A${LAST_ROW}"
    rows = f"ROW($L$2:$L${LAST_ROW})"
    return (
        f'=IF(AND({A}<>"",ISNUMBER({L}),{L}>90),'
        f'SUMPRODUCT(({lb}<>"")*ISNUMBER({la})*({la}>90)*'
        f'(({la}>{L})+(({la}={L})*({rows}<ROW({L})))))+1,"")'
    )

Which produces:

=IF(AND($A2<>"",ISNUMBER($L2),$L2>90),
   SUMPRODUCT(($A$2:$A$200<>"")*ISNUMBER($L$2:$L$200)*($L$2:$L$200>90)
   *(($L$2:$L$200>$L2)+(($L$2:$L$200=$L2)*(ROW($L$2:$L$200)<ROW($L2)))))+1,"")

Multiplying conditions together is the whole trick with SUMPRODUCT: count the rows that beat you, add one, and that’s your rank. Two details do real work here.

The ISNUMBER term can’t be dropped. Column L returns "" for rows with nothing to measure, and Excel evaluates text as greater than any number, so $L$2:$L$200>90 on its own would let every blank row into the ranking. Multiplying by the numeric test removes them.

Ties need handling too. Two rows with identical elapsed days would share a rank, and MATCH only ever returns the first hit, so the other row vanishes from the table. Adding (equal value) × (lower row number) puts the earlier row first and keeps ranks distinct.

Replace REGEXMATCH with an array constant and SUBSTITUTE

Password strength was checking character classes with REGEXMATCH(E2, "[A-Z]"). Excel has no regular expressions, so I switched to counting occurrences per class instead.

def char_class_count(cell, chars):
    """Build a formula counting how many characters of `cell` appear in `chars`."""
    arr = "{" + ",".join(f'"{c}"' for c in chars) + "}"
    return f'SUMPRODUCT(LEN({cell})-LEN(SUBSTITUTE({cell},{arr},"")))'

Handing SUBSTITUTE an array constant of {"A","B",…,"Z"} gives back 26 strings, each missing one letter. Sum the length deltas with SUMPRODUCT and you have the uppercase count. SUBSTITUTE is case-sensitive, so [A-Z] and [a-z] port over directly. Symbols come out as total length minus upper, lower, and digits.

My first version split the string with MID(E2, ROW(INDIRECT("1:"&LEN(E2))), 1) and matched each character with FIND. Verification rated every single password as weak. INDIRECT array expansion behaves differently across environments, and when it fails it lands in the IFERROR branch — so once again, quietly. Dropping INDIRECT cleared it.

Let the display side do nothing but INDEX/MATCH

With ranks sitting in a helper column, the dashboard only has to fetch the row holding rank n.

def lookup(rank_col, value_col, first_row, row):
    n = row - first_row + 1
    return (
        f'=IFERROR(INDEX({SHEET}!{value_col}$2:{value_col}${LAST_ROW},'
        f'MATCH({n},{SHEET}!${rank_col}$2:${rank_col}${LAST_ROW},0)),"")'
    )

=IFERROR(INDEX('パスワード一覧'!A$2:A$200,
         MATCH(1,'パスワード一覧'!$R$2:$R$200,0)),"")

The rank could come from a relative reference like ROWS(B$10:B10), but since Python is emitting a formula per row anyway, a literal integer is simpler and dodges another function-support difference. IFERROR shows up here as well, though now it carries only its intended meaning: no row has rank 3, so there are fewer than three matches.

Formulas aren’t the only thing that goes quiet

Writing sample data for the trial edition with openpyxl, I repeated the same class of mistake. Passing dates as the string "2019-04-02" blanked out days-elapsed, the needs-update count, and the top-5 rotation list all at once.

Column L computes INT(TODAY()-K2). When K2 holds text the subtraction fails, IFERROR absorbs it, and an empty string comes back. The helper column’s ISNUMBER then correctly excludes the row, MATCH finds nothing to fetch, and the dashboard prints “no matches.” No error surfaces anywhere along that chain.

ws.cell(row=r, column=10).value = datetime.date.fromisoformat(reg)  # a date, not a str
ws.cell(row=r, column=10).number_format = "yyyy-mm-dd"

IFERROR is worth having as long as it only catches the failures you anticipated. Once it starts folding unanticipated ones into the same appearance, a file can ship broken and stay that way.

Wrapping up

If you build in Google Sheets and ship .xlsx, open the export with something other than a spreadsheet at least once. Counting __xludf.DUMMYFUNCTION takes twenty lines, and the same twenty lines confirm the count reached zero after you fix it.

For the rewrite itself, dropping all the way down to helper columns and INDEX/MATCH suited a distributed file better than reaching for 365’s dynamic arrays. The formulas get long, but Excel 2016 and Google Sheets produce the same numbers. When a script generates them, formula length stops being a cost worth optimizing.

The same instinct — make breakage visible instead of trusting it not to happen — shows up in catching content rule violations at build time with Zod. The corrected file is live on the password manager sheet.

FAQ

What happens to Google-only functions when you export to .xlsx?

They survive the export. ARRAYFORMULA, FILTER, SORTN, QUERY, UNIQUE, and REGEXMATCH have no Excel equivalent, so the exported file stores them as a function called __xludf.DUMMYFUNCTION with the original formula kept as a string argument. openpyxl shows them plainly. Excel never evaluates them.

Why does it show ‘no matches’ instead of an error?

Because the original formula was already wrapped in IFERROR. On the Sheets side that wrapper existed to print an em dash when nothing matched. In Excel it also catches the “this function does not exist” error, so the cell returns an empty string or the fallback text and the sheet looks fine. Every one of the 203 broken cells in my file had that shape.

Can’t you just swap in Excel 365’s FILTER / SORT / UNIQUE?

Only if you can assume the buyer’s Excel version. Dynamic arrays don’t exist in Excel 2019 or 2016, so they’re off the table for something you ship to strangers. Putting the rank, sequence, and flag into per-row helper columns and reading them back with INDEX/MATCH produces the same result on Excel 2016 and Google Sheets alike.

How do you count character classes without REGEXMATCH?

With an array constant and the length delta from SUBSTITUTE. SUBSTITUTE($E2,{"A","B",…},"") returns one string per letter with that letter removed, so summing the LEN differences with SUMPRODUCT gives you the count. SUBSTITUTE is case-sensitive, which covers the upper-versus-lower distinction for free. Splitting the string with MID and INDIRECT is fragile across environments — in my case it silently rated every password as weak.

Idempotent notifications with an outbox — separate detection from sending, re-check right before send

Mon, 03 Aug 2026 00:00:00 GMT

Send a notification the instant you detect its condition, and the short gap between detection and delivery will bite you. Stock runs out, a price reverts, a user removes a watch — tens of seconds to minutes after detection, the alert may no longer apply.

I built and shipped a price watcher called Yasugoro on Cloudflare Workers + D1. It notifies you when a staple gets cheaper, so sending a “no longer cheaper” alert is the worst thing it can do. So I split detection from sending: detection only appends to an outbox, and a separate step re-checks right before it sends. This post is about how that outbox is built as a state machine on D1 — idempotent enqueue, a pre-send re-check, and retries — not about what gets detected (the buy-timing decision).

Detection only enqueues

Detection’s whole job is one INSERT into the outbox when a condition holds. Actually sending is a separate step (a per-minute dispatch). That decouples detection from delivery in time.

Make the enqueue idempotent. However many times the same watch’s same trigger fires in a day, the notification should collapse to one. A SQLite expression index plus ON CONFLICT DO NOTHING absorbs it.

-- migration: an expression index that collapses duplicates to one
CREATE UNIQUE INDEX uq_outbox_dedup
  ON outbox (watch_id, trigger, date(detected_at), channel);

// enqueue is idempotent; returns false when ON CONFLICT ignored it
export async function enqueueOutbox(db: D1Database, input: EnqueueInput): Promise<boolean> {
  const res = await db.prepare(
    `INSERT INTO outbox
       (watch_id, user_id, channel, trigger, status, detected_at, detected_payload_json, retry_count, created_at)
     VALUES (?, ?, ?, ?, 'pending', ?, ?, 0, ?)
     ON CONFLICT (watch_id, trigger, date(detected_at), channel) DO NOTHING`
  ).bind(/* ... */).run();
  return (res.meta?.changes ?? 0) > 0; // changes=1 on real insert, 0 when dedup ignored it
}

Keying on date(detected_at) is the trick: it folds “the same detection on the same day” into one row while still letting the next day’s detection through as a new one. Detection can run many times over without the outbox ballooning with duplicates.

Five states, one direction

Each outbox row moves one way through five states. An enqueued pending becomes sending → sent if it passes the pre-send re-check, or dropped if it broke. A transient send failure goes to retry, and to failed at the cap.

flowchart LR
  pending -->|re-check holds| sending
  pending -->|condition broke| dropped
  sending -->|delivered| sent
  sending -->|transient fail| retry
  retry -->|back to re-check| pending
  retry -->|cap reached| failed

sent, dropped, and failed are terminal. dropped isn’t an error — it’s a normal terminal state meaning “we learned we shouldn’t send, so we didn’t.” Keeping it distinct from errors lets you later count how many you correctly withheld.

Re-check right before sending, and drop what broke

The dispatch handles one row at a time: re-fetch → re-check whether the detection condition still holds → send only the ones that do. This is the core that stops stale alerts.

// processing one outbox row (skeleton)
async function processOutbox(o, now, deps) {
  // (0) pre-guards: watch removed / target paused -> dropped
  if (watchRemoved || listingPaused) { await markDropped(db, o.id, ...); return 'dropped'; }

  // (1) re-fetch the current value right before sending (split drop vs retry by error kind)
  const r = await adapter.getOne(itemCode);
  if (!r.ok) {
    if (r.error.kind === 'empty') { await markDropped(db, o.id, 'out_of_stock'); return 'dropped'; }
    await scheduleRetry(db, o, `refetch_${r.error.kind}`); return 'retry';
  }

  // (2) re-check that the detection condition still holds (the body is domain-specific)
  const recheck = recheckTrigger(/* current value */);
  if (!recheck.ok) { await markDropped(db, o.id, 'condition_lost'); return 'dropped'; }

  // (3) only for the ones that hold: build the body from the latest value and deliver
  await markSending(db, o.id, o.channel);
  const result = await deps.deliver(o.userId, buildPayload(/* re-fetched latest value */));
  // ... sent / retry / dropped
}

What the re-check contains (which trigger still “holds”) depends on the service. What’s universal: don’t send from the value you captured at detection — assert from the value you re-fetched at send time. Align the number in the message with the basis on which you decided to send. Let them drift and you ship “the body says one price, reality is another.”

Advance “notified” only on sent

The other half of idempotency is a rule: update last_notified (the baseline for the last alert sent to this watch) only on sent. Never on dropped or retry.

If a drop or retry moved the baseline, the next legitimate detection would be wrongly rejected as “already notified.” Don’t stamp “sent” on something you didn’t send; only advance when you actually delivered. That’s how “send exactly once, but things you couldn’t send still flow to the next round” both hold.

Retry transient failures with exponential backoff, fail at the cap

Rate limits on the re-fetch, network errors, and transient delivery failures aren’t discarded — they drop to retry. Bump retry_count, wait with exponential backoff, and terminate at failed at the cap.

export function backoffMs(retryCount: number): number {
  return Math.min(2 ** retryCount * 30_000, 30 * 60_000); // 30s, 60s, 120s … capped at 30 min
}

One implementation compromise here. Rather than add next_attempt_at / last_error columns, the retry schedule rides inside the payload JSON under a _retry key — the detection snapshot (listingId, etc.) stays immutable while the schedule piggybacks on it. The fetch queue coarsely pulls pending/retry in SQL and filters out any whose _retry.nextAttemptAt hasn’t arrived, app-side. Backoff is respected without growing the schema.

export async function pendingOutbox(db, limit, now) {
  const rows = await db.prepare(
    `SELECT * FROM outbox WHERE status IN ('pending','retry') ORDER BY created_at ASC LIMIT ?`
  ).bind(limit * 2).all();
  const ready = rows.results.filter((r) => {
    if (r.status === 'pending') return true;
    const meta = parseDetectedPayload(r.detected_payload_json)._retry;
    return !meta?.nextAttemptAt || meta.nextAttemptAt <= now; // exclude not-yet-due backoff
  });
  return ready.slice(0, limit);
}

Wrapping up

The more you want a notification to be correct, the more it paid off to stop “sending on detection” and put an outbox in between. Enqueue idempotently, send only what passed a pre-send re-check, and advance “notified” only when you actually delivered. Collapsing the states into a one-way pending → sending → sent / dropped / retry → failed makes “shouldn’t send, so didn’t (dropped)” and “want to send, haven’t yet (retry)” distinct in both code and operations.

How this outbox gets driven every minute is in running many jobs on one Workers cron, and the delivery channel (login-less Web Push) is in delivering Web Push by anonymous device token. The running app is Yasugoro.

FAQ

Why enqueue into an outbox instead of sending on detection?

Because the situation changes between detection and delivery, and the alert goes stale. You detect that a condition holds, but in the tens of seconds to minutes before you actually send, stock can vanish or the condition can revert. Send anyway and you deliver a notification that no longer applies. Letting detection only append one outbox row, then re-fetching and re-checking right before sending, lets you drop the stale ones instead of sending them.

What stops the same notification from firing twice?

The enqueue is idempotent. It runs INSERT … ON CONFLICT DO NOTHING against a UNIQUE(watch_id, trigger, date(detected_at), channel) expression index, so duplicates for the same day, watch, trigger, and channel collapse to one row. On top of that, the notified baseline (last_notified) advances only on sent, so a dropped or retried row never blocks the next legitimate detection.

What happens if the condition breaks right before sending?

It’s dropped (a normal terminal state). If the pre-send re-fetch shows stock gone or the detection condition no longer holding, the row is dropped as condition_lost and the notified baseline is left untouched. The message body is built from the freshly re-fetched values, not the ones captured at detection, so the number you assert and the basis for sending are the same current value.

How are transient send failures handled?

As retry. Rate limits on the re-fetch, network errors, and transient delivery failures bump retry_count and wait with exponential backoff (30s → 60s → … capped at 30 min), terminating at failed once it hits OUTBOX_MAX_RETRY. Rather than add a next_attempt_at column, the retry schedule rides inside the payload JSON under a _retry key, and the fetch queue filters out retries whose scheduled time hasn’t arrived.

One Cloudflare Workers cron, many jobs — a per-minute tick with time branching

Sat, 25 Jul 2026 00:00:00 GMT

The first wall I hit building scheduled work on Cloudflare Workers was the free plan’s cron trigger cap: five per account. Cut one cron per purpose within a single service — patrol, send, daily maintenance — and you reach five fast. Want another service on the same account, and there’s nothing left.

I built and shipped a price watcher called Yasugoro on Workers + D1. Its scheduled work spans wildly different cadences: sending notifications every minute, patrolling every five, running maintenance once a day. To keep all of it under five crons — and shareable with other services — I collapsed cron into one per-minute schedule that branches on time inside the tick. Here’s how it’s wired.

One per-minute cron, branching on time inside the tick

The move is simple: register a single * * * * * (per-minute) cron, and inside the scheduled() handler check the current time to decide what runs this minute. No new cron per purpose.

// the only cron expression (per minute); keep it in sync with wrangler.toml crons
export const CRON_TICK = '* * * * *';

export async function runScheduledTick(env: Env, now: Date): Promise<void> {
  if (isPatrolTick(now))      await runPatrol(env, now);       // every 5 min
  if (isKaidokiEveTick(now))  await runKaidokiEve(env, now);   // JST 23:50
  if (isKaidokiDayTick(now))  await runKaidokiDay(env, now);   // JST 00:00
  if (isHousekeepingTick(now)) {
    await runHousekeeping(env, now);                            // JST 03:30
    await runSaleIngest(env, now);
  }
  if (isRankingIngestTick(now)) await runRankingIngest(env, now); // JST 04:00
  await runDispatchCron(env, now);                              // every minute (send)
}

Each per-minute tick runs only the jobs whose predicate matches that minute. Six distinct schedules, and the cron budget spent stays at one. scheduled() sees it through with ctx.waitUntil(runScheduledTick(env, new Date())).

flowchart TD
  Cron["cron * * * * * one per minute"] --> Tick[runScheduledTick now]
  Tick --> Dispatch[send: every minute]
  Tick --> Patrol[patrol: minute%5==0]
  Tick --> Eve[buy-timing eve: UTC 14:50]
  Tick --> Day[buy-timing day: UTC 15:00]
  Tick --> House[maintenance+ingest: UTC 18:30]
  Tick --> Rank[ranking ingest: UTC 19:00]

Write the branch predicates in UTC

Time branching is written with Date’s UTC methods. The time handed to scheduled is UTC, so writing it as if it were local drifts. Jobs that should run on Japan time are converted as JST = UTC+9.

export const isPatrolTick      = (n: Date) => n.getUTCMinutes() % 5 === 0;
export const isKaidokiEveTick  = (n: Date) => n.getUTCHours() === 14 && n.getUTCMinutes() === 50; // JST 23:50
export const isKaidokiDayTick  = (n: Date) => n.getUTCHours() === 15 && n.getUTCMinutes() === 0;  // JST 00:00
export const isHousekeepingTick = (n: Date) => n.getUTCHours() === 18 && n.getUTCMinutes() === 30; // JST 03:30

“Every five minutes” is getUTCMinutes() % 5 === 0; a specific time is an hour-and-minute match. Pulling each predicate into its own function keeps runScheduledTick readable — a list of “if it matches, call it.” Tests can pass new Date('...Z') and check each branch in isolation.

Reliability is the same as several dedicated crons

You might worry that collapsing to one per-minute cron makes daily jobs flaky. It doesn’t — because the dependency is identical.

“A dedicated cron that fires at UTC 18:30” and “a per-minute cron whose UTC 18:30 firing you check” both depend, equally, on Cloudflare firing the cron for that minute. If Cloudflare drops a firing, the dedicated cron is missed too; if the per-minute tick arrives, the time check holds. Collapsing adds no new uncertainty.

The one thing that matters operationally is keeping wrangler.toml’s [triggers] crons and the cron constant in code character-for-character identical. Drift there and it either doesn’t fire or registers more than you meant. yasugoro keeps a single test that cross-checks the two so drift fails the build.

Fit each tick under the free-tier subrequest limit (50)

The free plan has a separate limit: at most 50 outgoing fetches (subrequests) per invocation. Collapsing crons doesn’t help here — if one tick’s work exceeds 50, it fails. So for each job you back the batch size out of “how many subrequests does one item cost?”

Patrol: roughly one request per item, so a batch of 40 (PATROL_BATCH_SIZE) leaves headroom under 50.
Send: one item costs a pre-send re-check plus a delivery — up to two subrequests. So the batch is capped at half the send constant (PUSH_BATCH_SIZE / 2).

// send: one item = re-check + deliver = 2 subrequests -> halve the batch
const batchSize = Math.floor(PUSH_BATCH_SIZE / 2);
await runDispatch(deps, batchSize, now);

A job that can’t clear its targets in one batch rotates them. Yasugoro’s ranking ingest splits its target categories into a few per tick, so firing at the same time each day still keeps one run to a handful of categories and sweeps the whole set over several days (effectively weekly). You clear the 50-subrequest wall by not doing everything at once.

Resolve job dependencies by call order

When jobs within a tick depend on each other, the order of calls in runScheduledTick resolves it. Yasugoro puts sending last.

That way notifications enqueued by an earlier job in the same tick (buy-timing, say) get picked up by the same run’s send, without waiting for the next tick. Reverse the order and those enqueued items sit for at least a minute. Splitting the heavy daily work — maintenance and ranking ingest — across different times (UTC 18:30 and 19:00) comes from the same instinct: don’t concentrate load and subrequests into one tick.

Wrapping up

When you want to run several schedules on Cloudflare Workers without eating the cron cap, collapsing to a single per-minute cron that branches on UTC time inside the tick was the straightforward path. Reliability matches lining up dedicated crons, and you spend one trigger instead of many. What actually bites is the other free-tier limit — 50 subrequests per run — which you handle by backing out batch sizes and not doing everything at once.

The send queue this cron drives every minute — the idempotent-notification outbox state machine — is in idempotent notifications with an outbox, and how the notifications get delivered without any login is in delivering Web Push by anonymous device token. The running thing is Yasugoro, which you can poke at with no login.

FAQ

Why collapse into one cron instead of several dedicated ones?

Because Cloudflare Workers’ free plan caps cron triggers at five per account. Run several services on one account and a single service burning five triggers leaves no room for the rest. yasugoro registers exactly one per-minute cron (* * * * *) and, inside it, checks the UTC time to dispatch each job. It spends only one trigger, so other services can share the budget.

Does a per-minute tick with time branching run daily jobs reliably?

Yes. “A dedicated cron fires at that minute (say UTC 18:30)” and “a per-minute cron fires at that minute and you check the time” depend on the exact same thing: Cloudflare firing the cron for that minute. So daily-job reliability equals a dedicated cron — if a firing is dropped, both are affected the same way. Write the predicates with getUTCHours() / getUTCMinutes(), converting JST as UTC+9.

How do you stay within the free-tier subrequest limit?

Outgoing fetches (subrequests) per invocation are capped at 50 on the free plan. Count how many subrequests one item costs, then back out the batch size. yasugoro patrols 40 items per batch, and for sending — where one item costs a re-check plus a delivery, i.e. two subrequests — it caps the batch at PUSH_BATCH_SIZE/2. Any ingest that can’t finish in one pass rotates its targets and covers the whole set over several days.

Can a job enqueued this tick be processed in the same run?

Yes, if you order the calls right. yasugoro runs sending (the outbox dispatch) last in the tick. That way notifications enqueued by an earlier job in the same tick (like buy-timing) get picked up and sent in the same run. Reverse the order and those enqueued items wait until the next tick, so dependent jobs are resolved by call order.

Web Push without login — delivering notifications by anonymous device token

Tue, 07 Jul 2026 00:00:00 GMT

The requirement was “no account, no email,” and Web Push trips on the very first step under it. Before you can send a notification you need something to hold the recipient — but with no login there is no user_id and no email to key on.

I built and shipped a price watcher called Yasugoro — it watches consumables and pushes a notification when one gets cheaper. The headline was “no login, anonymous, no email,” which meant the notification target had to live on an anonymous identifier rather than an account.

This post is only about how to deliver under that constraint, not what to deliver (the buy-timing decision). The anonymous device token, storing the subscription, handling revocation, and the two holes I fell into on real iOS and FCM devices. Once the recipient key is settled, the Web Push part itself is standard plumbing straight out of the spec.

The recipient is an anonymous device token, not email or an account

With no login, the natural identifier is an anonymous token minted per device. On the first visit Yasugoro generates one UUID v4 and stores it as a device token in a cookie.

The cookie is httpOnly + Secure + SameSite=Lax, scoped to the app’s subpath, with a one-year lifetime. Server middleware does exactly one thing: mint the token if the cookie is missing.

export function deviceTokenMiddleware() {
  return async (c, next) => {
    let token = getCookie(c, DEVICE_COOKIE);
    if (!token) {
      token = crypto.randomUUID();            // anonymous UUID v4
      setCookie(c, DEVICE_COOKIE, token, {
        path: '/yasugoro', httpOnly: true, secure: true,
        sameSite: 'Lax', maxAge: 60 * 60 * 24 * 365,
      });
    }
    c.set('deviceToken', token);
    await next();
  };
}

Taking no email means holding no personal data. The target is “this device,” not “this person” — that framing is the spine of the whole design. crypto.randomUUID() is the same API on Cloudflare Workers and Node.

Upsert a user on the device token, then hang subscriptions off it

Once the anonymous token exists, the rest is an ordinary relation. Upsert a users row keyed by a unique device token, and attach the Web Push subscriptions (push_subscriptions) to that user.

// idempotently secure a users row by device_token, return user.id
await db.prepare(
  `INSERT INTO users (device_token, settings_json, plan, created_at, updated_at)
   VALUES (?, '{}', 'free', ?, ?)
   ON CONFLICT(device_token) DO UPDATE SET updated_at = excluded.updated_at`
).bind(deviceToken, ts, ts).run();

A single user carries multiple subscription rows. Pull endpoint / p256dh / auth out of the subscription object that the browser’s PushManager.subscribe() returns, and store them. Register the same watch from a laptop and a phone, and both devices get the push.

Column	Holds	Note
`user_id`	users.id secured via device token	`ON DELETE CASCADE`
`endpoint`	push service URL	`UNIQUE` (one per device+browser)
`p256dh` / `auth`	subscription public key + auth secret	used to encrypt the payload
`ua`	User-Agent	tells iOS PWA apart (deliverability KPI)
`revoked_at`	revocation time	stamped on 404/410, never DELETE

Making endpoint UNIQUE means a device that re-subscribes doesn’t pile up new rows.

flowchart TD
  Visit[First visit] --> Cookie[Mint device_token / cookie]
  Cookie --> Allow[Allow notifications]
  Allow --> Sub[PushManager.subscribe]
  Sub --> Post[POST /api/subscribe]
  Post --> User[users UPSERT by device_token]
  User --> Save[Store in push_subscriptions]
  Save --> Send[Send from Workers with VAPID]

The confirm call, POST /api/subscribe, saves the subscription and registers the watch in one request. The server already has the device token from the cookie, so the client only sends what to watch plus the subscription object.

Send from Cloudflare Workers, with VAPID over webcrypto

Sending happens from Cloudflare Workers, hitting each device endpoint with a VAPID signature. Node’s web-push library doesn’t run on Workers, so I use one that works over webcrypto (@block65/webcrypto-web-push).

There are only three moves: pull the live subscriptions, encrypt the payload and build the headers, fetch the endpoint.

const req = await buildPushPayload(
  { data: toWebPushBody(payload) },   // JSON the SW receives
  toLibSubscription(sub),             // endpoint + p256dh + auth
  { subject, publicKey, privateKey }, // VAPID keys
);
const res = await fetch(req.endpoint, req.init);

The payload is aes128gcm-encrypted with each subscription’s keys. VAPID keys come from env secrets, and the subject is a mailto: or the app URL. That is the whole send path — the notification contents (title, body, target URL) arrive as a Payload assembled upstream. No “when or why to send” logic lives here.

Real-device holes: FCM’s VAPID header and iOS standalone

Code that passed locally failed silently on a real Chrome and a real iPhone — the usual Web Push initiation. Two spots caught me.

FCM rejects the VAPID Authorization header. The library emits the older draft-01 form, Authorization: WebPush <jwt>, but FCM (behind Chrome/Android) rejects it with invalid JWT provided (403). Rewriting it to the draft-02 form, vapid t=<jwt>, k=<publicKey>, got a 201. Apple and autopush accept draft-02 too, so I transform the header right before sending.

// draft-01 "WebPush <jwt>" -> draft-02 "vapid t=<jwt>, k=<key>"
const m = headers[authKey].match(/^WebPush\s+(.+)$/i);
if (m) headers[authKey] = `vapid t=${m[1].trim()}, k=${vapid.publicKey}`;

iOS won’t subscribe unless it’s added to the home screen. Safari on iOS/iPadOS only subscribes to Web Push when the PWA is added to the home screen and launched standalone. In a normal tab, PushManager.subscribe() fails. So the client detects an iOS device that isn’t standalone and shows the add-to-home-screen guide instead of attempting a subscribe.

if (isIOS() && !isStandalone()) {
  showHomeScreenGuide();   // don't attempt subscribe
  return;
}

Skip that branch and iPhone users get the most confusing failure of all: they tap allow and nothing happens. The standalone check reads both matchMedia('(display-mode: standalone)') and navigator.standalone.

Dead subscriptions get parked with revoked_at, not deleted

A subscription whose send returns 404 or 410 gets a revoked_at timestamp instead of a DELETE, because I want to measure deliverability. Hard-deleting would erase any way to count how many issued subscriptions are still alive.

Send results split three ways by HTTP status:

404 / 410 Gone — revoked. Stamp revoked_at and drop it from targets (don’t retry).
429 / 5xx — transient. Mark retryable and hand it back to the upstream dispatcher.
any other 2xx — success.

export function classifyPushStatus(status: number) {
  if (status === 404 || status === 410) return { revoked: true, retryable: false };
  if (status === 429 || status >= 500) return { revoked: false, retryable: true };
  return { revoked: false, retryable: false };
}

The live-subscription query is WHERE user_id = ? AND revoked_at IS NULL, with a partial index on revoked_at so only active rows are read. Send to all of a user’s devices; if even one succeeds, the push counts as delivered, and I only retry when every device failed and the failures were retryable. If only revocations remain, there is no retry — the recipient is gone, so waiting won’t help.

No login doesn’t excuse unguarded POSTs. Yasugoro derives the CSRF token as an HMAC keyed by the device token, so there is no server-side session to hold.

// derive the CSRF token from device_token via HMAC-SHA256 (no server state)
const key = await crypto.subtle.importKey('raw', secretBytes,
  { name: 'HMAC', hash: 'SHA-256' }, false, ['sign']);
const sig = await crypto.subtle.sign('HMAC', key,
  new TextEncoder().encode(`csrf:${deviceToken}`));

On a POST, on top of a same-origin check (Origin/Referer), I compare the submitted _csrf against the token derivable from the cookie’s device token, using a constant-time comparison. Even with no login session, a cookie-bound token confirms that this form came from this device.

Wrapping up

Login-less Web Push mostly dissolves the moment you make the recipient an anonymous device token. From there it’s standard plumbing: hang the subscription off a user, send from Workers with VAPID, park the dead ones. What actually bit on real devices was the other two things — FCM’s header format and the iOS standalone constraint.

What I covered here is only how to deliver, not what or when to deliver — the effective-price comparison and the buy-timing decision live in another layer of Yasugoro. The three-tier effective-price calculation is in the effective-price notes, and how I matched the same product across malls is in the JAN cross-matching post. The schedule that drives this sending — one cron collapsed into a per-minute tick — is in the Workers cron multiplexing post. The running app is Yasugoro — try it with no login.

FAQ

On the first visit I mint an anonymous device token (UUID v4) and store it in an httpOnly cookie. That token is the key for an upserted users row, and Web Push subscriptions (push_subscriptions) hang off that user. There is no email and no account — the only identifier is an anonymous token issued to the device. One user can carry several device subscriptions, so a watch set on a phone also fires on the desktop.

Yes. When the device token changes, the app treats it as a different device and the my-watches screen shows empty. That is the trade-off of having no login: clearing history or switching browsers creates a fresh anonymous user. The token is also mirrored into localStorage to survive some cookie loss, but that is a safety net, not a promise of permanence. For people who need durability there is a separate escape hatch (LINE linking).

Why soft-revoke dead subscriptions with revoked_at instead of DELETE?

When a send returns 404/410 Gone, I keep the row and stamp revoked_at rather than deleting it, because I want deliverability as a KPI. Hard-deleting would make it impossible to count how many issued subscriptions later expired. The active-subscription query filters on revoked_at IS NULL, backed by a partial index so live rows stay cheap to read.

Why don’t notifications arrive on iOS?

Safari on iOS/iPadOS can only subscribe to Web Push when the PWA is added to the home screen and launched standalone. In a normal browser tab the subscribe call fails, so for an iOS device that isn’t standalone I skip the subscribe API entirely and show the add-to-home-screen steps instead. Miss that branch and tapping allow silently does nothing.

How Rakuten's API hides the JAN, and the 3-tier cross-marketplace match

Sat, 27 Jun 2026 00:00:00 GMT

Trying to compare the same product across Rakuten’s and Yahoo!‘s APIs, the first wall wasn’t price — it was identity. Before you can compute which is cheaper, a machine has to decide that this listing on one marketplace and that listing on the other are the same product. Without that, there’s no comparison table to build.

Where it got stuck was the JAN — the 13- or 8-digit barcode number on a product. Yahoo! Shopping looks one up directly with jan_code. Rakuten’s item search API never returns it. So the obvious plan — pull both marketplaces by the same JAN and tie them together in one shot — breaks on one side.

I built and shipped a price watcher on my own: yasugoro, which compares point-inclusive prices across marketplaces and pings you when a watched staple gets cheaper. To make the cross-marketplace part hold together, I split matching into three certainty tiers, assuming up front that JANs won’t always line up. Here’s that design.

Rakuten and Yahoo! are asymmetric about the JAN

The two marketplaces simply differ in how easily you can get a JAN, and designing as if they were symmetric guarantees the Rakuten side falls apart.

Yahoo! Shopping’s item search takes a dedicated jan_code parameter, so you can query by JAN directly, and the response sometimes carries janCode back. If you know the JAN, you can resolve the listing.

Rakuten’s item search has no JAN parameter. You end up putting the JAN string into keyword — an indirect match. Worse, the JAN isn’t a structured field; it tends to be buried inside the item description (itemCaption). I saw exactly this on model-numbered goods like Anker. So on the Rakuten side you always need a follow-up step: did the keyword=JAN hit actually correspond to that JAN’s product?

That asymmetry is the starting point. Yahoo! you can trust into a confirmed JAN match; Rakuten you can’t use without proof.

Collapse it into three certainty tiers

I split product identity into three tiers, ordered by how certain the match is. Instead of forcing every item through a JAN match, only the ones I can tie down with confidence sit at the top.

Tier	Method	Used for comparison	Sends alerts
① JAN match	Yahoo! direct lookup + Rakuten keyword=JAN scored for confidence	Yes (confirmed only)	Yes
② Name fuzzy match	Scored on model no. / size / brand → candidates → user confirms	After user confirms	Confirmed only
③ No match	Treated as single-store (e.g. pasted URL)	No	Own price drops only

The lower the tier, the weaker the evidence, so the right to enter comparison and to fire alerts narrows with it. Only tier ① lands in the table as “the same product to everyone”; ② and ③ are demoted in proportion to how sure the match is.

flowchart TD
  In[Input: JAN / name / URL]
  Kind{Detect kind}
  Jan[① JAN match]
  Score{Confidence score}
  Fuzzy[② Name fuzzy match]
  Cand{Candidates?}
  Single[③ Single-store]
  Cross[Enters comparison, can notify]
  Pending[Candidates, awaiting user confirm]
  Watch[Track own price drops only]

  In --> Kind
  Kind -->|JAN| Jan
  Kind -->|name| Fuzzy
  Kind -->|URL| Single
  Jan --> Score
  Score -->|high ≥0.8| Cross
  Score -->|mid ≥0.5| Pending
  Score -->|low <0.5| Single
  Fuzzy --> Cand
  Cand -->|yes| Pending
  Cand -->|no| Single
  Pending -->|confirm| Cross
  Single --> Watch

A JAN input enters at ①, a name at ②, a URL at ③ — and a weak ① hit falls through to ② or ③. It’s a two-stage funnel, not a single gate.

① JAN match: prove Rakuten hits by scoring “JAN-ness”

Even inside the JAN tier, Rakuten needs proof that a keyword=JAN hit is really the same product. I made that an additive confidence score.

Yahoo! comes in via jan_code lookup, so it goes straight to confirmed (alerts allowed). Rakuten is the awkward one — each hit accumulates a score:

JAN string appears in the item name or caption … +0.5 (the most direct evidence)
price falls inside the product-search API’s price range … +0.3 / outside it … −0.4
a model number or size can be read off the hit … +0.2

Thresholds split the score three ways. ≥0.8 is confirmed (jan, can notify), ≥0.5 is needs-review (demoted to fuzzy, no notify), and anything under 0.5 is discarded (kept out of comparison). Refusing to treat a weak hit as confirmed turned out to be the single most effective guard against wrong merges.

// Rakuten single-hit JAN confidence (excerpt)
let s = 0;
if (hit.itemCaption.includes(jan) || hit.itemName.includes(jan)) s += 0.5;
if (range) {
  const lo = range.minPrice * 0.7;   // allow 30% below (old models, parallel imports)
  const hi = range.maxPrice * 1.5;   // allow 50% above (shipping-included, bundles)
  if (hit.itemPrice >= lo && hit.itemPrice <= hi) s += 0.3;
  else s -= 0.4;                      // far outside = likely a bad match
}
if (modelOrCapacityConsistent(hit)) s += 0.2;

The price range comes from Rakuten’s product search API (productCode=JAN), which returns a min and max price. I allow 0.7× below for old stock and parallel imports, and 1.5× above for shipping-included or bundled listings. A price far outside that band gets penalized — it’s probably a different size, a different color, or another product entirely.

The catch: in practice that product search API returned 404 a lot. The price range is missing for many hits, so I can’t lean on the range term. To cover that, I added a check anchored on the cross-marketplace counterpart — the same-JAN Yahoo! hit’s representative price. If a model-number token pulled from the Yahoo! side (alphanumeric, 5+ chars, so effectively unique) appears in the Rakuten hit’s name, and the price is within ±40% of the counterpart’s price, the hit is promoted to confirmed. The model token alone could collide by chance; the price band screens that out. Both conditions have to hold.

You can see the comparison screen running in yasugoro’s live app — free, no login. When the same product lines up across marketplaces, that’s a tier ① confirmation.

② Name fuzzy match: stop at candidates, don’t auto-confirm

When all you have is a product name, I score it on model number, size, and brand and stop at surfacing candidates. Nothing auto-confirms as the same product.

The fuzzy score breaks down like this — the model number is the strongest signal by far:

model number matches … +0.5
brand matches … +0.2
size matches … +0.2
overlap of remaining tokens (Jaccard) … +0.1 × coefficient

A model number like EH-NA0J is pulled from the name by regex, and a size like 54 sheets × 3 is decomposed into “54 per pack × 3 packs = 162 total” before comparing. Only candidates scoring ≥0.6 are surfaced; if none clear it, the single best one drops to tier ③.

The person confirms the candidate, not the algorithm. Only what a user marks as “same product” gets mastered. A wrong pairing can be rejected in one tap, which removes that pair from the candidate pool for good (it won’t resurface). For cross-marketplace comparison of household staples, putting a plain confirm path and a plain undo path in front of the user beats endlessly chasing higher auto-match accuracy.

③ No match: fall back to single-store drop alerts

Anything that couldn’t be tied down at any tier becomes a single-store item. It stays out of the cheapest-across-stores ranking, and only its own page’s price drops are tracked.

A single item added by pasting a URL starts here from the outset. It isn’t on the comparison footing, so it never claims “cheaper than another store” — but it can still say “cheaper than before.” I’d rather do the honest, smaller thing than force an incomparable item into the ranking and emit a wrong “cheapest.”

What stops wrong merges is “only confirmed tiers notify”

The real reason for three tiers was to never fire an alert on a bad identity. Notifications can only fire on JAN-confirmed (jan) and user-confirmed (user_confirmed) matches. The fuzzy needs-review state and single-store items are both kept out of the trigger condition.

Locking the right to notify behind identity certainty — before any price math — matters because otherwise a price drop on a different product gets reported as “your watched item got cheaper.” The trust in a cross-marketplace comparison rests, in the end, on that identity footing.

How I compute the point-inclusive real price (splitting price into three certainty layers) is its own post: the effective-price design notes. That one is “how the price is computed,” this one is “how the same product is identified” — you need both for the comparison to hold. And once something gets cheaper, “how the alert is delivered” (login-less Web Push) is split into delivering Web Push by anonymous device token. The running thing is yasugoro if you want to poke at it.

FAQ

Can you not get the JAN from Rakuten’s API at all?

There’s no JAN field in the item search response and no JAN-specific search parameter. In practice you throw the JAN string into the keyword parameter, and the JAN itself usually sits inside the item caption rather than a structured field. Yahoo! Shopping, by contrast, has a jan_code parameter that looks it up directly and returns janCode in the response — and that asymmetry is where the whole cross-marketplace problem starts.

Why not just match everything by JAN?

Force a JAN match on items where the JAN isn’t available and you start merging different products as one — the failure I most wanted to avoid. Rakuten doesn’t return the JAN directly, so a keyword=JAN hit needs separate proof that it’s really that product. Only items I can tie down by JAN go into the confirmed tier; weaker evidence drops to candidate suggestions, and unmatchable items drop to single-store alerts.

Why doesn’t the fuzzy match confirm automatically?

A matching model number, size, and brand still only give you a probability, and auto-confirming a guess means alerts fire on a wrong merge. The fuzzy tier surfaces candidates and only masters the ones a person confirms; a wrong pairing can be rejected in one tap, which permanently drops that pair from the candidate pool. Notifications fire only on JAN-confirmed and user-confirmed matches.

What happens to items that don’t match?

They become single-store items: dropped from cross-marketplace comparison, with only their own page’s price drops tracked. Anything you add by pasting a URL starts here too. They never enter the cheapest-across-stores ranking — they’re watched as a standalone price, and alerts only say “cheaper than before”, never “cheaper than the other store”.

How I compute the real, point-inclusive price — design notes from a 3-layer model

Mon, 15 Jun 2026 00:00:00 GMT

I built and shipped a price watcher on my own. It compares the real price — sticker price minus the reward points you earn — across marketplaces, and pings you when a watched staple gets cheaper. Going in, I figured “display price minus points” was a one-liner. The moment I started writing it, that one line turned out to be the hardest part.

Three reasons. Subtract every point and you inflate the number. Mix points you’ll definitely get with points you might get, and the listing that looks cheapest turns out to be conditional. Reward rates use different units per marketplace, and get the rounding wrong and you land a few yen off the real grant.

So here are the design decisions I landed on for that real price. In short: split points into three certainty layers, floor each layer before summing, and rank on the certain layer only. If you’re computing “which is cheaper once points are in” yourself, maybe these help.

First, define the real price

A real price is the sticker price minus the points that purchase earns you. A ¥1,000 item that returns 100 points is ¥900 real. It’s closer to what actually leaves your wallet than the sticker number is.

Points aren’t quite cash, though — they expire and they’re restricted in where you can spend them. So I fixed an assumption up front: one point equals one yen on your next purchase. Without that assumption nailed down, a cross-store comparison doesn’t even hold together. Folding the assumption into the definition was where the design started.

Where the naive version broke

The first thing I wrote was this:

const effectivePrice = displayPrice - totalPoints; // this inflates

Short, and it had three holes.

It subtracts points you might not get. Fold in entry-required campaigns or your assumed loyalty multiplier, and the cheapest-looking store becomes “the price you reach only if every condition is met.”
Reward rates use different units per marketplace. Rakuten’s multiplier, Yahoo’s campaign add-ons, limited-time points. Multiply them in one pass and the magnitude drifts.
Rounding. Points usually accrue floored. Sum several rewards and then floor, and you land a few yen off the real grant.

A few yen is enough to swap first and second place — which changes the answer to “where should I buy.” Don’t inflate, and don’t drift. A real price has to do both, or it decays into a rough hint. So I rebuilt it.

The core decision: split points by certainty into three layers

The axis I settled on was splitting points by how certain you are to receive them into three layers.

Layer	Contents	Certainty	Used for ranking
① Confirmed	Actual points the API returns	Certain	Yes
② Conditional	Known campaigns (e.g. “5 and 0 days”)	Varies (entry, etc.)	No
③ Assumed	Your own loyalty multiplier / shop-hopping	Self-reported	No

Rank on the confirmed layer alone, and show ② and ③ as “the most it could drop.” The sort order is the same for everyone — decided by certain information — while the upside flexes with each person’s conditions.

flowchart TD
  Price[Sticker price incl. tax]
  L1[① Confirmed: API points]
  L2[② Conditional: known campaigns]
  L3[③ Assumed: user settings]
  Rank[Ranking rankKey = price − ①]
  Eff[Displayed real price = price − ①②③]

  Price --> L1 --> Rank
  Price --> L2 --> Eff
  Price --> L3 --> Eff
  L1 --> Eff

Each result carries the weakest layer it includes as a certainty tier. In types, that’s all it is:

type CertaintyTier = 'confirmed' | 'conditional' | 'assumed';

A rule hangs off that: a price that includes layer ③ (assumed) never fires a notification. You can’t tell someone “buy now” on the strength of self-reported points. Notifications fire on confirmed only — or on conditional once the entry requirement is disclosed.

Floor each layer before summing

Don’t floor once at the end across layers. Floor inside each layer, then add.

let pts = Math.floor(base * rate); // floor per layer
if (cap != null && pts > cap) {
  pts = Math.min(pts, cap); // cap applies to floored point-yen
}

Real grants are usually floored per campaign too, so summing and flooring afterward lands a yen or two off. The cap behaves the same way: take Math.min against the floored point-yen. Cap first, floor second, and you’re off by one again. Unglamorous ordering, but this is what the displayed number’s credibility rested on.

There’s a quirk in how rates are stored, too. A campaign’s “+4×” is held as the added grant on top of the tax-inclusive price — an increment. The base 1× is increment zero. Multiply the price by the raw multiplier and you double-count the standard points everyone already gets. So I store only the increment as a rate.

Separate the ranking key from the displayed price

The key that sorts and the real price I show subtract different layers.

const rankKey = base - confirmedPoints;                 // ① only
const total = confirmedPoints + conditionalPoints + assumedPoints;
const effectivePrice = Math.max(0, base - total);       // negative guard

rankKey subtracts the confirmed layer only, so the order is identical for everyone. effectivePrice subtracts every layer and represents “the most this could drop” with your conditions folded in. On high-reward items the total points can exceed the price and go negative, so Math.max(0, …) guards it.

I pulled the order apart from the displayed amount on purpose: numbers that vary by reader (conditional, assumed) shouldn’t move the ranking. Let them in and you get the queasy behavior where whoever inflated their loyalty settings most reshuffles the top.

Leave shipping out of the real price, on purpose

No marketplace search API returns the shipping amount, so I decided to keep it out of the real price. Add a guess and you import error to paper over data you don’t have.

Instead, carry a flag — shipping-included, shipping-separate, or unknown — and compare only shipping-included listings on the same footing.

// only shipping-included listings enter the main ranking
function isComparableForRank(listing: Listing): boolean {
  return listing.postageFlag === 0; // 0=included, 1=separate, null=unknown
}

Mix separate-shipping and included-shipping listings into one “cheapest” and you’ll miss the ones that flip once shipping lands. Deciding not to include data you can’t get — and surfacing that decision — is more honest than half-guessing it.

Looking back

A real price reads as “sticker minus points” in one line, yet making it trustworthy stacked up quiet decisions: split by certainty, floor per layer, separate the ranking key, drop shipping. None of it is flashy. Do it sloppily, though, and “the cheapest one was actually conditional” costs you the reader’s trust. Most of the time I spent went into making this calculation’s footing solid.

This three-layer calculation runs on watched staples every day, pinging only when the confirmed layer says something genuinely got cheaper. The watcher I built around it — now live — is yasugoro: free, no login required.

FAQ

Why not just subtract all the points and compare that number?

Subtracting everything inflates the number. A price with entry-required campaigns and your own loyalty multiplier baked in is “the price you reach if every condition is met” — not a floor everyone can count on. Rank on the points the API actually confirms, and show conditional and assumed points as a separate range.

How do you handle limited-time points and entry-required campaigns?

They don’t go into the effective-price total. They live in a conditional layer, computed separately and flagged for “entry required” or “cap reached”. Only API-confirmed points are asserted as a firm amount. Conditional points are shown as “this is the most it could drop”, and they’re excluded from what triggers a notification.

If shipping isn’t in the price, won’t the cheapest option flip?

It can. But no marketplace search API returns the shipping amount, so adding a guess just imports error. I kept shipping out of the effective price and instead carry a flag — shipping-included, shipping-separate, or unknown — comparing only shipping-included listings on the same footing. Separate-shipping listings drop out of the main ranking and are disclosed in a note.

Why use different layers for ranking versus the displayed price?

The sort order should look the same to everyone. The ranking key uses only the confirmed layer, while the displayed effective price subtracts confirmed plus conditional plus assumed. That way “cheapest first” is decided by certain information alone, and the per-person upside shows up as a range on the displayed price.

Generating llms.txt and llms-full.txt in Astro for a Bilingual Site

Sun, 14 Jun 2026 00:00:00 GMT

llms.txt is a proposed entry point for letting AI search and LLMs read a site. It sits in the same spot as robots.txt and sitemap.xml — a plain-text file at the site root. The spec isn’t settled, and no major LLM vendor officially promises to consume it. I still generate one, because the cost of doing so is small.

This post is the implementation record for generating llms.txt and llms-full.txt in Astro, split into English and Japanese. I’m not going to argue whether it works — I have no measurable traffic to point at and nothing to assert. This is only about what gets generated and how. It continues the same thread as enforcing rules with a Zod schema and reciprocal hreflang: treat frontmatter and Content Collections as the single source of truth.

What llms.txt actually is

llms.txt is a Markdown index for LLMs, placed at the site root. Where sitemap.xml is a machine-readable list of URLs, llms.txt describes — in natural language, with one-line notes — what the site is and where to start reading.

The llmstxt.org spec sets a loose shape: an H1 site name and a blockquote summary at the top, then H2 sections of annotated links below. Each link follows - [Title](URL): one-line note.

The spec also names a second convention, /llms-full.txt. That one isn’t an index but an expansion of each page’s body, so an LLM can grasp the content from a single file without following links. Aulvem ships both.

	llms.txt	llms-full.txt
Role	Light index (links + notes)	Full-text-ish index (excerpts + headings + tags)
Posts	Capped at seven featured	Every post
Language	Split per language (EN / JA)	One file, EN + JA mixed, each entry language-tagged
Size	Small	Large

The minimal Astro version

Astro’s file-based API routes let you return text by dropping a .txt.ts file under src/pages/. Return a Response from a GET handler with a text/plain content type.

// src/pages/llms.txt.ts
import type { APIContext } from "astro";
import { renderLlmsTxt } from "../lib/llmsTxt";

export async function GET(_context: APIContext) {
  const body = await renderLlmsTxt({ docLang: "en" });
  return new Response(body, {
    status: 200,
    headers: {
      "Content-Type": "text/plain; charset=utf-8",
      "Cache-Control": "public, max-age=3600",
    },
  });
}

The .txt.ts extension is what matters: it builds to the URL /llms.txt. I keep the generation logic in a separate file (src/lib/llmsTxt.ts) and leave the route as a thin entry point, so a per-language endpoint can reuse the same logic.

The stack is short.

	Package	Use
	`astro` (API routes + `astro:content`)	Route definitions and Content Collections access
	`typescript`	Typing the renderer

Driving it from Content Collections

The post list inside llms.txt isn’t written by hand. It’s built on the fly from Content Collections — getCollection for posts and products. A hand-kept list goes stale: you add a post, forget the index, and llms.txt drifts from the actual content.

// src/lib/llmsTxt.ts (excerpt)
export async function renderLlmsTxt(opts: LlmsTxtOptions): Promise<string> {
  const services = await getCollection("services");
  const blog = await getCollection("blog", ({ data }) => !data.draft);

  // newest first; products order by status (active → preparing → archived)
  blog.sort((a, b) => b.data.pubDate.getTime() - a.data.pubDate.getTime());
  // ...assemble sections and return join("\n")
}

The detail to get right is the ({ data }) => !data.draft filter that drops draft: true. Skip it and a half-written draft lands in llms.txt, advertising a URL you haven’t published. I keep the same exclusion that sitemap and RSS use.

For each post’s note, I reuse the frontmatter summary (falling back to description). Every post already carries a summary written as the short answer I want quoted by AI, so the llms.txt note and the post’s own TLDR come from one place. Frontmatter stays the single source here too.

Splitting by language

This is the part that matters for a multilingual site. Aulvem drives the renderer on two axes:

filterLang: which language’s posts and products to list (ja lists only Japanese entries)
docLang: which language the document’s own text — headings, summary, usage terms — is written in

Separating those two lets one renderer emit three endpoints.

// src/pages/llms.txt.ts        → English headings, posts from all languages
renderLlmsTxt({ docLang: "en" });

// src/pages/ja/llms.txt.ts     → Japanese headings, Japanese posts only
renderLlmsTxt({ filterLang: "ja", docLang: "ja" });

The English route /llms.txt leaves filterLang unset on purpose. I treat the English version as the whole-site entry point, so it can surface posts in either language. The Japanese version /ja/llms.txt closes to the Japanese surface with filterLang: "ja".

flowchart LR
  CC[Content Collections<br/>blog + services]
  R[renderLlmsTxt<br/>filterLang / docLang]
  EN["/llms.txt<br/>docLang=en"]
  JA["/ja/llms.txt<br/>filterLang=ja, docLang=ja"]

  CC -->|getCollection| R
  R --> EN
  R --> JA

Featured posts narrow to docLang

One design call sits inside this. The English /llms.txt can surface posts from both languages, but the Featured posts section alone is narrowed to docLang.

// featured posts only in the doc language
const featuredSource = filteredBlog.filter(
  (p) => entryLangLocal(p.id) === opts.docLang,
);

Listing both halves of a translation pair in the featured slots spends two slots on one piece of content and halves the unique signal in a limited list. So the featured slots align to the document language, and the cross-language full set lives in llms-full.txt. A bounded list (seven featured posts) narrows by language; a full dump with loose size limits carries both. That’s the split.

llms-full.txt — full index with language cross-references

llms-full.txt is a separate route (src/pages/llms-full.txt.ts) that expands every post’s excerpt, headings and tags into one file. The MDX body runs through stripMdx to drop the syntax, then clip to 500 characters.

Each entry gets a language tag ((ja) / (en)), and if a translation pair exists, a Lang-Alt line cross-references the other language’s URL.

// add the paired language version as Lang-Alt
const alt = altByBase.get(slug);
if (alt) {
  const otherLang = lang === "ja" ? "en" : "ja";
  const otherPath = alt[otherLang];
  if (otherPath) {
    lines.push(`Lang-Alt (${otherLang}): ${SITE}${otherPath}`);
  }
}

This is the same idea as hreflang. I wrote up emitting reciprocal alternate links in HTML only for posts that exist in both languages in the hreflang post; llms-full.txt makes the same statement to an LLM — “this post has a paired version over here.” The aim is to keep the English and Japanese halves of one post from being counted as two unrelated documents.

How it sits next to robots.txt

llms.txt doesn’t replace robots.txt. robots.txt states whether a crawler may fetch; llms.txt states what the site contains and where the good parts are. The final permission to train or quote stays with robots.txt.

Aulvem puts a usage-and-citation section inside llms.txt: quoting is welcome, link back to the source URL, training is permitted for now. The section ends by pointing back to robots.txt as the authoritative bot policy — the machine-readable authority lives there, and llms.txt is a note, not an override. I’d keep that framing.

Leave it vague and you get a mismatch: llms.txt says “no training” while robots.txt allows everything. The two files’ stances have to be kept aligned by hand.

Pitfalls and operating rules

A few things I nearly tripped on:

Apply the draft exclusion across all three routes: put the getCollection filter in the shared renderer so llms.txt, ja/llms.txt and llms-full.txt all pass the same exclusion. Let one route through raw and drafts leak
Don’t conflate filterLang and docLang: “which posts to list” and “what language to write in” are different axes. The English version leaves filterLang unset to surface Japanese posts too — don’t read that later as a bug and add a filter
Featured language narrowing: only Featured narrows by docLang. Drop that and translation pairs fill the slots
Set Cache-Control: all three use max-age=3600. Generation is cheap, but there’s no need to rebuild on every request
Keep usage terms and robots.txt in sync: change the stance and edit both. Don’t let llms.txt stand alone

Closing

llms.txt, generated from Content Collections as an Astro API route, drops the manual upkeep every time you add a post. For multiple languages, splitting on filterLang (which posts) and docLang (the wording) lets one renderer emit the English, Japanese and full-text versions.

Whether it does anything is a separate question I’m leaving alone. How llms.txt feeds AI-search traffic isn’t a settled convention or a measured one yet. Aulvem’s stance is to generate it because it’s cheap, and to keep the authority in robots.txt.

Treating frontmatter as the single source of truth is the thread that runs from the Aulvem blog architecture post onward, if you want the wider context.

Counting which AI clients actually came for it is a separate problem, because the User-Agent they present cannot be taken at face value. The split I use is in AI crawler user agents are self-reported.

FAQ

What’s the difference between llms.txt and llms-full.txt?

llms.txt is a light index — annotated links that say what the site is and where to start. llms-full.txt carries each post’s excerpt, headings and tags as a closer-to-full-text dump. Aulvem treats the first as a curated list capped at seven featured posts, and the second as a full excerpt dump of every post.

Will publishing llms.txt get my site read by AI and bring traffic?

I can’t claim that. llms.txt is a proposed convention, and no major LLM vendor officially guarantees it consumes the file. robots.txt and sitemap.xml are still the authoritative bot signals. Aulvem keeps llms.txt because it’s cheap to generate, not because there’s measurable traffic to point to.

Should a multilingual site split llms.txt per language?

Aulvem serves a Japanese version (/ja/llms.txt) separate from the English root (/llms.txt). I want the headings and usage terms in the reader’s language, and the featured posts narrowed to that language. The full-text llms-full.txt stays as a single bilingual file, kept as a separate track from the two light indexes.

If I already have robots.txt, do I need llms.txt too?

They do different jobs. robots.txt states what crawlers may fetch; llms.txt describes what the site contains and where to read. The authoritative permission for training and citation lives in robots.txt — llms.txt is a guide on top of it. Aulvem’s usage section even points back to robots.txt as the final word.

Emitting reciprocal hreflang only when both languages exist — Aulvem's i18n customisation

Sat, 30 May 2026 00:00:00 GMT

Astro 5’s i18n gives you routing, not hreflang.

A naive “emit one <link rel="alternate"> per supported locale” implementation breaks the moment a post only exists in one language: the alternate points at a 404, Google sees the reciprocal pointer is missing, and the hreflang is quietly ignored. Worst case, it weakens the site’s multilingual signal as a whole.

Aulvem walks Content Collections to detect which language versions actually exist for a given slug, then emits <link rel="alternate" hreflang> conditionally from Seo.astro. This is the third follow-up to How this blog is built, picking up the FAQ note that hreflang and sitemap lastmod were both custom-built on top of Astro’s defaults.

What breaks when you emit alternates for every locale

The naive pattern: a 2-language site emits two <link rel="alternate" hreflang> lines on every page, one for each locale. It looks correct at first.

The break happens when a specific post only exists in one language:

A JA-only post emits a hreflang pointing at the EN URL → that URL returns 404
Google notices the EN URL doesn’t reciprocate (no JA pointer because it doesn’t render)
The hreflang is dropped, and the site’s overall multilingual signal weakens

Google’s hreflang documentation makes both rules explicit: “reciprocal pointers required” and “alternates must point at valid URLs”. A naive implementation fails both and effectively becomes a no-op — or worse, an active liability.

The two rules from Google

Two requirements have to hold for hreflang to count:

Reciprocal pointers: if A → B is emitted, B → A must exist
Valid URLs: 404 or noindex targets disqualify the alternate

The cleanest way to satisfy both is “only emit an alternate when both languages actually have a published entry”. That requires walking the content at build time to know which entries are bilingual.

Three helpers in `src/lib/posts.ts`

Aulvem keeps three helpers for the three page types (single post, product page, paginated category):

import { getCollection } from "astro:content";
import type { Lang } from "../i18n";

const LANGS: Lang[] = ["en", "ja"];

// Per-post: which languages have the entry
export async function blogAvailableLangs(slugWithoutLang: string): Promise<Lang[]> {
  const all = await getCollection("blog", ({ data }) => !data.draft);
  return LANGS.filter((lang) =>
    all.some((e) => e.id === `${lang}/${slugWithoutLang}`),
  );
}

// Per-product: same, against the services collection
export async function serviceAvailableLangs(slugWithoutLang: string): Promise<Lang[]> {
  const all = await getCollection("services");
  return LANGS.filter((lang) =>
    all.some((e) => e.id === `${lang}/${slugWithoutLang}`),
  );
}

// Per-paginated-page: does page N exist in each language?
export async function categoryPageAvailableLangs(
  category: string,
  pageNum: number,
  pageSize: number,
): Promise<Lang[]> {
  const results: Lang[] = [];
  for (const lang of LANGS) {
    const posts = await getPublishedBlog(lang);
    const inCat = posts.filter((p) => p.data.category === category);
    const lastPage = Math.max(1, Math.ceil(inCat.length / pageSize));
    if (pageNum <= lastPage) results.push(lang);
  }
  return results;
}

The draft: true filter inside getCollection matters — treating a draft-only language as “available” would emit a hreflang pointing at a URL that doesn’t render on the published build.

Unlike the sitemap lastmod implementation inside astro.config.mjs, these helpers run from layouts during page rendering. The Content Collections API is already initialised by then, so getCollection is the right tool here.

Conditional emission in `Seo.astro`

Seo.astro accepts an availableLangs prop (defaulting to both) and uses includes("en") / includes("ja") to decide which <link rel="alternate"> to emit:

---
interface Props {
  // ... other props
  availableLangs?: Lang[];
}
const { availableLangs = ["en", "ja"], lang: langProp } = Astro.props;
const lang: Lang = langProp ?? getLangFromUrl(Astro.url);

const currentPath = Astro.url.pathname;
const otherLangPath = altPathForOtherLang(currentPath, lang);
const enPath = lang === "en" ? currentPath : otherLangPath;
const jaPath = lang === "ja" ? currentPath : otherLangPath;
const enHref = new URL(enPath, Astro.site).toString();
const jaHref = new URL(jaPath, Astro.site).toString();
const hasEnVersion = availableLangs.includes("en");
const hasJaVersion = availableLangs.includes("ja");
// x-default points to EN when available; otherwise the only available lang.
const xDefaultHref = hasEnVersion ? enHref : jaHref;
---
{hasEnVersion && <link rel="alternate" hreflang="en" href={enHref} />}
{hasJaVersion && <link rel="alternate" hreflang="ja" href={jaHref} />}
<link rel="alternate" hreflang="x-default" href={xDefaultHref} />

The default value ["en", "ja"] matches the top-level pages (/, /about/, etc.) which are always bilingual. The contract is: per-entry layouts must override availableLangs. Forgetting the override on a per-entry layout means broadcasting a 404 URL through hreflang, which is the exact failure mode the helpers are designed to prevent.

x-default always falls back to EN when both languages exist, since EN is the site’s default locale. Single-language posts still get an x-default pointing at the only available language — Google’s guideline says to declare at least one supported language explicitly.

Wiring layouts to the helpers

BlogPostLayout.astro is the typical caller:

---
import { entrySlug, blogAvailableLangs } from "../lib/posts";

const slug = entrySlug(entry); // "ja/2026-05-30-foo" → "2026-05-30-foo"
const availableLangs = await blogAvailableLangs(slug);
---
<Seo
  title={data.title}
  description={data.description}
  lang={lang}
  availableLangs={availableLangs}
  // ...
/>

entrySlug strips the language prefix (en/ or ja/) before handing the slug to the helper. Forgetting the strip turns the lookup into a full mismatch and silently returns an empty array — quiet failure that doesn’t fail the build.

Paginated routes (/blog/build/[...page].astro) call categoryPageAvailableLangs(category, page.currentPage, 12) instead.

The flow

flowchart LR
  MDX[blog MDX en+ja]
  Helper[blogAvailableLangs<br/>serviceAvailableLangs<br/>categoryPageAvailableLangs]
  Layout[BlogPostLayout<br/>ServiceLayout<br/>category route]
  Seo[Seo.astro<br/>availableLangs prop]
  HTML[hreflang 0~2 + x-default]

  MDX -->|getCollection| Helper
  Helper --> Layout
  Layout --> Seo
  Seo --> HTML

The helper walks Content Collections once and returns a Lang[] to the layout. The layout passes that array to Seo.astro, which emits zero, one, or two <link rel="alternate"> lines plus an x-default.

Routing stays on the official integration

Astro 5’s i18n routing — automatic /ja/... prefixing, locale detection on request — is left untouched. It’s mature enough that there’s no incentive to rewrite it.

Aulvem’s stance is to leave anything the official integration handles well alone and patch only the gaps. Hreflang is purely an SEO-layer concern, so the override lives entirely inside Seo.astro and the helpers. When Astro eventually ships hreflang generation out of the box, peeling the override off is a one-file change.

Pitfalls

A short list of things that almost broke:

availableLangs defaults to ["en", "ja"]: appropriate for top-level pages, but per-entry layouts must override. Forgetting the override is the exact failure mode this whole setup prevents
entrySlug placement: Content Collections IDs are ja/2026-05-30-foo. Hand the prefixed ID to the helper and the lookup quietly returns [], which then emits no hreflang at all — silent regression
draft: true filtering: handled inside the helper’s getCollection filter. Without it, a draft-only language can be marked “available” and the published build has nothing at that URL
Paginated edge case: /blog/build/3/ can exist in EN but not JA if the categories diverge in length. categoryPageAvailableLangs handles this — without it, page 3 of a category would emit JA alternates pointing at 404
x-default for single-language posts: still emitted, pointing at the one available language. Google recommends declaring at least one supported locale explicitly

Wrap-up

Hreflang only counts when reciprocal pointers and existence checks both hold. Walking Content Collections at build time gives both, and the override fits in roughly 30 lines split between three helpers and a Seo.astro change. Routing stays on the official integration.

The same “frontmatter as the single source of truth” idea drives the Zod schema enforcement and the sitemap lastmod override — three follow-ups to the stack overview. The reciprocal-pointer idea carries over to LLM indexes too, in generating a bilingual llms.txt. Pointing a canonical the other way — from someone else’s platform back at your domain — is a separate problem, measured in is cross-posting duplicate content.

FAQ

Why doesn’t Astro emit hreflang on its own?

Astro 5’s i18n is scoped to routing — generating URLs and detecting the request locale. SEO meta-tag generation isn’t part of that responsibility. @astrojs/sitemap will emit hreflang inside the sitemap.xml, but the <link rel="alternate"> in the <head> is your job to write. Aulvem keeps that responsibility inside Seo.astro.

Aren’t the hreflang entries from `@astrojs/sitemap` enough?

The xhtml:link entries inside sitemap.xml are emitted, but the <link rel="alternate" hreflang> in HTML <head> is a separate channel. Google reads both, and missing the <head> channel weakens multilingual recognition at the snippet level. So both belong.

What should `x-default` point to?

Google describes it as the fallback for users whose target language can’t be determined. Aulvem has EN as the default locale, so x-default points at the EN version when one exists, and falls back to the only available language otherwise. For posts that have both languages, defaulting to EN is fine.

Does omitting hreflang for one-sided posts hurt SEO?

Not emitting hreflang isn’t itself a penalty. What hurts is emitting an alternate that points at a non-existent URL — Google flags the reciprocal mismatch and weakens the whole site’s multilingual signal. Single-language posts are safer with a clean canonical and no alternate than with a broken one.

Reading sitemap lastmod from MDX frontmatter — customising Astro's sitemap integration

Mon, 25 May 2026 00:00:00 GMT

@astrojs/sitemap won’t read updatedDate out of MDX frontmatter.

Leave it at the defaults and the sitemap’s lastmod ends up being the build time, which broadcasts the noise signal that “every post was updated on every build” to search engines and AI search. The freshness hint stops being a hint and starts being misleading.

Aulvem solves this inside astro.config.mjs by walking the blog folder, building a path-to-date map with a lightweight regex, and feeding it into the serialize hook of @astrojs/sitemap. Paginated noindex pages are dropped from the sitemap in the same pass via filter. This is the second follow-up to How this blog is built.

Why `lastmod` matters

lastmod is the last-modified date in the sitemap protocol, and search engines use it as a re-crawl priority hint. AI search engines — ChatGPT Search, Perplexity, Claude Search — read it as a freshness anchor for citation eligibility.

Two specific reasons to get it right:

Crawl priority: recently updated URLs should get re-fetched sooner
AI citation: “this article was updated N days ago” feeds into the LLM’s decision to cite the page as current information

A bogus lastmod works the other way. Google’s own Sitemap docs note that unreliable lastmod values erode site-wide trust. Painting build-time onto every URL teaches Google that this site’s lastmod is noise, which is worse than not setting one at all. Pair the implementation with a strict updatedDate policy (see content rules A-2-1) and the signal stays meaningful.

Walking MDX at config evaluation time

The lastmod map is built once, inside astro.config.mjs, by reading every .mdx / .md under src/content/blog/{en,ja}/.

import { readdir, readFile } from "node:fs/promises";
import { join } from "node:path";

async function buildBlogLastmodMap() {
  const map = new Map();
  for (const lang of ["en", "ja"]) {
    const dir = join(process.cwd(), "src", "content", "blog", lang);
    let files = [];
    try {
      files = await readdir(dir);
    } catch {
      continue;
    }
    for (const file of files) {
      if (!file.endsWith(".mdx") && !file.endsWith(".md")) continue;
      const slug = file.replace(/\.(mdx|md)$/, "");
      const raw = await readFile(join(dir, file), "utf8");
      const fm = /^---\n([\s\S]*?)\n---/.exec(raw);
      if (!fm) continue;
      const front = fm[1];
      if (/^draft:\s*true/m.test(front)) continue;
      const updated = /^updatedDate:\s*(\S+)/m.exec(front);
      const pub = /^pubDate:\s*(\S+)/m.exec(front);
      const dateStr = (updated && updated[1]) || (pub && pub[1]);
      if (!dateStr) continue;
      const d = new Date(dateStr);
      if (Number.isNaN(d.getTime())) continue;
      const path = lang === "ja" ? `/ja/blog/${slug}/` : `/blog/${slug}/`;
      map.set(path, d.toISOString());
    }
  }
  return map;
}

const blogLastmod = await buildBlogLastmodMap();

The reason getCollection("blog") isn’t used here: astro.config.mjs evaluates before the content loader is initialised. The Content Collections API simply isn’t available when the config is being read.

Frontmatter is read with a light regex. There’s no need to fully parse MDX — only updatedDate and pubDate matter for the lastmod map, and adding a YAML parser as a dependency for two fields isn’t worth it. Type-safety is sacrificed at this layer, but the Zod schema validates every post during the actual build, so the safety net moves up one step rather than disappearing.

Skipping draft: true posts here is important. If they leak through, draft URLs end up in the sitemap, which is the opposite of what draft: true is supposed to express.

Feeding the map into `serialize`

The serialize option on @astrojs/sitemap is a hook that rewrites each emitted URL entry. The map gets looked up by URL path and the result is dropped into item.lastmod.

sitemap({
  i18n: {
    defaultLocale: "en",
    locales: { en: "en", ja: "ja" },
  },
  serialize(item) {
    const url = new URL(item.url);
    // Strip the /ja/ prefix so /ja/blog/foo/ and /blog/foo/ both
    // hit the same changefreq / priority branch. lastmod lookup
    // still uses the original url.pathname so the map keys match.
    const pathname = url.pathname.replace(/^\/ja\//, "/").replace(/^\/ja$/, "/");
    if (pathname === "/") {
      item.changefreq = "daily";
      item.priority = 1.0;
    } else if (pathname === "/blog/") {
      item.changefreq = "daily";
      item.priority = 0.9;
    } else if (pathname.startsWith("/blog/")) {
      item.changefreq = "monthly";
      item.priority = 0.7;
      const lastmod = blogLastmod.get(url.pathname);
      if (lastmod) item.lastmod = lastmod;
    } else if (pathname.startsWith("/products/")) {
      item.changefreq = "monthly";
      item.priority = 0.8;
    } else {
      item.changefreq = "monthly";
      item.priority = 0.5;
    }
    return item;
  },
});

changefreq and priority go in the same hook so each path category gets a consistent value. priority is officially “ignored” by Google these days, but Bing and the AI crawlers still read it, so I keep it consistent rather than dropping it entirely.

Stripping /ja/ for the branch decision but not for the lookup matters: the lookup needs to keep the language prefix because the map was built with /ja/blog/foo/ and /blog/foo/ as distinct keys.

Filtering out paginated `noindex` pages

Page 2 and beyond of the category listings (/blog/build/2/, /blog/reviews/3/, …) ship a <meta name="robots" content="noindex, follow">. Only page 1 should be indexed; later pages are crawlable but not index-eligible.

Leave those URLs in the sitemap and the signal collides. “Listed in sitemap” reads as “please index this”, which contradicts noindex. Google and Bing both treat that conflict as a quality smell, and there’s no upside to keeping them in.

sitemap({
  filter: (page) => {
    if (page.endsWith("/404/") || page.endsWith("/404")) return false;
    // Paginated category pages (/blog/build/2/, /blog/reviews/3/ ...)
    // are noindex, follow — drop them from the sitemap to avoid
    // contradicting the meta robots tag.
    if (/\/blog\/(build|reviews)\/\d+\/?$/.test(new URL(page).pathname)) return false;
    return true;
  },
  // ...
});

If you set noindex on paginated pages, dropping them from the sitemap is the matching half of the change. Doing one without the other leaves a half-fixed signal.

How the pieces connect

flowchart LR
  MDX[src/content/blog/&lt;lang&gt;/&lt;slug&gt;.mdx]
  Map[blogLastmod Map<br/>path → ISO date]
  Pages[@astrojs/sitemap<br/>URL list]
  Filter[filter:<br/>drop noindex]
  Serialize[serialize:<br/>lastmod / changefreq / priority]
  SM[sitemap.xml]

  MDX -->|fs.readdir + regex| Map
  Map -->|lookup| Serialize
  Pages --> Filter
  Filter --> Serialize
  Serialize --> SM

Two pipelines run inside astro.config.mjs. One walks MDX before the build starts and produces the lastmod map. The other is the standard Astro routing, whose URL list is filtered down and then enriched per-entry in serialize. Both converge into the final sitemap.xml.

Things I almost got wrong

A short list of pitfalls picked up while writing this:

Top-level await: works because astro.config.mjs is evaluated as ESM. Older repo layouts that still use astro.config.cjs won’t accept top-level await, and would need either a .mjs rename or an async wrapper inside defineConfig
draft: true filtering: skipping drafts in the map builder is necessary. Otherwise draft URLs land in the sitemap, which contradicts the whole point of marking them as drafts
Regex tightness: /^updatedDate:\s*(\S+)/m reads unquoted YAML like updatedDate: 2026-05-25. Quoted strings still work because \S+ captures "2026-05-25" whole and new Date() can parse it, but be aware if you ever use multi-line or list-style date values
Per-language folder merge: en and ja are walked separately and merged into a single map. Keys are kept distinct (/blog/<slug>/ vs /ja/blog/<slug>/) so the lookup inside serialize resolves correctly
updatedDate discipline: implementation alone isn’t enough — bumping updatedDate for trivial edits will still poison the signal. The matching policy lives in content rules A-2-1

When this should move back into the official integration

Honestly, this code is the shape of a @astrojs/sitemap feature request. Coupling Content Collections to the sitemap helper is a need other Astro users will have, and there’s room to generalize it as an option.

For now it lives inside Aulvem’s config at ~30 lines, which is a size I’m comfortable maintaining. If the official integration ever ships a Content Collections bridge, the bespoke code drops away. Keeping it isolated to astro.config.mjs makes that removal cheap.

Wrap-up

lastmod is a freshness signal that matters in both classic SEO and AI search, so feeding it real values from updatedDate is worth the 30 lines of bespoke code. Drop paginated noindex pages from the sitemap at the same time — handling them separately leaves the signal half-fixed.

The first follow-up in this series is on using the Zod schema to enforce operational rules at build time → Pushing operational rules into Astro Content Collections with Zod.

FAQ

Why doesn’t `@astrojs/sitemap` read `updatedDate` on its own?

The integration takes the list of built URLs and emits a sitemap.xml — it doesn’t reach into Content Collection frontmatter. Pulling MDX shape into a generic sitemap integration would mean coupling it to collection names, schemas, and loader types, which doesn’t sit well as a default responsibility.

Isn’t reading frontmatter through `fs.readdir` + regex unsafe compared to the Content Collections API?

It’s not type-safe. But astro.config.mjs evaluates before the content loader is initialised, so the API isn’t available at this point. The only fields the lastmod map needs are updatedDate and pubDate, so a light regex pass keeps the dependency surface flat. The real type check runs at build time through the Zod schema, so there are two nets, not one.

Is it really wrong to ship `noindex` URLs in the sitemap?

It’s a contradictory signal. A sitemap is a list of URLs you want indexed, so a noindex entry collides with that intent. Google and Bing both treat the conflict as a quality smell that can drag the site-wide signal down.

Does putting a real date in `lastmod` actually move AI search citations?

I haven’t measured the lift directly. What the docs do say — Google, Bing, and the AI search vendors — is that lastmod is a freshness hint and that bogus values erode trust. Treating it as insurance against losing existing trust feels more honest than claiming a citation uplift.

Pushing operational rules into Astro Content Collections with Zod

Sat, 23 May 2026 00:00:00 GMT

Written rules get forgotten.

When you run a blog alone, the writer (you) breaks the writer’s (your) own rules sooner or later. Forgetting affiliate: true on a category: reviews post — and shipping it without the ad disclosure — is the kind of mistake a README check-list doesn’t catch.

Aulvem leans on Astro Content Collections and Zod to make rules that a machine can verify into build failures. What follows is where I draw the line between “let the schema catch it”, “let a lint script catch it”, and “leave it to review”.

This post is a follow-up to How this blog is built, which sets up the stack overview.

Why a README isn’t enough

A rule that lives only in a README has roughly zero enforcement.

The editor, reviewer, and writer are the same person, so there’s no second pair of eyes
Weeks later, the README isn’t going to be re-read line by line
Past me I trust; future me will plainly forget

I started with a checklist under docs/ and assumed that would be enough. By the third post I’d already missed an item. With no reviewer in the loop, the only way to force a rule is to put the check outside the writing flow — in the build. That’s why Aulvem leans on schema enforcement.

The schema, in one file

The Content Collections schema sits in src/content.config.ts with two collections (blog and services) declared together. Keeping them in one file means shared shapes (pubDate, updatedDate, heroImage) can be moved into local helpers without crossing file boundaries.

The blog schema looks like this (abridged):

import { defineCollection, z } from "astro:content";
import { glob } from "astro/loaders";

const blog = defineCollection({
  loader: glob({ pattern: "**/[^_]*.{md,mdx}", base: "./src/content/blog" }),
  schema: z
    .object({
      title: z.string(),
      description: z.string(),
      summary: z.string().optional(),
      pubDate: z.coerce.date(),
      updatedDate: z.coerce.date().optional(),
      category: z.enum(["build", "reviews"]),
      tags: z.array(z.string()).default([]),
      draft: z.boolean().default(false),
      affiliate: z.boolean().default(false),
      heroImage: z.string().optional(),
      heroAlt: z.string().optional(),
      howto: z.object({ /* ... */ }).optional(),
      faq: z.array(z.object({
        question: z.string(),
        answer: z.string(),
      })).optional(),
    })
    .refine((data) => (data.category === "reviews") === data.affiliate, {
      message: "affiliate must be true iff category is 'reviews'",
      path: ["affiliate"],
    }),
});

z.enum pins the category to two values, z.coerce.date swallows whatever date-string format the YAML emits, and .optional() + .default() are explicit about which fields the writer can leave blank. Plain tools cover most of the surface. The interesting work is the .refine() at the bottom, which is what the next section is about.

Coupling two fields with `.refine()`

When two fields must move together, I tack a .refine() onto the end of the schema. The case in Aulvem: a category: reviews post is, by affiliate-network rules, advertising, and has to carry the disclosure banner. Aulvem injects the banner (and adds rel="sponsored noopener noreferrer") only when affiliate: true. So the failure mode is publishing a review post with affiliate left at its default — disclosure missing, sponsored rel missing, ASP terms broken.

Here’s the one-liner that catches it:

.refine((data) => (data.category === "reviews") === data.affiliate, {
  message: "affiliate must be true iff category is 'reviews'",
  path: ["affiliate"],
})

Both sides compared with === means changing only one is a build failure. Reading the predicate as “these two are always equal” is clearer than writing the XOR explicitly — it’s the same logic, easier to scan three months later.

The build error reads like this:

[ContentEntryInvalidError] Content config error in `blog → 2026-05-...`:
affiliate must be true iff category is 'reviews'
  at affiliate

message ends up in the output verbatim, so it’s worth writing the message as instructions for future-you.

`.refine` vs `.superRefine`

Once you need more than one independent constraint on the same object, or per-field error messages, .superRefine() is the easier shape:

.superRefine((data, ctx) => {
  if (data.category === "reviews" && !data.affiliate) {
    ctx.addIssue({
      code: z.ZodIssueCode.custom,
      message: "reviews posts must set affiliate: true",
      path: ["affiliate"],
    });
  }
  if (data.draft && data.updatedDate) {
    ctx.addIssue({
      code: z.ZodIssueCode.custom,
      message: "draft posts should not carry updatedDate",
      path: ["updatedDate"],
    });
  }
})

For a single relationship between two fields, .refine() is still the lighter option.

Typing structured data inside frontmatter

I keep the data for the HowTo and FAQPage JSON-LD blocks in frontmatter rather than parsing the body. The reasons:

Parsing MDX headings to reconstruct structured data is brittle — a heading rename quietly breaks JSON-LD
Frontmatter is what Zod validates, so the shape is enforced for free
The JSON-LD generator can trust frontmatter without re-reading MDX

The schema looks like:

howto: z.object({
  name: z.string().optional(),
  description: z.string().optional(),
  totalTime: z.string().optional(),
  steps: z.array(z.object({
    name: z.string(),
    text: z.string(),
    image: z.string().optional(),
  })),
}).optional(),
faq: z.array(z.object({
  question: z.string(),
  answer: z.string(),
})).optional(),

Anything malformed — a howto with no steps, a faq entry missing answer — fails the build.

Where Zod can’t reach: body parity

Zod only inspects frontmatter. The body MDX is outside its scope.

Google’s quality guidelines flag “JSON-LD that has no body counterpart” as structured-data mismatch and pull the rich-result eligibility. A post with FAQ questions only in frontmatter — never echoed in the body — passes the schema and silently disqualifies itself.

The fix is a separate layer. Aulvem runs a validate-schema-match.mjs script that grep-checks the body for each faq[].question and howto.steps[].name:

if (Array.isArray(data?.faq)) {
  for (const [i, item] of data.faq.entries()) {
    if (!item?.question) continue;
    if (!bodyNorm.includes(normalise(item.question))) {
      mismatches.push(
        `faq[${i}].question not found in body: "${item.question}"`,
      );
    }
  }
}

It normalises whitespace and case, then checks for substring presence. If something is missing, the script returns exit code 1 and the build is blocked at the pre-commit / CI layer.

What it can’t catch: meaning. If the question matches but the answer underneath got swapped during editing, this layer says it’s fine.

What I gave up automating

Things neither the schema nor the grep validator can see:

Whether the FAQ answer is factually correct
How strong the disclosure language is (affiliate networks differ on what counts)
Whether the closing paragraph still reads like an AI draft
Whether the conclusion lines up with the article’s claim

These need a human read-through. Trying to encode them as a schema rule would cost more than the manual review they replace. I keep them in the pre-publish checklist and only push the patterns mechanically catchable (lint-banned-phrases.mjs for AI-tells) into automation.

The point isn’t “automate everything I can”. It’s deciding where build failures buy you more than they cost. With one person on the keyboard, automation time is finite, and that ratio matters.

The schema / lint / review split

Splitting rules across three layers makes “which layer should hold this?” answerable:

Layer	When it fires	What it catches	What it misses
Zod schema	`astro build`	types, enums, required/optional, field-to-field relationships	meaning, body-to-frontmatter parity
Lint script	pre-commit, CI	banned phrases, body/frontmatter substring parity	meaning
Review	pre-publish checklist	meaning, disclosure strength, AI-tell judgment	not automatable

Rule of thumb: if a higher layer can catch it, don’t push it down. Writing field-level constraints into a lint script means the build doesn’t fail and you have to remember to run the script. Pushing simple substring checks into Zod, conversely, makes the schema type definition harder to read.

Wrap-up

Structural constraints a machine can check go in the schema. Substring presence goes in lint. Meaning stays with review. Once that split is in place, every new operational rule can be sized against “which layer holds this?” instead of “where do I update the README?”.

The full Aulvem stack overview is in How this blog is built, if you want the bigger picture.

The same “frontmatter as the single source of truth” idea also drives the sitemap lastmod setup — written up separately in Reading sitemap lastmod from MDX frontmatter.

The instinct to make breakage visible applies outside the build too. A spreadsheet I was shipping had been quietly broken in Excel for weeks, and finding it took counting cells with openpyxl rather than a schema — A Google Sheets .xlsx breaks silently in Excel.

FAQ

When do you reach for `.superRefine` over `.refine`?

When one object needs more than one independent constraint, or when each issue needs its own message and path. For a single relationship between two fields, .refine() is enough. Aulvem only ties category and affiliate together, so it stays on .refine().

Why keep howto and faq inside the frontmatter instead of separate files?

Frontmatter gives me one source of truth that’s typed. Splitting them across files means the body, JSON-LD, and metadata each have their own home and have to be kept in sync by hand. Frontmatter as the source, JSON-LD generated from it, and the body echoing the same strings is the configuration with the fewest seams.

What happens to existing posts when I change the schema?

The build breaks. Adding a required field means every existing post needs the field before astro build will pass again. That’s intentional — a schema change forces you to confirm the rollout reach across every post.

Why isn’t the body-to-JSON-LD parity check wired into `astro build`?

To keep build time small. The parity check is a separate grep-based script run in pre-commit and CI. Folding it into the build itself would slow astro dev startup, which would hurt the writing loop.

How this blog is built — Aulvem on Astro 5 and Content Collections

Sun, 17 May 2026 00:00:00 GMT

This post is the starting point for documenting how Aulvem is built — not as a stack tour, but as a map of what we enforce through the schema and what we deliberately leave out. The stack itself is just Astro + MDX + Content Collections — nothing special there. What I want to write about is the design work on top of that: how operational rules a writer tends to forget get pushed into build-time checks.

The overview hub goes here. The individual decisions (schema enforcement, structured-data parity, custom sitemap lastmod) are split into follow-up posts.

Stack overview — Astro 5 + MDX + Content Collections + R2

Aulvem runs on Astro 5 (static output) + MDX + Content Collections + Tailwind 3 + Pagefind + Cloudflare R2. There’s no dynamic server — the site is just pre-built HTML served from Cloudflare Pages.

I narrowed the selection criteria to three points:

the build output is fully static
post metadata can be locked down by a schema
the core has a small dependency tree

Astro 5 met all three. The other features — partial hydration, UI framework integration, and so on — aren’t being used right now.

Whether Astro is the right pick depends on the requirements. If you want to layer on dynamic auth, comments, or subscriptions, something like Next.js is a better fit.

Definition: what Content Collections is

Aulvem defines two collections: blog (posts) and services (product pages). The schema lives in a single file and is validated at build time for both collections.

Full stack — the main packages

Runtime dependencies are kept to eight packages.

Package	Use	Notes
`astro` ^5	SSG core	static output only
`@astrojs/mdx` ^4	MDX support	all posts are `.mdx`
`@astrojs/sitemap` ^3	sitemap generation	`lastmod` injected manually
`@astrojs/rss` ^4	RSS generation	full `content:encoded`
`@astrojs/tailwind` ^6	Tailwind integration	`applyBaseStyles: false`
`rehype-external-links` ^3	rel attributes on external links	`noopener noreferrer`
`rehype-mermaid` ^3	Build-time mermaid → SVG	`inline-svg` strategy
`tailwindcss` ^3.4	styling	holding off on v4

On the dev side: pagefind (full-text search), sharp (local image processing), playwright (build-time SVG rendering for mermaid), typescript, and @types/node. No React, no Vue, no Vite plugins.

The opening rule is: don’t add a dependency on the hope it’ll be useful later. Unused dependencies show up in both build time and security-alert noise, so before adding one I check whether the use case is articulable in a sentence.

Three pieces I wouldn’t drop

1. Enforce rules through the schema

Operational rules don’t live in the README — they live in the frontmatter schema. I wanted the build to fail when a writer (including me) forgets, instead of leaning on memory.

For example, category: reviews posts have to be flagged as advertising under affiliate-network rules, so I require affiliate: true on them. Instead of leaving this as a line in the README, Zod’s .refine() forces category === "reviews" and affiliate === true to always be equivalent. Changing only one side breaks the build, so memory isn’t the load-bearing piece.

The same idea applies to the howto / faq structured data, but that belongs in a follow-up post.

2. Keep structured data and body text in sync

The JSON-LD (howto / faq) content is generated from the frontmatter, but the same content must also appear in the body. This avoids Google’s structured-data mismatch penalty — pumping out elaborate JSON-LD that has no corresponding body text violates Google’s quality guidelines, and getting caught costs you the structured-data display eligibility.

Aulvem’s rule is to write the same content on both sides — frontmatter and body — so a one-sided update can’t slip through. That alone closes off the failure mode where you fill in the frontmatter FAQ and forget the body.

3. Read sitemap lastmod from frontmatter yourself

Astro’s official sitemap integration doesn’t read updatedDate from MDX frontmatter. So Aulvem walks every MDX file with custom logic and pipes updatedDate ?? pubDate into the sitemap as lastmod.

Without that, the sitemap lastmod defaults to the build time, which broadcasts the noise signal that “every post was updated every build” to search engines. lastmod is also used as a freshness input by AI search, so it’s not a part to be sloppy on.

At the same time, paginated pages that return noindex, follow are dropped from the sitemap. Submitting a noindex URL through the sitemap is a contradictory signal, so both have to be handled together.

The implementation — walking MDX inside astro.config.mjs, wiring @astrojs/sitemap serialize and filter together, and the operational pitfalls — is in the follow-up post Reading sitemap lastmod from MDX frontmatter.

How the three pieces connect

flowchart LR
  FM[frontmatter] --> Zod
  FM --> Sitemap
  FM --> Sync
  Body[MDX body] --> Sync

  Zod["Zod schema check<br/>category ⇔ affiliate"] --> Build
  Sitemap["sitemap lastmod injection<br/>reads updatedDate"] --> Build
  Sync["frontmatter ↔ body<br/>parity"] --> Build

  Build[Astro build]

  Build --> HTML[HTML + JSON-LD + SVG]
  Build --> SM[sitemap.xml]

Two inputs — frontmatter and body — fan out through three build-time pieces and converge at Astro build. Frontmatter runs Zod schema validation and sitemap lastmod injection, while both frontmatter and body feed the parity check that makes sure the same content lives on both sides.

All results are unified by Astro build into the HTML, JSON-LD, and sitemap.xml. If any single piece fails, the build fails — so polishing one side alone won’t get you to publish.

Operational flow lives in a single source of truth

Operational flows — adding a post, adding a product, retiring a post — are anchored in one doc as the single source of truth. From there, the scaffolding and integrity-check scripts are wired in, and the rule is to follow the same steps every time. I think this absorbs most of the “the approach drifts run to run” kind of variance. What’s left is the warmth of the prose, which automation shouldn’t try to normalize anyway.

There’s also a rule: don’t bump updatedDate just because you added a preface note — bump it only on substantive revision. Search engines and AI search read dateModified as a freshness signal, so handling it loosely lowers the whole site’s authority. The criteria are fixed up front.

FAQ

Why Astro 5?

I narrowed it down to four constraints — Markdown-centric, fully static output, schema-typed frontmatter, and a small core dependency tree. Astro 5 was the best fit. Content Collections and MDX are built in, so there’s no need for extra plugins to get typed content.

Why no headless CMS (Sanity, Contentful, etc.)?

Publishing once or twice a week with one editor, MDX + git is faster than a CMS admin UI. A CMS adds the cost of draft locations, schema migrations, and API auth from day one. I’d revisit this when readership crosses ~10k, or when more editors come in.

How is full-text search implemented?

Pagefind runs after astro build and produces a static binary search index. The front end pulls it through a tens-of-KB WASM loader. No serverless function, no external API.

Does i18n use Astro’s built-in support as-is?

Routing uses Astro 5’s i18n (defaultLocale: en / prefixDefaultLocale: false) untouched. hreflang and sitemap lastmod, on the other hand, are read out of the MDX frontmatter through a custom helper. The details belong in a follow-up post.

Wrap-up

That’s the overview. Building and running this, what mattered most wasn’t which stack to pick — it was deciding how tightly to lock the frontmatter down with the schema.

Figma basics, part 3 — color, font, and shadow styles, then components and variants

Tue, 05 Apr 2022 00:00:00 GMT

2026-05-16 update: A few years on from the original. The skeleton still holds up, but the tool UI and version match when it was written and differ from the current Figma.

This post covers the Figma features that make a file easier to manage and easier to adjust.

Building up to the design file itself is covered in the earlier posts, so use those as the starting point.

The finished design from this post looks like this.

Create a color style

The screen elements are already in place from the earlier parts, so the next step is to register the colors and make them reusable.

Start with the header background color.

Point

Select the element whose color you want to register

Find the Fill section in the right sidebar

Click the four-dot icon to the right of Fill

Click the + icon to the right of Color Styles

Type the name you want to register in the modal input

Click Create style

That’s the color registered.

Registered colors pay off in places where the same color is reused across elements.

The header background color is also used for the footer background, so apply the style we just registered to the footer.

Point

Select the element you want to apply the registered color to

Find the Fill section in the right sidebar

Click the four-dot icon to the right of Fill

Under Color Styles, click the color you want to apply

The registered footer color is now in place.

Registering each color in Color Styles and applying it everywhere makes color changes much easier later on.

The GIF below edits the main color while every other color and element is already mapped to Color Styles.

Point

Click an empty area with no element selected

Find the Color Styles section in the right sidebar

Hover over the color you want to edit in Color Styles

The adjust icon appears on the right — click it

The color code shows below — click it

Pick the color you want in the color picker

As a design tightens up, the urge to fine-tune colors goes up too, so registering colors early tends to save time.

Color Styles also surfaces which color is doing which job and how many times each one is used, which makes it less likely an odd color slips in.

Create a font style

The same kind of registration exists for font settings.

Point

Select the element whose font you want to register

Find the Text section in the right sidebar

Click the four-dot icon to the right of Text

Click the + icon to the right of Text Styles

Type the name you want to register in the modal input

Click Create style

The font style is now registered.

Applying and editing a font style works the same way as Color Styles.

Set up the shadow

So far this has been about manageability. Next, add a shadow to an element.

The card frame is a good test case.

Point

Select the element you want to add a shadow to

Find the Effects section in the right sidebar

Click the + icon to the right of Effects

Click the sun icon to the right of Effects

Adjust the values under Drop shadow until they look right

Arrow keys (↑↓) nudge by 1 pixel

The card frame now has a shadow.

Shadows can be saved as a style the same way.

Point

Select the element whose shadow you want to register

Find the Effects section in the right sidebar

Click the four-dot icon to the right of Effects

Click the + icon to the right of Effect Styles

Type the name you want to register in the modal input

Click Create style

The shadow style is now registered.

Applying and editing a shadow style works the same way as Color Styles.

Combine components with auto layout

A button on a real web page changes color or content the moment it’s clicked.

To express those state changes inside Figma, the Variants feature is the right tool.

As things stand, Variants is awkward to use, so the prep step is to keep auto layout on the button and turn it into a separate component.

This section originally included a roughly 32 MiB GIF showing the component-plus-auto-layout operation. It’s been dropped because of the hosting (Cloudflare Pages) file size limit. For the actual behavior, see the Figma official documentation.

Point

Select the element you want to turn into a component

Move it outside the frame

Wrap in this order: component / auto layout / element

If the top of the element is a frame, add auto layout to it

Move the child component back to where it was in step 1

The flow is a bit involved, but the button is now a component with auto layout intact.

Create variants

A button is now ready outside the frame as a component.

From here, use Variants to express the button’s state changes.

Point: applying variants

Select the component-ised element you want to add variants to

Find the Variants section in the right sidebar

Click the + icon to the right of Variants

In the input field showing something like “Variant2”, type the sub-state name you want to register

Select the elements inside that state

Change colors or other properties

Point: switching variant state

Select the element whose state you want to change

Find the section in the right sidebar that shows the component name

Click the dropdown next to something like “Property1”

Pick the state to switch to

Adding variants outlines the component with a dotted border and duplicates the element.

Editing those duplicates is how the sub-states get expressed.

One caveat: Variants only applies to elements that are already components.

By default only one variant on top of the default can be registered, so add another state.

Point: adding a variant

Select the variant element you want to add another from

A + icon appears nearby — click it

New variants get added directly below the selected element, so starting from the bottom-most variant tends to keep the layout clean.

Finished

That’s part 3 of the Figma basics.

How to use Figma — basics, part 2

Fri, 01 Apr 2022 00:00:00 GMT

2026-05-16 update: A post from a few years back. The bones still hold up, but the UI screenshots and Figma version are from when it was written and don’t match the current build.

This post walks through the Figma basics, part 2.

I covered getting up to a design file in the earlier posts, so use those as a starting point.

The finished design from this post looks like the image below.

Setting a stroke

The earlier post stopped at the header and footer.

This time I start with the cards inside the content area.

The GIF below begins right after the frame has been created.

Point

Select the element you want to add a stroke to

Find the Stroke section in the right panel

Click the + icon to the right of Stroke

Click the color code that appears below

Pick a color from the picker, or from the swatch history under it

Set the line thickness with the number next to the three-line icon under the color code

That’s the stroke set.

Creating different elements

With the card outline in place, the next step is the contents.

I drop in a title and body text as shown in the GIF below.

Then a divider line and an avatar icon underneath, both made from primitive shapes.

Point: drawing a line

Click the ∨ next to the rectangle icon in the toolbar

Pick Line from the list

Move the cursor to where you want the line to start

Drag to set the length

Point: drawing a circle

Click the ∨ next to the rectangle icon in the toolbar

Pick Ellipse from the list

Move the cursor to where you want the shape

Drag to set the size

Snapping kicks in around centers and edges
Hold Shift while dragging to lock the X or Y axis

The divider and avatar icon are now in place under the content.

Grouping

The card frame is starting to hold a lot of elements, so I bundle them.

Point

Select the elements you want to group (multi-select)

Right-click to open the menu

Choose Group selection

In the left panel you can Shift-click to select a range

With elements selected, Ctrl + G also groups them

The card elements now live inside one group.

Changing the corner radius

The last piece inside the card is a button next to the avatar icon.

I drop a frame next to the avatar and give it a background color.

I want rounded corners on the button, so I round it off here.

Point

Select the element you want to round

Find the Frame section in the right panel

Click the number inside the red outline on the Frame row

Type the value you want

Arrow keys ↑↓ nudge by 1 pixel

That’s all the pieces inside the card built.

Creating a component

Cards usually appear more than once, so I duplicate it.

But instead of plain copy-paste like last time, I use Figma’s component feature — a stronger form of duplication.

Point

Select the elements you want to turn into a component

Right-click to open the menu

Choose Create component

With elements selected, Ctrl + Alt + K also creates a component

As the GIF shows, components do a few useful things at once:

Edits to the master component flow through to every instance
Per-instance overrides (no effect on the master)
- Toggle visibility per element
- Override text per element

Turning frames you’ll likely reuse into components — and reaching for them by default — is one of the biggest payoffs Figma gives you later.

Adding auto layout

At some point I want to swap the button label for something longer.

This is where auto layout comes in.

Point: turning on auto layout

Select the element you want auto layout on

Right-click to open the menu

Choose Add auto layout

With elements selected, Shift + A also enables auto layout

Normally, longer text would overflow the frame. With auto layout, as the GIF shows, the padding on all four sides stays put.

Reach for auto layout when an element’s width or height needs to grow with its content.

The button in that example grows to the right, which eats into the right-side margin.

One more tweak keeps the right margin intact while the button still grows.

Point: changing the start point

Select the element whose start point you want to change

Find the Constraints section in the right panel

In the small diagram, click the blue lines to set the start point

The button’s auto layout now anchors to the bottom-right, so the right margin stays put as it grows.

The labels (left, right, etc.) next to the diagram set the same thing if you prefer text.

Auto layout has another useful piece.

I use it here to set the padding around each card and the gap between cards.

Point: adjusting auto layout

Multi-select the elements you want under auto layout

Enable auto layout

Find the Auto layout section in the right panel

Click the rightmost square in the Auto layout row

Adjust the top / bottom / left / right padding in pixels

Click the three-line icon

Adjust the gap between elements

The Auto layout section in the right panel handles padding and spacing.

When the same element appears more than once, auto layout makes it easy to dial in outer padding and inter-element gaps in one place — worth reaching for by default.

The finished design

That covers Figma basics, part 2.

Next up: the features that make a Figma file easier to maintain over time.

Figma basics, part 1 — frames, fills, text, and alignment

Wed, 30 Mar 2022 00:00:00 GMT

2026-05-16 update: A post a few years old. The bones still hold up, but the tool UI and version are from when it was written and don’t match the current Figma.

Figma basics, part one. The previous post got as far as creating a design file. From here, I’m building out the screen itself.

The finished version looks like the image below. A plain screen, but it covers the core Figma basics: frames, fills, text, and alignment.

The basic layout

Rough map of the screen:

The header is the tooling
The left sidebar is the objects you’ve created
The right sidebar adjusts the object you’ve selected

Frame

First, the screen bounds.

In Figma, a screen is expressed using a part called a frame.

Frames are versatile — they’re used for the screen bounds, as outer containers for elements, and more. There are templates available, so the easiest place to start is picking one of those.

Point

Click the frame icon in the header

The frame templates appear in the right sidebar — pick the one you want

Clicking creates an iPhone 13 Pro Max frame in the center of the canvas.

The right sidebar shows the new frame labeled # iPhone 13 Pro Max.

With the screen bounds in place, next is the elements inside the screen.

Starting with the header.

The previous step used a frame template — this time, drawing a frame freely.

Point

Click the frame icon in the header

Move the cursor to where you want the frame

Drag to draw

Freehand drawing isn’t pixel-precise, so aim for roughly the right size and fine-tune later.

Use freehand drawing for larger blocks like headers, and for the base shape of small elements like buttons.

Change the background fill

The freehand-drawn frame comes out plain white. To make it easier to see, I’ll give the header frame a background fill.

Fine adjustments like fill color happen in the right sidebar.

Point

Select the element to adjust

Find the Fill row in the right sidebar

Click the + icon to the right of Fill

Click the color code shown below

Pick a color in the color picker

That switches the fill to black.

Move and resize elements

The header has its color, but the position and size are off.

Time to move it and resize it.

Point: repositioning

Select the element to adjust

Drag to where you want it

Smart guides snap to center and edges

Shift + drag locks movement to the x or y axis

Point: resizing

Select the element to adjust

Move the cursor to the bounding-box handle

Drag to resize

That puts the header roughly where it belongs, at roughly the right size.

But the height is a few pixels off, so a quick fine-tune.

Point: fine-tuning element height

Select the element to adjust

Click the H field under Frame in the right sidebar

Edit the value

Arrow up / down nudges by 1 pixel

Under Frame, X and Y are position; W and H are width and height — all editable.

Enter text

With the header frame in place, next are the elements inside the header.

For the walkthrough, putting some text in the header.

Point

Click the T icon in the header

Click where you want the text

Type once the cursor goes into edit mode

That puts the word “Header” inside the header.

Changing the font size

The text is too small as-is, so bumping the font size up to something readable.

Same as before — fine adjustments happen in the right sidebar.

Point

Select the text to adjust

Find the Text row in the right sidebar

Find the numeric field in the lower right of the Text row

Click the number

Edit the value

Arrow up / down nudges by 1 size unit

That gets the font size to something readable.

Align elements

Font size sorted. Next, the text position.

Same as before — fine adjustments happen in the right sidebar.

Point: horizontal center

Select the element to adjust

In the right sidebar, click the icon with a vertical line through the center

Point: vertical center

Select the element to adjust

In the right sidebar, click the icon with a horizontal line through the center

The other icons on the same row are the rest of the alignment options.

Header done. Next, the footer.

Same design as the header, so copying the header across.

Point: duplicating an element

Select the element to duplicate

Press Ctrl + C

Press Ctrl + V

That gives the footer copy.

It sits at the same position as the header, though — needs to move to the bottom.

Point: align to bottom

Select the element to adjust

In the right sidebar, click the icon with a horizontal line at the bottom

Footer done.

That covers the first set of Figma operations. Part two picks up the core features worth learning early when you start using Figma seriously.

How to use Figma — setup guide

Tue, 29 Mar 2022 00:00:00 GMT

2026-05-16 update: This post is a few years old now. The skeleton — create an account, create a team, create a file — still holds, but the UI details and labels are left as they were at the time. Today’s Figma has added Dev Mode, Variables, stronger Auto layout, and other features that didn’t exist back then. Read it alongside the current screens.

This is a walkthrough of what Figma is, and what it takes to get to your first design file. Three steps — account, team, file — in order, and the UI design canvas is right there.

Understand what Figma is

Figma is a UI design tool that runs in the browser. No install needed; open a URL and the editor is right there.

There are other prototyping tools out there, but for getting the shape of Figma, two characteristics stand out:

The work lives in the cloud
Multiple people can edit at the same time

The work lives in the cloud

Figma files live in the cloud, and editing happens inside the browser.

As long as the account exists, any PC with a browser can sign in and read or edit the current file. There’s nothing to install locally.

Multiple people can edit at the same time

The defining feature of Figma is that multiple people can edit the same file simultaneously.

For solo use the benefit is thin, but for prototyping at work it removes the usual friction of zipping files, mistaking which copy is the latest, and so on.

Note: it’s especially good in remote meetings. You can share your screen and fix things live as the conversation happens.

Create an account

To actually start using it, the first step is creating a Figma account.

Open the official site and head to signup.

https://www.figma.com/

The page above loads up. Click the Sign up button in the red box.

That opens the new-account modal.

Signing up with a Google account is the least friction and what I’d recommend. Email also works fine.

Create a team

Once the account is created and signed in, the next step is to create a team.

A team is similar to a folder on a PC. You create the team folder first, and design files go inside it — that’s the hierarchy.

Click the area highlighted in red below.

A team-name input screen comes up.

Type any team name and click the Create team button.

Next is a screen for inviting users to the team.

Since this is a solo setup on the free plan, click Skip for now at the bottom of the screen.

The last step is choosing a Figma plan.

Going with the free plan, so click Choose Starter in the red box. That finishes team creation.

Create a design file

With the team in place, the next step is creating the actual workspace — a design file.

Note that on the free plan you can only create up to three design files (limit as of writing — check the official pricing page for the current cap).

Creating the file itself is simple.

The team name shows up on the left side. Selecting it opens the screen below.

Clicking the area in the red box creates a new design file.

Once created, this screen opens — and that’s the end of setup.

So that covers what Figma is, and getting through to an empty canvas.

What to actually do inside that canvas is the subject of the next post — the basics guide.

Design thinking, and how I actually use it at work

Sat, 19 Mar 2022 00:00:00 GMT

2026 update: A few years on from the original post. The shape of the argument still holds, but any tool UIs and version numbers mentioned are from when it was written and may have drifted from current releases.

What is design thinking?

I’ve been running into “design thinking” more often at work lately, so this is me writing down how I actually understand it.

Design thinking started as a way designers approached creative work, then got generalized as a problem-solving approach for other fields.

Where logical thinking leans on theory and numbers to solve the problem in front of you,

design thinking starts from empathy and emotion and tries to find the real problem underneath.

The example people reach for is the Ford one. The user request — “I want a faster horse” — gets read not as the literal object (a horse) but as the underlying desire (“I want to move from A to B”), and the answer becomes a car.

That’s the move: doubt the problem as presented, and go after the real problem sitting behind it.

The design thinking process

The process runs as a pattern of diverge and converge. This is the Double Diamond model.

It breaks into four stages.

Diverge and converge to find the right problem — “discover” and “define”
Diverge and converge to find the right solution — “develop” and “deliver”

You spin through these four stages, fast, and the product gets sharper each loop.

Diverging to find the right problem

Discover

The discover phase is where the core idea of design thinking — searching for the real problem — does the most work.

The job here is to observe and interview actual users and explore their needs.

Why did they take that action?
What was their emotional state, before and after touching the service?
And so on.

These are the questions you use to diverge the problem space.

You let it run wide — from the obvious things that feel too dumb to say, all the way out to weird and tangential ones — and you keep them all on the table.

Define

Once you have a spread of hypotheses, you narrow.

Which problem is the one users feel most strongly?
What’s the underlying cause, and how do you abstract it?

Bring in the marketing and sales angles alongside the user angle, and converge from more than one axis.

Diverging to find the right solution

From here, the work shifts closer to ordinary logical thinking.

For ideation, I try to keep a few things in mind:

More ideas is better than fewer
The more obvious it seems, the more worth asking about
Locking in a decision too early is a real risk
Don’t shoot ideas down

The downsides of design thinking

It costs time and money

By design (so to speak), the process eats time and money.

Real product development has deadlines, so the practical move is to put a hard time limit on each loop before you start spinning.

The bar to actually run it is high

User perspective and empathy carry the whole thing, and neither is something you pick up overnight. It takes patient practice — observation, interviews, repeat.

On top of that you need the imagination to spread ideas wide, so trying to run the full loop as a beginner is rough.

It’s overkill for small improvements

Because design thinking re-examines the problem from scratch, it’s strong at producing new solutions but expensive for incremental tweaks to existing features. For pure improvement work, plain logical thinking is often the faster path.

Putting it into practice

User perspective, empathy, idea generation — all of these need training, so running the full set from tomorrow morning isn’t realistic.

The version I’ve landed on is to pull out the essence and mix it into ordinary work, a bit at a time.

What the idea is really getting at

The core, for me, is “search for the real problem” — or, said differently, step back from the problem as it was handed to you.

User perspective is the strongest lever, but you don’t have to be locked to it. Just imagining the context behind the subject opens up a different angle. Treating design thinking as a tool for shifting your viewpoint is what makes it usable without ceremony.

Training the user’s-eye view

A quick way to train the user’s-eye view is to actually pick up the competing products you use in daily life and try them.

You start noticing the differences against what you’re used to, and the comparison becomes a habit. That’s the first step into the user’s-eye view.

Design thinking, in microdoses

You don’t have to plant a flag that says “we’re doing design thinking now”. You can slip a fragment of the mindset into ordinary work.

For example, on a given meeting agenda: is the agenda itself even framed correctly? Is the perspective too local? Practice it on small units, and the resolution of the discussion shifts, little by little.

How to Choose a microSD Card: Markings by Use and a Bitrate Chart

Sat, 26 Feb 2022 00:00:00 GMT

Choose an SD Express microSD card for the Switch 2, A1 or better for phone apps, and for video up to 240 Mbps, V30 or above by bitrate.

Games: see the Switch 2 section right below
Phones: see “Phone apps want A1 or A2”
Video: see “Choose video cards by bitrate”
Dashcams: see “Dashcams need a high-endurance card”

The Switch 2 only takes microSD Express

Per Nintendo’s US support page, the Nintendo Switch 2 only works with microSD Express cards, up to 2 TB.

Per Nintendo’s US support page, a regular microSD card from the original Switch can’t store games or save data on the Switch 2, though you can still copy off its screenshots and videos
Samsung’s 256GB card for the Switch 2 reads at up to 800 MB/s. In UHS-I mode it reads at 90 MB/s

A card’s maximum speed does not tell you how much faster a game will load. Load times were not tested on a real console.

Nintendo’s US support page gives no game size guide. Check the size of each game you plan to play.

Phone apps want A1 or A2

Check that your phone supports A2 before you pay for it.

Per the SD Association, A1 and A2 are markings for running apps. IOPS is how many small reads or writes a card can do per second. You only get A2 speed when the device supports A2 too.

Marking	Minimum random read	Minimum random write
A1	1,500 IOPS	500 IOPS
A2	4,000 IOPS	2,000 IOPS

Choose video cards by bitrate

In this calculation, the write speed a recording needs in MB/s is its bitrate in Mbps divided by 8. The table compares Video Speed Class markings and gives the lowest V rating that covers it, plus the recording time on a 256 GB card. Match the marking family your device asks for.

In this calculation, V60 covers above 240 Mbps up to 480 Mbps, and V90 covers above 480 Mbps up to 720 Mbps. It’s borderline for V60 above 432 Mbps up to 480 Mbps, and for V90 above 648 Mbps up to 720 Mbps.

Recording bitrate	Write speed needed	Lowest V rating	Recording time (at most)
50 Mbps	6.25 MB/s	V10	about 682 min
100 Mbps	12.5 MB/s	V30	about 341 min
150 Mbps	18.75 MB/s	V30	about 227 min
200 Mbps	25 MB/s	V30	about 170 min
240 Mbps	30 MB/s	V30, borderline	about 142 min
300 Mbps	37.5 MB/s	V60 (UHS-II or faster)	about 113 min
400 Mbps	50 MB/s	V60 (UHS-II or faster)	about 85 min
600 Mbps	75 MB/s	V90 (UHS-II or faster)	about 56 min
800 Mbps	100 MB/s	Beyond V90	about 42 min

Per the SD Association, V60 and V90 are meant for a UHS-II or faster bus, so both the device and the card need UHS-II

Use the bitrate from your device’s manual. The tool below turns your use and bitrate into the marking you need.

Which microSD marking you need (from the use and bitrate)

Pick the use (Switch 2, phone, video or dashcam) and a capacity to see which card type and markings you need. For video and dashcams it turns the recording bitrate into the write speed needed, the lowest speed class that covers it, and how long the card records. It also shows one card that fits.

Dashcams need a high-endurance card

Put high-endurance cards first for a dashcam. There are two reasons.

Per Samsung’s official pages, Samsung’s consumer PRO Plus and microSD Express cards exclude dashcams, security cameras and other write-heavy use from their warranty
Per SanDisk’s official pages, the SanDisk High Endurance 256GB is rated for up to 20,000 hours of recording. That’s for Full HD; 4K gets fewer hours

The official pages give no lifespan in years.

Capacity, speed marks and buses

Which capacity works in my device?

Per the SD Association, the capacity standards are:

SDHC: above 2 GB up to 32 GB
SDXC: above 32 GB up to 2 TB, for devices that support SDXC or SDUC
SDUC: above 2 TB up to 128 TB
SDHC cards use FAT32, and SDXC and SDUC cards use exFAT

Which speed marking do I match?

Class 10, U1 and V10 all list 10 MB/s on paper, but mixing symbol families isn’t guaranteed to give the expected write speed. Match the family your device asks for.

U3 and V30: at least 30 MB/s
V60: at least 60 MB/s
V90: at least 90 MB/s

For recording, look at the write figure, not the read figure.

A speed class is the standard way a card states a minimum write speed. Some makers also state one for a card, such as SanDisk’s 210 MB/s for its microSD Express 256GB.

Does a faster card or bus help?

An “up to” speed is a maximum, not a minimum. The same SanDisk Extreme (SDSQXH9) is rated up to 170 MB/s write at 256GB but 120 MB/s at 128GB. Both are maximums.

A bigger card is not necessarily slower. In that example, the larger card has the higher maximum write speed.

The bus is the path between the device and the card: UHS-I, then UHS-II, then SD Express, from slowest to fastest. A slow bus in the device holds back a fast card.

Picks by condition

Check that your device accepts the capacity before you buy. The conditions:

The maker’s official page lists the markings
Sold new by Amazon.com or the maker
At least 100 ratings and 4 stars or more
Not a marked-up third-party import

Among current models from major brands, each pick is the cheapest listing that met these. Check prices on the store.

Switch 2

It is marked microSD Express, with up to 650 MB/s write.

Whether the rating count is shared with other models on the listing wasn't checked.

Video up to 240 Mbps

V30 is marked, and it is built for dashcams and security cameras. Recording from above 216 Mbps up to 240 Mbps is borderline.

Phone

A2, U3 and V30 are marked. A2 only helps if your phone supports it.

Dashcam or security camera

It is a high-endurance card whose maker publishes recording hours.

FAQ

Which microSD card works with the Nintendo Switch 2?

Per Nintendo’s US support page, the Nintendo Switch 2 only works with microSD cards built to the SD Express standard, up to 2 TB.

Are Class 10, U1 and V10 the same?

Class 10, U1 and V10 all list 10 MB/s on paper, but mixing symbol families isn’t guaranteed to give the expected write speed. Match the family your device asks for.

How do I choose between V30, V60 and V90?

In this calculation, recording at up to 240 Mbps is covered by a V30 card (at least 30 MB/s). Above 216 Mbps up to 240 Mbps the headroom is under 10%, so it’s borderline.

Above that, see the table for V60 and V90.

Is the “up to” speed printed on a card guaranteed?

No. An “up to” speed is a maximum, not a minimum. A speed class is the standard way a card states a minimum write speed, and some makers also state one for a specific card.

One SanDisk card line is rated up to 170 MB/s write at 256GB but 120 MB/s at 128GB. Both are maximums.

Do I need an A2 microSD card for my phone?

Check that your phone supports A2 before you pay for it. Per the SD Association, A1 and A2 are markings for running apps.

IOPS is how many small reads or writes a card can do per second. You only get A2 speed when the device supports A2 too.

Does a dashcam need a high-endurance card?

Put high-endurance cards first for a dashcam. Per Samsung’s official pages, some of Samsung’s consumer cards exclude dashcams, security cameras and other write-heavy use from their warranty.

Per SanDisk’s official pages, one of its high-endurance 256GB cards is rated for up to 20,000 hours of recording. That’s for Full HD; 4K gets fewer hours.

Data and formulas used (for verification)

How this was checked: Worked out from the official pages of the SD Association and the makers as of October 2026. No card was tested in person.

Sources: SD Association speed class ／ SD Association application performance class ／ SD Association bus speed ／ SD Association capacity standards ／ Nintendo Switch 2 microSD (US) ／ SanDisk Extreme ／ SanDisk High Endurance ／ SanDisk microSD Express ／ Samsung microSD Express datasheet ／ Samsung PRO Plus

Method

Write speed needed (MB/s) = recording bitrate (Mbps) ÷ 8. The bitrate is the device’s rated value for the recording mode; swings during recording aren’t modeled.
The lowest marking is the lowest symbol in the SD Association’s speed class table whose minimum write speed is at least the speed needed. Video Speed Class: V6 = 6 MB/s, V10 = 10 MB/s, V30 = 30 MB/s, V60 = 60 MB/s, V90 = 90 MB/s; UHS Speed Class: U1 = 10 MB/s, U3 = 30 MB/s; Speed Class: Class 10 = 10 MB/s (read from the table image on the official page). The bitrate limits are 240 Mbps for V30, 480 Mbps for V60 and 720 Mbps for V90. If the headroom (marking minimum − speed needed) ÷ marking minimum is under 10%, it’s marked borderline (this article’s own threshold).
The SD Association says combinations across symbol families (for example a Class 10 host with a U1 card) aren’t guaranteed to give the expected write speed, so different families aren’t treated as interchangeable. V60 and V90 are meant for a UHS-II or faster bus, so they’re shown as conditional.
Recording time (minutes) = capacity (GB) × 8000 ÷ bitrate (Mbps) ÷ 60, rounded down. 1 GB counts as 1 billion bytes (as in SanDisk’s footnotes and on card labels); the space lost to the file system isn’t subtracted, so it’s an upper estimate.
Capacity classes follow the SD Association’s table (SDHC above 2GB up to 32 GB, FAT32; SDXC above 32GB up to 2 TB, exFAT; SDUC above 2TB up to 128 TB). A1 and A2 follow the Application Performance Class table (A1: 1500 IOPS random read, 500 IOPS random write; A2: 4000 IOPS and 2000 IOPS; both 10 MB/s sustained write). The Switch 2 rules come from Nintendo’s US support page (microSD Express only, up to 2 TB).
Makers’ read and write speeds are maximums, not guarantees. Picks are cards whose markings were confirmed on the maker’s official page or datasheet, among new listings sold by Amazon or the maker directly with at least 100 ratings and 4 stars, where the model bought matches the official model number; the cheapest including shipping is chosen. No card was tested in person.

Official values used

Item	Value	Source	Checked
SDHC capacity upper limit	32 GB	sdcard.org Capacity table: SDHC more than 2GB up to 32GB (FAT32); SDXC more than 32GB up to	2026-10-08
SDXC capacity lower bound (more than this)	32 GB	sdcard.org Capacity table: SDHC more than 2GB up to 32GB (FAT32); SDXC more than 32GB up to	2026-10-08
SDXC capacity upper limit	2 TB	sdcard.org Capacity table: SDHC more than 2GB up to 32GB (FAT32); SDXC more than 32GB up to	2026-10-08
SDUC capacity upper limit (more than 2TB)	128 TB	sdcard.org Capacity table: SDHC more than 2GB up to 32GB (FAT32); SDXC more than 32GB up to	2026-10-08
SDHC file system	FAT32	sdcard.org Capacity table: SDHC more than 2GB up to 32GB (FAT32); SDXC more than 32GB up to	2026-10-08
SDXC and SDUC file system	exFAT	sdcard.org Capacity table: SDHC more than 2GB up to 32GB (FAT32); SDXC more than 32GB up to	2026-10-08
Hosts that accept an SDXC card	SDXC card works in SDXC and SDUC hosts only	sdcard.org SDXC memory cards can be used with SDXC devices and SDUC Host Devices	2026-10-08
UHS-I bus speed (SDR104)	104 MB/s	sdcard.org Bus speed table (re-read on the original page, 2026-10-08)	2026-10-08
UHS-II bus speed (full duplex)	156 MB/s	sdcard.org Bus speed table (re-read on the original page, 2026-10-08)	2026-10-08
UHS-II bus speed (half duplex)	312 MB/s	sdcard.org Bus speed table (re-read on the original page, 2026-10-08)	2026-10-08
UHS-III bus speed (full duplex, lower)	312 MB/s	sdcard.org Bus speed table (re-read on the original page, 2026-10-08)	2026-10-08
UHS-III bus speed (full duplex, upper)	624 MB/s	sdcard.org Bus speed table (re-read on the original page, 2026-10-08)	2026-10-08
SD Express bus speed (PCIe Gen3 x1)	985 MB/s	sdcard.org Bus speed table (re-read on the original page, 2026-10-08)	2026-10-08
SD Express bus speed (PCIe Gen4 x1)	1970 MB/s	sdcard.org Bus speed table (re-read on the original page, 2026-10-08)	2026-10-08
SD Express bus speed (PCIe Gen4 x2)	3940 MB/s	sdcard.org Bus speed table (re-read on the original page, 2026-10-08)	2026-10-08
Class 2 minimum write speed	2 MB/s	sdcard.org Table image (SpeedClass9.1EN.png, opened 2026-10-08): minimum sequential write s	2026-10-08
Class 4 minimum write speed	4 MB/s	sdcard.org Table image (SpeedClass9.1EN.png, opened 2026-10-08): minimum sequential write s	2026-10-08
Class 6 minimum write speed	6 MB/s	sdcard.org Table image (SpeedClass9.1EN.png, opened 2026-10-08): minimum sequential write s	2026-10-08
Class 10 minimum write speed	10 MB/s	sdcard.org Table image (SpeedClass9.1EN.png, opened 2026-10-08): minimum sequential write s	2026-10-08
UHS Speed Class U1 minimum write speed	10 MB/s	sdcard.org Table image (SpeedClass9.1EN.png, opened 2026-10-08): minimum sequential write s	2026-10-08
U3 minimum write speed	30 MB/s	sdcard.org Table image (SpeedClass9.1EN.png, opened 2026-10-08): minimum sequential write s	2026-10-08
Video Speed Class V6 minimum write speed	6 MB/s	sdcard.org Table image (SpeedClass9.1EN.png, opened 2026-10-08): minimum sequential write s	2026-10-08
V10 minimum write speed	10 MB/s	sdcard.org Table image (SpeedClass9.1EN.png, opened 2026-10-08): minimum sequential write s	2026-10-08
V30 minimum write speed	30 MB/s	sdcard.org Table image (SpeedClass9.1EN.png, opened 2026-10-08): minimum sequential write s	2026-10-08
V60 minimum write speed	60 MB/s	sdcard.org Table image (SpeedClass9.1EN.png, opened 2026-10-08): minimum sequential write s	2026-10-08
V90 minimum write speed	90 MB/s	sdcard.org Table image (SpeedClass9.1EN.png, opened 2026-10-08): minimum sequential write s	2026-10-08
SD Express Speed Class E150 minimum write speed	150 MB/s	sdcard.org Table image (SpeedClass9.1EN.png, opened 2026-10-08): minimum sequential write s	2026-10-08
E300 minimum write speed	300 MB/s	sdcard.org Table image (SpeedClass9.1EN.png, opened 2026-10-08): minimum sequential write s	2026-10-08
E450 minimum write speed	450 MB/s	sdcard.org Table image (SpeedClass9.1EN.png, opened 2026-10-08): minimum sequential write s	2026-10-08
E600 minimum write speed	600 MB/s	sdcard.org Table image (SpeedClass9.1EN.png, opened 2026-10-08): minimum sequential write s	2026-10-08
Matching host and card speed class symbols	Choose the same or higher class symbol than the host’s; different symbol families are not guaranteed to substitute (e.g. host Class 10 with a U1 card)	sdcard.org Best Combination between Speed Class Host and Card	2026-10-08
Bus mode needed for V60 / V90	V60 and V90 are used in UHS-II mode or faster; U1/U3 in SDR50/DDR50 or faster; E-classes in SD Express mode	sdcard.org Regarding bus mode, it is necessary to use a bus mode fast enough that does not	2026-10-08
A1 minimum random read	1500 IOPS	sdcard.org Specification table image (img_application.jpg, opened 2026-10-08): A1 1500 / 50	2026-10-08
A1 minimum random write	500 IOPS	sdcard.org Specification table image (img_application.jpg, opened 2026-10-08): A1 1500 / 50	2026-10-08
A2 minimum random read	4000 IOPS	sdcard.org Specification table image (img_application.jpg, opened 2026-10-08): A1 1500 / 50	2026-10-08
A2 minimum random write	2000 IOPS	sdcard.org Specification table image (img_application.jpg, opened 2026-10-08): A1 1500 / 50	2026-10-08
A1/A2 minimum sustained sequential write	10 MB/s	sdcard.org Specification table image (img_application.jpg, opened 2026-10-08): A1 1500 / 50	2026-10-08
When A2 performance is available	A2 performance is available only with an A2-capable host and an A2 card; at least A1 performance is available with A1 and A2 combinations	sdcard.org Specification table image (img_application.jpg, opened 2026-10-08): A1 1500 / 50	2026-10-08
microSD cards usable in Switch 2	Nintendo Switch 2 is only compatible with microSD Express cards	en-americas-support.nintendo.com ‘Nintendo Switch 2 is only compatible with microSD Express cards.’ (opened 2026-	2026-10-08
Old Switch microSD in Switch 2	A microSD card used with Nintendo Switch cannot save or load digital games or save data on Switch 2; screenshots and videos from the original Switch can still be retrieved	en-americas-support.nintendo.com Inserting a microSD card previously used on a Nintendo Switch console into Switc	2026-10-08
Switch 2 microSD Express capacity limit	2 TB	en-americas-support.nintendo.com microSD Express cards with a capacity of up to 2TB are supported	2026-10-08
First-use system update	A system update via the internet is required the first time a microSD Express card is used	en-americas-support.nintendo.com Please note: a system update via the internet is required the first time you use	2026-10-08
Express card in the original Switch	On a Nintendo Switch console a microSD Express card reads and writes at the speed of a microSD card	en-americas-support.nintendo.com On a Nintendo Switch console, the data reading and writing speed for a microSD E	2026-10-08
SanDisk Extreme (SDSQXH9) 256GB max read	245 MB/s	sandisk.com [256GB-1TB] Up to 245MB/s read speed (beyond UHS-I 104MB/s; compatible devices r	2026-10-08
SanDisk Extreme (SDSQXH9) 256GB max write	170 MB/s	sandisk.com [256GB-1TB] Up to 170MB/s write speed	2026-10-08
Same card, 128GB max write (varies by capacity)	120 MB/s	sandisk.com [128GB] Up to 120MB/s write speed	2026-10-08
Previous Extreme 128GB max write	90 MB/s	sandisk.com Footnote: For 128GB: Up to 90MB/s write speeds (256GB-1TB: up to 130MB/s)	2026-10-08
SanDisk High Endurance max read	100 MB/s	sandisk.com Up to 100MB/s read speed	2026-10-08
SanDisk High Endurance max write	40 MB/s	sandisk.com Up to 40MB/s write speed	2026-10-08
SanDisk High Endurance 256GB recording hours (Full HD only, up to)	20000 hours	sandisk.com Up to 20,000 hours for 256GB. Full HD video only; total hours less for 4K UHD vi	2026-10-08
SanDisk MAX ENDURANCE 256GB recording hours (Full HD only, up to)	120000 hours	sandisk.com Up to 120,000 hours/over 13 years for 256GB. Full HD video only	2026-10-08
SanDisk MAX ENDURANCE max write	40 MB/s	sandisk.com Up to 100MB/s read speed, up to 40MB/s write speed	2026-10-08
SanDisk microSD Express 256GB max write	650 MB/s	sandisk.com [256GB] Up to 880MB/s read, up to 650MB/s write	2026-10-08
SanDisk microSD Express 256GB minimum sustained write	210 MB/s	sandisk.com [256GB] 210MB/s minimum sustained write speed	2026-10-08
Same card 512GB minimum sustained write	220 MB/s	sandisk.com [512GB] 220MB/s minimum sustained write; [1TB] 745MB/s	2026-10-08
Samsung microSD Express 256GB max read	800 MB/s	download.semiconductor.samsung.com Sequential Read Speed (SD Express) 800 MB/s, (UHS-I) 90 MB/s; Speed Class Class	2026-10-08
Same card read speed in UHS-I mode	90 MB/s	download.semiconductor.samsung.com Sequential Read Speed (UHS-I) 90 MB/s	2026-10-08
Lexar PLAY PRO Express max read	900 MB/s	americas.lexar.com Up to 900MB/s read and 600MB/s write	2026-10-08
Lexar PLAY PRO Express max write	600 MB/s	americas.lexar.com Up to 900MB/s read and 600MB/s write	2026-10-08
Samsung PRO Plus 256GB max read	180 MB/s	samsung.com Read: up to 180MB/s, Write: up to 130MB/s; Speed Class U3, V30, A2	2026-10-08
Samsung PRO Plus 256GB max write	130 MB/s	samsung.com Read: up to 180MB/s, Write: up to 130MB/s	2026-10-08
Dashcam use excluded from warranty on consumer cards	Samsung’s warranty for PRO Plus and microSD Express does not extend to dashcam, CCTV, surveillance and other write-intensive uses	samsung.com Warranty does not extend to dashcam, CCTV, surveillance camera and other write-i	2026-10-08
How card capacity is labeled (1 GB = 1 billion bytes)	1GB = 1,000,000,000 bytes (1TB = 1 trillion bytes); actual usable capacity is less	sandisk.com	2026-10-08

Which microSD marking you need (from the use and bitrate): quick reference

The calculator in this article shows the result for your own settings. Settings can be set in the URL; each link below opens the calculator with those settings.

Settings	Result	Calculator
Switch 2, 256GB	✓ You need microSD Express. The Switch 2 only works with microSD Express cards. Supported up to 2TB. A regular microSD card from the original Switch can’t store games or save data (screenshots and videos can still be copied off). The first time you use a microSD Express card, the console needs a system update over the internet. 256GB is SDXC (exFAT). An SDXC card works only in devices that support SDXC or SDUC	Open with these settings
Phone, 256GB	✓ Pick A1 or better. A1 means at least 1500 random-read and 500 random-write IOPS, which is what apps lean on. A2 means at least 4000 and 2000 IOPS, but you only get A2 speed if the phone supports A2 too. 256GB is SDXC (exFAT). An SDXC card works only in devices that support SDXC or SDUC. Check your phone’s spec sheet for the largest card it takes	Open with these settings
Video (camera), 50Mbps, 256GB	✓ V10 or better is enough. Write speed needed: 6.25 MB/s (bitrate ÷ 8). Lowest marking: Video Speed Class V10 (at least 10 MB/s). If your device specifies a UHS Speed Class: U1 (at least 10 MB/s). If your device specifies a Speed Class: Class 10. Mixing symbol families isn’t guaranteed to give the expected speed, so match the family in your device’s manual. Recording time on 256GB: about 682 min (about 11 h 22 min, an upper estimate). 256GB is SDXC (exFAT). An SDXC card works only in devices that support SDXC or SDUC. This counts 1 GB as 1 billion bytes; usable space is a little less than the label	Open with these settings
Video (camera), 100Mbps, 256GB	✓ V30 or better is enough. Write speed needed: 12.5 MB/s (bitrate ÷ 8). Lowest marking: Video Speed Class V30 (at least 30 MB/s). If your device specifies a UHS Speed Class: U3 (at least 30 MB/s). Mixing symbol families isn’t guaranteed to give the expected speed, so match the family in your device’s manual. Recording time on 256GB: about 341 min (about 5 h 41 min, an upper estimate). 256GB is SDXC (exFAT). An SDXC card works only in devices that support SDXC or SDUC. This counts 1 GB as 1 billion bytes; usable space is a little less than the label	Open with these settings
Video (camera), 150Mbps, 256GB	✓ V30 or better is enough. Write speed needed: 18.75 MB/s (bitrate ÷ 8). Lowest marking: Video Speed Class V30 (at least 30 MB/s). If your device specifies a UHS Speed Class: U3 (at least 30 MB/s). Mixing symbol families isn’t guaranteed to give the expected speed, so match the family in your device’s manual. Recording time on 256GB: about 227 min (about 3 h 47 min, an upper estimate). 256GB is SDXC (exFAT). An SDXC card works only in devices that support SDXC or SDUC. This counts 1 GB as 1 billion bytes; usable space is a little less than the label	Open with these settings
Video (camera), 200Mbps, 256GB	✓ V30 or better is enough. Write speed needed: 25 MB/s (bitrate ÷ 8). Lowest marking: Video Speed Class V30 (at least 30 MB/s). If your device specifies a UHS Speed Class: U3 (at least 30 MB/s). Mixing symbol families isn’t guaranteed to give the expected speed, so match the family in your device’s manual. Recording time on 256GB: about 170 min (about 2 h 50 min, an upper estimate). 256GB is SDXC (exFAT). An SDXC card works only in devices that support SDXC or SDUC. This counts 1 GB as 1 billion bytes; usable space is a little less than the label	Open with these settings
Video (camera), 240Mbps, 256GB	△ V30 just covers it. Write speed needed: 30 MB/s (bitrate ÷ 8). Lowest marking: Video Speed Class V30 (at least 30 MB/s). If your device specifies a UHS Speed Class: U3 (at least 30 MB/s). Mixing symbol families isn’t guaranteed to give the expected speed, so match the family in your device’s manual. Only 0% to spare, so it’s borderline. Recording time on 256GB: about 142 min (about 2 h 22 min, an upper estimate). 256GB is SDXC (exFAT). An SDXC card works only in devices that support SDXC or SDUC. This counts 1 GB as 1 billion bytes; usable space is a little less than the label	Open with these settings
Video (camera), 300Mbps, 256GB	△ You need V60 (UHS-II or faster). Write speed needed: 37.5 MB/s (bitrate ÷ 8). Lowest marking: Video Speed Class V60 (at least 60 MB/s). V60 and V90 are meant for a UHS-II or faster bus. Check that both the device and the card support UHS-II. Recording time on 256GB: about 113 min (about 1 h 53 min, an upper estimate). 256GB is SDXC (exFAT). An SDXC card works only in devices that support SDXC or SDUC. This counts 1 GB as 1 billion bytes; usable space is a little less than the label	Open with these settings
Video (camera), 400Mbps, 256GB	△ You need V60 (UHS-II or faster). Write speed needed: 50 MB/s (bitrate ÷ 8). Lowest marking: Video Speed Class V60 (at least 60 MB/s). V60 and V90 are meant for a UHS-II or faster bus. Check that both the device and the card support UHS-II. Recording time on 256GB: about 85 min (about 1 h 25 min, an upper estimate). 256GB is SDXC (exFAT). An SDXC card works only in devices that support SDXC or SDUC. This counts 1 GB as 1 billion bytes; usable space is a little less than the label	Open with these settings
Video (camera), 600Mbps, 256GB	△ You need V90 (UHS-II or faster). Write speed needed: 75 MB/s (bitrate ÷ 8). Lowest marking: Video Speed Class V90 (at least 90 MB/s). V60 and V90 are meant for a UHS-II or faster bus. Check that both the device and the card support UHS-II. Recording time on 256GB: about 56 min (an upper estimate). 256GB is SDXC (exFAT). An SDXC card works only in devices that support SDXC or SDUC. This counts 1 GB as 1 billion bytes; usable space is a little less than the label	Open with these settings
Video (camera), 800Mbps, 256GB	✕ Even V90 falls short. Write speed needed: 100 MB/s (bitrate ÷ 8). That’s more than V90’s minimum of 90 MB/s. Recording time on 256GB: about 42 min (an upper estimate). 256GB is SDXC (exFAT). An SDXC card works only in devices that support SDXC or SDUC. This counts 1 GB as 1 billion bytes; usable space is a little less than the label	Open with these settings
Video (camera), 100Mbps, 128GB	✓ V30 or better is enough. Write speed needed: 12.5 MB/s (bitrate ÷ 8). Lowest marking: Video Speed Class V30 (at least 30 MB/s). If your device specifies a UHS Speed Class: U3 (at least 30 MB/s). Mixing symbol families isn’t guaranteed to give the expected speed, so match the family in your device’s manual. Recording time on 128GB: about 170 min (about 2 h 50 min, an upper estimate). 128GB is SDXC (exFAT). An SDXC card works only in devices that support SDXC or SDUC. This counts 1 GB as 1 billion bytes; usable space is a little less than the label	Open with these settings
Dashcam, 25Mbps, 256GB	✓ V6 or better is enough. Write speed needed: 3.13 MB/s (bitrate ÷ 8). Lowest marking: Video Speed Class V6 (at least 6 MB/s). If your device specifies a UHS Speed Class: U1 (at least 10 MB/s). If your device specifies a Speed Class: Class 10. Mixing symbol families isn’t guaranteed to give the expected speed, so match the family in your device’s manual. Recording time on 256GB: about 1,365 min (about 22 h 45 min, an upper estimate). Dashcams overwrite the oldest footage, so this is roughly how far back it keeps. Some consumer cards exclude write-heavy use like dashcams from their warranty (Samsung PRO Plus and microSD Express). High-endurance cards publish a recording-hours rating (SanDisk High Endurance 256GB: up to 20,000 hours, Full HD). 256GB is SDXC (exFAT). An SDXC card works only in devices that support SDXC or SDUC. This counts 1 GB as 1 billion bytes; usable space is a little less than the label	Open with these settings
Dashcam, 50Mbps, 128GB	✓ V10 or better is enough. Write speed needed: 6.25 MB/s (bitrate ÷ 8). Lowest marking: Video Speed Class V10 (at least 10 MB/s). If your device specifies a UHS Speed Class: U1 (at least 10 MB/s). If your device specifies a Speed Class: Class 10. Mixing symbol families isn’t guaranteed to give the expected speed, so match the family in your device’s manual. Recording time on 128GB: about 341 min (about 5 h 41 min, an upper estimate). Dashcams overwrite the oldest footage, so this is roughly how far back it keeps. Some consumer cards exclude write-heavy use like dashcams from their warranty (Samsung PRO Plus and microSD Express). High-endurance cards publish a recording-hours rating (SanDisk High Endurance 256GB: up to 20,000 hours, Full HD). 128GB is SDXC (exFAT). An SDXC card works only in devices that support SDXC or SDUC. This counts 1 GB as 1 billion bytes; usable space is a little less than the label	Open with these settings

Manfrotto PIXI EVO review — a ball-head mini tripod that holds a full-frame body and still fits in a bag

Wed, 16 Feb 2022 00:00:00 GMT

2026 update: Ported from the old VuePress blog. The PIXI EVO is still in Manfrotto’s current line-up in 2026 as a staple model. The angles reviewed here — load rating, two-stage leg angle plus four-section legs, stiff ball-head lock — still hold up as practical buying criteria. The numbers are the manufacturer’s published figures from the original writing; current official specs may differ.

When I picked up a mirrorless body, I went looking for a single mini tripod that could cover desk, café, and outdoor snap shooting. Conditions: it had to stay rigid with a full-frame body on top, fit in a bag, and span low and high heights within reason.

I landed on the Manfrotto PIXI EVO. It is one step heavier than the plain PIXI from the same family, and in exchange you get four-section telescoping legs and two leg-angle modes. After more than two years of carrying it from indoor product shots to travel snaps, this is the write-up.

The verdict — a strong pick if you want one mini tripod that takes a full-frame body

Short answer: a rated 2.5 kg load covers a full-frame body, and two leg angles plus four-section telescoping legs span desk-low to just-below-eye-level. As all-rounder ball-head mini tripods go, it is one of the few practical ones.

The reasoning is simple. Mini tripods that combine “carries real weight” with “covers a real height range” are rarer than the category suggests. Most mini tripods rate at around 1 kg, have fixed-length legs, and either cannot take a full-frame body or can but never change height. The PIXI EVO addresses both gaps at once.

That said, some buyers will be happier elsewhere:

People who want to wrap legs around a pillar or handrail: the flexible-leg JOBY GorillaPod fits that use case better
People counting every gram: at 270g it is heavy for its class
People who need 1° video levelling: the ball-head lock is stiff, so micro-adjustments are awkward

Build — red logo on top, stubby integrated-head silhouette

Short answer: matte-black legs with the red Manfrotto logo on the ball head — a familiar Manfrotto palette. The head and legs are one piece, and the head lock is the unusual “turn the logo on top” mechanism.

Categorically it is a 1/4-inch ball-head mini tripod. The legs carry an angle-mode switch at the base and a push-button telescoping lock on each leg. The silhouette is distinctive enough to stick in memory.

Item	Value
Own weight	About 270g (rated)
Maximum load	About 2.5 kg (rated)
Head	Ball head (1/4-inch screw top plate)
Leg sections	4
Leg-angle modes	2 (normal / low)

You lock the head by turning the red Manfrotto logo on the front. It is firm enough that it does not slip on its own. The 1/4-inch screw tightens via the dial on the head. An Arca-Swiss compatible clamp will screw on, but it is close to twice the head’s width, so the rig leans top-heavy.

A slide switch at the base of the legs flips between normal and low leg-angle modes. The two modes are printed under the switch, so first-time use is unambiguous.

In low mode the legs splay almost flat to the surface. Useful when you are shooting product photos on a desk and want the camera right at subject height.

Folded, it is a slim cylinder. The backs of the three legs are cut at an angle so they sit flush against each other when closed. The result fits into a gap in a camera bag or the side pocket of a jacket.

Each leg has a push-button telescoping lock. Press, pull, release — that is the whole interaction.

The legs telescope in four sections. They lock at intermediate stops too, but in practice almost every shoot ends up using either fully extended or fully retracted. Treat the in-between stops as fine adjustment and the design reads more cleanly.

The camera body was in use for these review shots, so a GoPro stands in. For a GoPro-class load the PIXI EVO is wildly overspec’d, but the stability feels reassuring all the same.

What worked

Short answer: it stays rigid with a full-frame body on top; the two leg-angle modes plus four-section legs give it a real working height range; folded, it collapses into a slim stick that goes in a bag.

A 2.5 kg load rating — full-frame bodies are realistic

Unusually for a mini tripod, the head does not give up under a full-frame body with a standard zoom. A pro-grade telephoto zoom puts the centre of gravity too high to recommend, but anything up to a standard zoom is fine for desk product shots, table photography, and low-position street snaps.

The leg tips are rubber and grip on flooring and desk tops well. On gravel or concrete steps outdoors you have to adjust the leg angle so they bite, but that is a mini-tripod constraint in general, not a PIXI EVO one.

Two leg-angle modes plus four-section legs — real working height range

A typical mini tripod has fixed-length legs, which means it has exactly one usable height: “sitting on a desk.” The PIXI EVO sits flat on the floor in low mode, and reaches just below eye level in normal mode with the legs fully extended.

On a desk, low mode puts the camera at subject height for product shots. Outdoors, normal mode plus fully extended legs hits hip height for snap shots. Switching between scenes is fast. If you only want to carry one mini tripod across two or three shooting contexts, that flexibility earns its keep.

Folds into a stick — fits in the gaps in a bag

The angled cuts on the back of each leg mean the closed unit becomes a slim cylinder. It slides into the divider gap in a camera bag, or into the side pocket of a daypack. As “carryable in spite of being built for full-frame bodies” goes, it lands in the realm of “heavy but you will still bring it.”

Doubles as a selfie stick

With the legs folded, you can hold the column as a handle and get a passable selfie stick. Extending the legs while keeping them closed buys another few dozen centimetres. It will not match a dedicated selfie stick for length or weight, but on a trip the fact that you do not have to pack both a tripod and a selfie stick adds up.

What didn’t

Short answer: 270g is heavy for the class; the stiff ball-head lock makes 1° tweaks awkward; the bare 1/4-inch screw on top is exposed.

270g own weight — you notice it in the pouch

This is the flip side of “rigid enough for a full-frame body.” For a mini tripod it is on the heavy end. Drop it into a sacoche or a small pouch and you get a “do I really need to bring this today?” moment before heading out.

If portability is the deciding factor, a sub-60g flexible-leg option like the JOBY GorillaPod JB01542-PKK is the lighter answer. Load capacity and rigidity are separate concerns, though — the choice is “lighter, or higher load.”

Stiff ball-head lock, awkward at fine adjustments

The red logo plate that doubles as the head lock is firm. From a “do not let the camera fall” angle that is the right design choice, but when you are tightening a composition by one degree at a time, loosening the lock often shifts the framing more than you intended.

For video work where you want strict horizon levelling, or panning at a controlled speed, a three-way head or a dedicated ball head with independent lock pressure is the better tool. For stills, it is rarely a problem.

The bare 1/4-inch screw on top

There is no quick-release plate — the 1/4-inch screw sits exposed on the head’s top plate. Take the camera off and toss the tripod into a bag, and the screw rubs against the bag’s inner lining. A fabric inner case is enough to dodge the issue, but it is a design quirk.

Comparison

Short answer: heavier than a flexible-leg tripod but more rigid; heavier and pricier than the plain PIXI but with more load capacity and height range; less tall than a typical telescoping mini tripod, but more compact head-and-all.

vs JOBY GorillaPod JB01542-PKK — “light and unusual placements” vs “rigid and high load”

Angle	Manfrotto PIXI EVO	JOBY GorillaPod JB01542-PKK
Intended gear	Phone to full-frame standard zoom	Phone, webcam, GoPro
Own weight	About 270g (rated)	About 52g (rated)
Maximum load	About 2.5 kg (rated)	About 325g (rated)
Mounting site	Flat floor or desk	Floor, pillar, handrail, shelf edge
Height adjustment	2 leg angles + 4-section legs	Bend the legs as needed
Ball-head lock	Yes (firm)	Yes (firm, stepless)

The PIXI EVO is for planting heavy gear on a flat surface. The GorillaPod is for hooking light gear onto a non-flat one. The use cases are opposite, so the decision is “do I want to carry weight, or to place it in odd spots.” For a full comparison, see the JOBY GorillaPod JB01542-PKK review.

vs the plain PIXI — “lighter and cheaper” vs “more load and more height”

Angle	PIXI EVO	Plain PIXI
Intended gear	Up to full-frame standard zoom	Up to mirrorless standard zoom
Maximum load	About 2.5 kg (rated)	About 1 kg (rated)
Leg sections	4	None (fixed length)
Leg-angle modes	2	1
Price tier	Mid	Entry

If a phone, a compact, or a mirrorless body with a kit zoom is the full scope of use, the plain PIXI is enough. If “I might bring a full-frame body one day” or “I want to switch between low and high” is part of the plan, the EVO’s price premium is meaningful.

vs a typical telescoping mini tripod

Angle	PIXI EVO	Typical telescoping mini tripod
Maximum height	4-section legs + normal angle	Centre column reaches higher
Head	Integrated ball head	Sometimes three-way heads
Fine adjustment	Weak (lock is firm)	Three-way heads handle this well
Folded form	Legs flush, slim cylinder	Head tends to protrude
Portability	Fits gaps in a bag	Depends on size

If “slim when folded” and “switch between desk-low and just-below-eye-level” matter, the PIXI EVO wins. If “tallest possible” and “smooth video adjustment” matter, you are looking at a different category. Same “mini tripod” label, different target situations.

FAQ

Q. Should I pick the plain PIXI or the PIXI EVO? A. Decide by the gear you plan to mount. For a phone, a compact, or a mirrorless body with a kit zoom, the plain PIXI is enough. If a full-frame body with a standard-to-short-telephoto zoom is in scope, the PIXI EVO’s 2.5 kg rating and four-section legs are the safer choice. Conversely, if you only ever sit it on a desk to shoot phone vlogs, the EVO’s extra weight and bulk are overkill.

Q. Can it double as a selfie stick for a GoPro or a phone? A. With the legs closed, you can grip the column as a handle and treat it as a makeshift selfie stick. Extending the legs while keeping them closed buys another few dozen centimetres of reach. It will not match a dedicated selfie stick for length or weight, so think of it as a tripod that occasionally pulls double duty.

Q. Will it carry an Arca-Swiss compatible clamp? A. The head’s top plate is a 1/4-inch screw, so a standard Arca-Swiss clamp screws on directly. The clamp is usually close to twice the width of the head itself, so the rig ends up top-heavy. If Arca-Swiss is your default, picking the smallest clamp you can live with is steadier than swapping the head outright.

Q. Can I use it outdoors with a full-frame DSLR or mirrorless? A. Within the stated 2.5 kg load it carries fine, but ground and wind matter. On gravel or concrete steps, adjust the leg angle so the rubber feet bite. For long exposures, keep the centre of gravity low — no centre-column extension, no strap wrapped around the body. On windy days, hang a bag under the tripod as ballast.

Verdict — a strong pick if you want exactly one mini tripod

The PIXI EVO is one of the few mini tripods that simultaneously delivers load capacity, rigidity, a working height range, and a slim folded form. It is heavier than phone-only ultralights, and it cannot wrap legs around a pillar like a flexible-leg tripod, but it does make the realistic case — “carry one tripod that handles full-frame bodies indoors and outdoors” — actually work.

In more than two years of daily-ish use, neither the head lock nor the telescoping locks have loosened with age. For someone who would rather carry a little more weight in exchange for a tripod that lasts, it is the right tool.

If “as light as possible” or “wrap it around a pillar” is the real requirement, a different category will be a happier purchase. If the use case is “plant slightly heavy gear on a flat surface,” this is a hard one to regret.

Minecraft PC Requirements: RAM, GPU and What to Buy

Tue, 27 Apr 2021 00:00:00 GMT

Per Mojang’s requirements table, Java Edition recommends 16 GB of RAM and a CPU such as a Core i5-12400. The GPU should have 6 GB of VRAM (its own memory), such as a GeForce RTX 2060. The minimum is 8 GB of RAM (12 GB with integrated graphics), 2 GB of VRAM and a CPU with 4 physical cores.

The GPU is the part that draws the picture. The figures come from official sources as of October 2026, and no hardware was tested.

The mod and shader numbers have no primary source, so they are this guide’s assumptions. Shaders are extra lighting and shadow effects.

Minimum and recommended specs for Java and Bedrock

Per Mojang’s requirements table, Bedrock Edition needs 4 GB of RAM at minimum and recommends 8 GB, half of Java’s 8 GB minimum and 16 GB recommendation.

Per the FAQ on Mojang’s store page, the PC edition includes both Java and Bedrock. Bedrock runs on Windows, and Java runs on Windows, Mac and Linux.

Free storage: Java needs 2 GB minimum and 4 GB recommended; Bedrock needs 4 GB minimum and 8 GB recommended. Whether an SSD is required, and how much space modpacks need, couldn’t be confirmed from the sources.

To play with consoles or phones, Bedrock is the edition that crosses over; Java only plays with Windows, Mac and Linux. The mods and shaders in this guide are about Java Edition.

The Java and Bedrock columns use only the RAM and VRAM in Mojang’s requirements table (Bedrock: RAM only), so compare GPU and CPU models yourself. The mod and shader columns have no official numbers, so they compare against this guide’s assumptions (6–12 GB allocated to a heavy modpack, 8–16 GB VRAM for shaders). The 8–16 GB comparison is for discrete GPUs; shared memory on integrated graphics isn’t assessed.

Setup	Java	Bedrock	Heavy mods (assumed)	Shaders (assumed)	Allocation ceiling
Integrated GPU, 8 GB RAM	RAM below the minimum	RAM meets recommended (borderline)	Below the assumed range	Not assessed for integrated graphics	4 GB
Integrated GPU, 16 GB RAM	RAM meets recommended (borderline)	RAM meets recommended	Part of the assumed range	Not assessed for integrated graphics	8 GB
4 GB VRAM, 8 GB RAM	Exactly at the minimum (borderline)	RAM meets recommended (borderline)	Below the assumed range	Below the assumed range	4 GB
6 GB VRAM (RTX 4050 Laptop), 16 GB RAM	Meets recommended (borderline)	RAM meets recommended	Part of the assumed range	Below the assumed range	8 GB
8 GB VRAM (e.g. RTX 5060), 16 GB RAM	Meets recommended (borderline)	RAM meets recommended	Part of the assumed range	Part of the assumed range	8 GB
16 GB VRAM, 32 GB RAM	Meets recommended	RAM meets recommended	Whole assumed range	Whole assumed range	16 GB

How to check whether your PC meets the specs

Capacity and performance are two separate checks. Compare RAM and VRAM as numbers, and compare CPU and GPU models with Mojang’s examples.

Edition	Spec level	CPU examples	GPU examples
Java	Minimum (discrete GPU)	Intel Core i3-10100, AMD Ryzen 3 3100, Snapdragon X or Apple M1 (4 or more physical cores)	Discrete GPU: NVIDIA GTX 950 or newer, AMD Radeon RX 460 or newer, Intel Arc A310 or newer (Vulkan 1.3, 2 GB or more of VRAM)
Java	Minimum (integrated GPU)	Intel Core i3-10100, AMD Ryzen 3 3100, Snapdragon X or Apple M1 (4 or more physical cores)	Integrated GPU: Intel i5-6400 or newer, Apple M1 or newer, Snapdragon X (Vulkan 1.3, 2 GB or more of VRAM)
Java	Rec.	Intel Core i5-12400, AMD Ryzen 5 5600 or Apple M2 Pro	NVIDIA RTX 2060, AMD RX 5600 XT, Intel Arc A580 or Apple M2 Pro (6 GB of VRAM)
Bedrock	Minimum	Intel Celeron J4105, AMD FX-4100	Intel HD Graphics 4000, AMD Radeon R5
Bedrock	Rec.	Intel Core i7-6500U, AMD A8-6600K	NVIDIA GeForce 940M, AMD Radeon HD 8570D

Check your installed RAM. In Windows, go to Start > Settings > System > About. Compare it with the numbers at the top and in the table above (the capacity check).
Look up your CPU and GPU models. For the GPU, also check the VRAM on the maker’s spec page. Integrated graphics is the GPU built into the CPU, and a discrete GPU is a separate graphics card.
Compare the models with the examples above. Where a model is listed with “or newer”, check whether yours falls in that range. Java needs at least 4 physical cores and 2 GB of VRAM at minimum, 6 GB recommended.
Per Mojang’s requirements table, this guide doesn’t judge the performance of models not in these examples. Before buying, give the store or maker the CPU and GPU models and your edition, and ask whether it meets Mojang’s recommended specs.
The allocation ceiling is half your installed RAM. With 16 GB of RAM installed, you can allocate up to 8 GB.

In this guide’s tool, pick your edition, how you play and your current PC (laptop or desktop, RAM, GPU). It shows whether you meet the minimum and recommended RAM and VRAM, what to add, and how much memory to give the game. It doesn’t check the CPU, so compare CPU models yourself.

Minecraft PC spec check

Pick your edition (Java or Bedrock), how you play and your current PC (laptop or desktop, GPU, RAM, VRAM) to see whether it meets Mojang's minimum and recommended specs, what to add, and how much memory to give the game.

How much RAM to have and how to allocate it

Per the Java 26.1 release notes, the default memory allocation, meaning the RAM you let the game use, is 4 GB from 26.1, up from 2 GB before. Per Mojang’s Help Center, the recommended maximum allocation on a 64-bit OS is half your RAM: up to 8 GB with 16 GB installed, or 16 GB with 32 GB.

Whether giving the game more memory makes it faster couldn’t be confirmed from the sources; check a modpack’s download page for its recommended amount.

Under this guide’s assumptions, 24 GB of installed RAM covers the full heavy-mod allocation range, with no headroom. If the pack takes 6–12 GB, the top of that range fits within half your RAM from 24 GB, and exactly 24 GB is borderline. With 16 GB installed you can give it up to 8 GB.

To change the allocation, open the launcher’s Installations tab and pick More options. In the JVM arguments field (the setting passed to the part that runs Java), change the number in “-Xmx4G” to the amount you want. For 8 GB, that is “-Xmx8G”.

How to choose a CPU and GPU

Per the FAQ on Mojang’s store page, Minecraft is more CPU-intensive than GPU-intensive. If you’re unsure, start with a CPU that reaches the recommended models.

Yes. The integrated GPUs in Mojang’s minimum GPU examples are the Intel i5-6400 or newer, Apple M1 or newer and Snapdragon X. That covers the GPU only, so check the CPU separately and compare Bedrock’s GPU by model.

Java’s recommended GPU examples are all discrete cards except the Apple M2 Pro.

Per Mojang’s official pages, Bedrock ray tracing is available on DirectX-capable devices such as GeForce RTX and select AMD GPUs; check Minecraft’s official notes for supported GPUs. Vibrant Visuals on PC needs Minecraft to run on DX12, and it is planned for Java Edition later.

GPU and VRAM for shaders

Under this guide’s assumptions, for shaders, aim for 8 GB of VRAM or more, and 16 GB for headroom. An official figure wasn’t found in the sources checked, so this is an estimate over 8–16 GB, not a measurement.

Per the Iris FAQ, a VRAM figure for shaders wasn’t found in the sources checked (Iris site, Mojang). Mojang’s recommended 6 GB of VRAM is for the Fancy preset without shaders.

Per NVIDIA’s published specs, the GeForce RTX 5060 Ti comes in 8 GB and 16 GB versions, so check for “16G” or “16GB” in the model name when you buy.

Iris is the mod that runs shaders. Iris supports Fabric on 1.16.5 and higher and NeoForge on 1.21.1 and higher, not Forge, and ships with Sodium.

Per Modrinth’s public data, 26.x support looks like this.

Mod	26.x support
Iris (Fabric)	Builds available for 26.1, 26.2 and 26.3
Iris (NeoForge)	Not verified
Sodium	Lists versions through 26.3 (loader-by-loader support not checked)

For MODs and shaders on a laptop, this only compares capacity with the assumptions, and performance wasn’t tested. With 16 GB installed the allocation ceiling is 8 GB, so the top of the heavy-mod assumption (6-12 GB) doesn’t fit. Against the shader assumption (8-16 GB), the RTX 4050 Laptop’s 6 GB of VRAM falls short and the RTX 5060 Laptop’s 8 GB covers part of it.

The same FAQ says, Optifine, Canvas and Vulkan mods can’t be used with the Iris shader mod and crash the game.

Streaming

The memory comparisons here are this guide’s assumptions, not measurements.

Per OBS’s official requirements, OBS says CPU needs vary a lot with the encoder, resolution, fps and scene, so there’s no fixed number. RTX 50 cards have a 9th-generation NVENC (the GPU’s video encoder) that can also encode AV1. If you stream, leaving 4–8 GB for Windows, OBS and a browser on top of the game’s 4 GB still fits in the recommended 16 GB.

Budget

Settle on a build that meets the recommended RAM and VRAM first, then raise it for mods and shaders. Check prices on the store page.

Requirement changes and Vulkan

Per Mojang’s announcement, both the minimum and recommended Java Edition requirements went up on July 21, 2026. The minimum targets 1080p at 30 FPS on the Fast preset, and the recommended spec targets 1080p at 60 FPS on Fancy.

The same announcement says, a PC below the minimum should still launch the game. Performance and visual quality are no longer guaranteed, and it may stop launching as the game moves to Vulkan (a newer way for the GPU to draw).

Per Mojang’s requirements table, the requirements table asks for a Vulkan 1.3 capable GPU. The separate in-game Vulkan renderer (added experimentally in 26.2) needs Vulkan 1.2, and that can change. Older GPUs without Vulkan run the game on OpenGL.

Mods that render with OpenGL need extra work to support Vulkan; check each mod’s page for its status.

Per the Java 26.1 release notes, from 26.1 the game requires Java 25.

From 26.1 Java 25 comes bundled with the game, so with the bundled version you don’t need to install Java yourself. With a third-party launcher, check that it uses Java 25. For an older modpack, check the Minecraft and Java versions its page supports.

Picks by need

Where a listing met the bar, one is picked per need. The bar is a new item sold by Amazon or the maker’s own store, with 100 or more ratings and 4 stars or higher. In the US, there is no complete gaming PC pick among the products checked.

Among this guide’s picks, this guide’s RAM and graphics card picks are desktop parts and won’t fit a laptop. Laptop memory uses a different, smaller module type called SO-DIMM. The RAM picks are DDR5, so first check that your motherboard supports DDR5.

DDR5 memory kit

It meets the recommended 16 GB, and the top of the heavy-mod assumption (12 GB) fits within half of it.

Graphics card with 16 GB of VRAM

It covers the upper end of the shader assumption (16 GB). Per ASUS’s published specs, the recommended PSU is 550W, it needs 1 8-pin power connector, and it is 202 mm long. Check your PSU and case first.

The selector also lists an 8GB PRIME model and an RX 7600 model; choose Dual Fans, RX9060XT|16GB|Black.

FAQ

What are the recommended specs for Minecraft Java Edition?

Java Edition recommends 16 GB of RAM and a CPU such as a Core i5-12400. The GPU should have 6 GB of VRAM (its own memory), such as a GeForce RTX 2060.

The minimum is 8 GB of RAM (12 GB with integrated graphics), 2 GB of VRAM and a CPU with 4 physical cores.

Did the Minecraft system requirements change? Can my old PC still run it?

Both the minimum and recommended Java Edition requirements went up on July 21, 2026. The same announcement says, a PC below the minimum should still launch the game.

Performance and visual quality are no longer guaranteed, and it may stop launching as the game moves to Vulkan (a newer way for the GPU to draw).

Is 16 GB of RAM enough for Minecraft?

For Java without mods, yes, because 16 GB is Mojang’s recommended figure. Under this guide’s assumptions, 24 GB of installed RAM covers the full heavy-mod allocation range, with no headroom.

If the pack takes 6–12 GB, the top of that range fits within half your RAM from 24 GB, and exactly 24 GB is borderline. With 16 GB installed you can give it up to 8 GB.

How much RAM should I allocate to Minecraft?

Per the Java 26.1 release notes, the default memory allocation, meaning the RAM you let the game use, is 4 GB from 26.1, up from 2 GB before. The recommended maximum allocation on a 64-bit OS is half your RAM: up to 8 GB with 16 GB installed, or 16 GB with 32 GB.

Per Mojang’s Help Center, whether giving the game more memory makes it faster couldn’t be confirmed from the sources; check a modpack’s download page for its recommended amount.

What GPU and how much VRAM do I need for shaders?

Per the Iris FAQ, a VRAM figure for shaders wasn’t found in the sources checked (Iris site, Mojang). Mojang’s recommended 6 GB of VRAM is for the Fancy preset without shaders.

Can I play Minecraft on integrated graphics?

Yes. The integrated GPUs in Mojang’s minimum GPU examples are the Intel i5-6400 or newer, Apple M1 or newer and Snapdragon X.

That covers the GPU only, so check the CPU separately and compare Bedrock’s GPU by model.

Data and formulas used (for verification)

How this was checked: Requirements: Mojang’s requirements table, the requirements announcement, the Java 26.1 announcement. Allocation: Mojang’s Help Center, CurseForge, Distant Horizons. GPU, streaming and mods: NVIDIA, AMD, OBS, Iris.

Method

Java and Bedrock verdicts use only the RAM and VRAM in Mojang’s requirements table (the Java requirements were revised on 2026-07-21). Java minimum: 8 GB of RAM with a discrete GPU, 12 GB with integrated graphics, and 2 GB of VRAM. Java recommended: 16 GB of RAM and 6 GB of VRAM. Bedrock needs 4 GB of RAM at minimum and 8 GB recommended. CPU and GPU models are compared with the examples in the table, not calculated.
The allocation ceiling is installed RAM divided by 2 (Mojang’s Help Center, 64-bit OS), lowered to 2 GB if you have 4 GB or less. The Java default is 4 GB from 26.1 (it was 2 GB before). For example, 16 GB installed gives 8 GB and 32 GB gives 16 GB.
The heavy-mod allocation is an assumption of 6 GB to 12 GB, set from CurseForge’s general guidance (Vanilla 2-4 GB; light modded 4-6 GB; heavy modpacks 6-10 GB or more) and the Distant Horizons guidance (Distant Horizons FAQ: for CPUs with 4-8 cores try 6-8 GB; for CPUs with 10+ cores try 12-16 GB (when the game stutters)). It is not a measurement. If half your RAM is 12 GB or more (installed 24 GB or more), the whole range fits; if half is 6 GB or more, part of it fits.
The shader VRAM is an assumed band of 8 GB to 16 GB, from the 8 GB of the current RTX 5060 and 5050 up to the 16 GB versions of the RTX 5060 Ti and RX 9060 XT. Neither Iris nor Mojang gives a number, so a card inside the whole band gets a check mark and one inside part of it gets a triangle. For streaming, the assumption is the game’s default 4 GB plus 4 to 8 GB left for Windows, OBS and a browser, which all fits from 12 GB installed. OBS gives no fixed number (OBS Studio on Windows needs Windows 10 or 11 and a DirectX 10.1 compatible GPU; CPU requirements vary considerably with encoder, resolution, FPS and scene complexity).
The 8-16 GB shader comparison is for discrete GPUs; shared memory on integrated graphics is not assessed (integrated rows show “Not assessed”).
VRAM figures are NVIDIA’s and AMD’s published values (RTX 5060 Ti 16 GB version 16 GB, 8 GB version 8 GB, RTX 4050 Laptop 6 GB, RTX 5050 Laptop 8 GB). Picks come from new listings on Amazon.com with 100 or more ratings, 4 stars or higher, sold by Amazon or the maker’s own store, whose official specs meet the condition; the pick is the cheapest with shipping included. Listings whose rating count is shared with a different configuration are not counted. The hardware was not tested.

Official values used

Item	Value	Source	Checked
Date Mojang changed the Java Edition requirements	2026-07-21	minecraft.net Published 7/21/26. ‘The recommended system requirements change on July 21’ / ‘Mi	2026-10-09
Performance target of the Java minimum spec	1080p @ 30 FPS, Fast preset	minecraft.net Java System Requirements Target Minimum: ‘1080p @ 30 FPS, Fast preset’	2026-10-09
Performance target of the Java recommended spec	1080p @ 60 FPS, Fancy preset	minecraft.net Java System Requirements Target Recommended: ‘1080p @ 60 FPS, Fancy preset’	2026-10-09
Java minimum RAM (discrete GPU)	8 GB	minecraft.net Memory Minimum: ‘8 GB (with discrete GPU) or 12 GB (with integrated GPU)‘	2026-10-09
Java minimum RAM (integrated GPU)	12 GB	minecraft.net Memory Minimum: ’… or 12 GB (with integrated GPU)‘	2026-10-09
Java recommended RAM	16 GB	minecraft.net Memory Recommended: ‘16 GB’	2026-10-09
Java minimum CPU physical cores	4	minecraft.net Processor Minimum: ’… (4 physical cores minimum)‘	2026-10-09
Java minimum CPU	Intel Core i3-10100, AMD Ryzen 3 3100, Qualcomm Snapdragon X, or Apple M1 (4 physical cores minimum)	minecraft.net Java System Requirements Processor Minimum	2026-10-09
Java recommended CPU	Intel Core i5-12400, AMD Ryzen 5 5600, or Apple M2 Pro	minecraft.net Java System Requirements Processor Recommended	2026-10-09
Java minimum GPU	Vulkan 1.3 capable GPU with at least 2 GB VRAM: NVIDIA GTX 950 or newer, AMD Radeon RX 460 or newer, Intel Arc A310 or newer (discrete) or Intel i5-6400 or newer (integrated), Apple M1 or newer, or Qualcomm Snapdragon X	minecraft.net Java System Requirements Graphics Minimum	2026-10-09
Java recommended GPU	NVIDIA GeForce RTX 2060, AMD Radeon RX 5600 XT, Intel Arc A580, or Apple M2 Pro; 6 GB VRAM	minecraft.net Java System Requirements Graphics Recommended	2026-10-09
Java minimum VRAM	2 GB	minecraft.net ‘Vulkan 1.3 capable GPU with at least 2 GB VRAM’	2026-10-09
Java recommended VRAM	6 GB	minecraft.net Recommended graphics: ’… 6 GB VRAM.‘	2026-10-09
Vulkan version required in the Java requirements table	Vulkan 1.3	minecraft.net Minimum Graphics: ‘Vulkan 1.3 capable GPU’	2026-10-09
What Java 26.2’s experimental Vulkan renderer needs	Vulkan 1.2 with dynamic rendering and push descriptors	minecraft.net Vulkan Support: ‘The current requirement is Vulkan 1.2 with dynamic rendering an	2026-10-09
Java minimum OS	Windows 10 64-bit, Windows on ARM, macOS 12+, or Linux 64-bit	minecraft.net Java System Requirements OS Minimum	2026-10-09
Java recommended OS	Windows 11 64-bit, macOS 14+, or Linux 64-bit	minecraft.net Java System Requirements OS Recommended	2026-10-09
Java minimum storage	2 GB	minecraft.net FAQ ‘How much storage do I need’: ‘Java Edition: 2 GB minimum, 4 GB recommended’	2026-10-09
Java recommended storage	4 GB	minecraft.net FAQ: ‘Java Edition: 2 GB minimum, 4 GB recommended’	2026-10-09
How Mojang treats PCs below the minimum	Java Edition should still launch below the minimum, but performance and visual quality are no longer guaranteed	minecraft.net ‘If your computer falls below the new minimum requirements, Java Edition should	2026-10-09
CPU or GPU load (Mojang FAQ)	Minecraft is more CPU-intensive than GPU-intensive	minecraft.net FAQ: ‘Minecraft is more CPU-intensive than GPU-intensive.‘	2026-10-09
Bedrock minimum CPU	Intel Celeron J4105 ／ AMD FX-4100	minecraft.net Bedrock System Requirements Processor Minimum	2026-10-09
Bedrock recommended CPU	Intel i7-6500U ／ AMD A8-6600K	minecraft.net Bedrock System Requirements Processor Recommended	2026-10-09
Bedrock minimum GPU	Intel HD Graphics 4000 ／ AMD Radeon R5	minecraft.net Bedrock System Requirements Graphics Minimum	2026-10-09
Bedrock recommended GPU	NVIDIA GeForce 940M ／ AMD Radeon HD 8570D	minecraft.net Bedrock System Requirements Graphics Recommended	2026-10-09
Bedrock minimum RAM	4 GB	minecraft.net Bedrock Memory Minimum: 4 GB	2026-10-09
Bedrock recommended RAM	8 GB	minecraft.net Bedrock Memory Recommended: 8 GB	2026-10-09
Bedrock OS (same for minimum and recommended)	Windows 10 version 19041.0 or higher (ARM or x64)	minecraft.net Bedrock System Requirements OS: Windows 10 version 19041.0 or higher; Architectu	2026-10-09
Bedrock minimum storage	4 GB	minecraft.net FAQ: ‘Bedrock Edition: 4 GB minimum, 8 GB recommended’	2026-10-09
Bedrock recommended storage	8 GB	minecraft.net FAQ: ‘Bedrock Edition: 4 GB minimum, 8 GB recommended’	2026-10-09
GPUs for Bedrock ray tracing (Mojang wording)	DirectX capable devices, such as GeForce RTX GPU and select AMD GPUs	minecraft.net	2026-10-10
Vibrant Visuals requirement on PC (Mojang)	PC: running Minecraft on DX12	minecraft.net ‘Vibrant Visuals graphics upgrade is currently only available for Minecraft: Bed	2026-10-09
Vibrant Visuals on Java (Mojang)	Vibrant Visuals is planned for Java Edition in the future	minecraft.net ‘plans to bring it to Java Edition in the future’	2026-10-09
Controller support (Mojang FAQ)	Java Edition has no native controller support; Bedrock supports controller and touch	minecraft.net FAQ: ‘Java Edition doesn’t currently have native controller support at this time	2026-10-09
Crossplay scope (Mojang FAQ)	Bedrock: Windows 10, Windows 11, Xbox, Nintendo Switch, PS5, and mobile; Java: Windows, Mac, Linux only	minecraft.net FAQ ‘Can you play Minecraft cross-platform?‘	2026-10-09
PC edition bundles Java and Bedrock (Mojang)	Minecraft: Java & Bedrock Edition for PC includes both Bedrock and Java; Bedrock runs on Windows only, Java on Windows, Mac and Linux	minecraft.net Store page: Standard Edition lists ‘Minecraft: Bedrock Edition’, ‘Minecraft: Jav	2026-10-09
Java runtime version required (from 26.1)	The game requires Java 25 from 26.1	minecraft.net Published 12/16/25 (26.1 Snapshot 1): ‘The game now requires Java 25’; marks the	2026-10-09
Java 26.1 release date	2026-03-24	minecraft.net Published 3/24/26	2026-10-09
Java default memory allocation (26.1 onward)	4 GB	minecraft.net Default JVM Options: ‘The game now allocates 4 GB of RAM by default (up from 2 G	2026-10-09
Recommended allocation ceiling (Mojang Help)	64-bit OS: half of your RAM is the recommended maximum	help.minecraft.net ‘64-bit operating systems: Half of your RAM is the recommended maximum.’ Change	2026-10-09
Plan to remove OpenGL (Mojang, Feb 18, 2026)	Mojang will remove OpenGL once Vulkan is stable; players will be notified in advance and the minimum requirements will be updated	minecraft.net Published 2/18/26: ‘Once we’re happy with the performance and stability of Vulka	2026-10-09
Experimental Vulkan rendering (Java 26.2)	26.2 (June 16, 2026) added experimental Vulkan rendering and a Graphics API video setting	minecraft.net Published 6/16/26: ‘Added experimental support for rendering the game through Vu	2026-10-09
Impact on mods (Mojang)	Mods that render with OpenGL will take modders more effort to update to Vulkan	minecraft.net ‘Switching from OpenGL to Vulkan will have an impact on the mods that currently	2026-10-09
What Iris supports (Iris official FAQ)	Iris supports Minecraft 1.16.5 and higher; Fabric on 1.16.5+, NeoForge on 1.21.1+; Forge is not supported; Iris ships with Sodium	irisshaders.dev FAQ ‘What versions does Iris support?’ / ‘What mod loaders does Iris support?‘;	2026-10-09
What does not work with Iris (Iris FAQ)	Optifine, Canvas, or Vulkan mods are incompatible with Iris and will crash the game; resource packs that use core shaders will not work with a shader pack enabled	irisshaders.dev FAQ ‘Does insert mod work with Iris?’: ‘Optifine, Canvas, or Vulkan mods: Incomp	2026-10-09
Iris 26.x availability (Modrinth data)	Iris for Fabric has builds for Minecraft 26.1, 26.2 and 26.3 (latest 1.11.7+26.3-fabric, published 2026-09-30)	api.modrinth.com Modrinth API (loaders=fabric): 1.11.7+26.3-fabric 2026-09-30; 1.11.4+26.2-fabric	2026-10-09
Sodium 26.x availability (Modrinth data)	Sodium lists game versions up to 26.3 (project updated 2026-09-20)	api.modrinth.com Modrinth API: game_versions end with 26.2, 26.3-rc-1, 26.3-rc-2, 26.3; updated 2	2026-10-09
CurseForge’s official allocation guidance	Vanilla 2-4 GB; light modded 4-6 GB; heavy modpacks 6-10 GB or more	blog.curseforge.com Published Aug 14, 2025: Vanilla ‘2–4 GB’, Light modded gameplay ‘4–6 GB’, Heavy	2026-10-09
Distant Horizons’ own RAM guidance	Distant Horizons FAQ: for CPUs with 4-8 cores try 6-8 GB; for CPUs with 10+ cores try 12-16 GB (when the game stutters)	modrinth.com Modrinth page FAQ ‘My game is stuttering, help!’: 1. use a concurrent GC such as	2026-10-09
OBS Studio official requirements	OBS Studio on Windows needs Windows 10 or 11 and a DirectX 10.1 compatible GPU; CPU requirements vary considerably with encoder, resolution, FPS and scene complexity	obsproject.com System Requirements: Windows ‘DirectX 10.1 compatible GPU’, ‘Windows 10, or Wind	2026-10-09
RTX 50 series encoder (NVIDIA)	RTX 50 series: 9th-generation NVENC, AV1 encode supported	nvidia.com RTX 50 series table: NVIDIA Encoder (NVENC) 9th Generation (RTX 5060 Ti / 5060 /	2026-10-09
GeForce RTX 5060 Ti (16 GB model) VRAM	16 GB	nvidia.com Standard Memory Config (RTX 5060 Ti): ‘16 GB / 8 GB GDDR7’	2026-10-09
GeForce RTX 5060 Ti (8 GB model) VRAM	8 GB	nvidia.com RTX 5060 Ti: ‘16 GB / 8 GB GDDR7’	2026-10-09
GeForce RTX 5060 VRAM	8 GB	nvidia.com RTX 5060: ‘8 GB GDDR7’	2026-10-09
GeForce RTX 5050 VRAM	8 GB	nvidia.com RTX 5050: ‘8 GB GDDR6’	2026-10-09
GeForce RTX 5070 VRAM	12 GB	nvidia.com RTX 5070: ‘12 GB GDDR7’	2026-10-09
GeForce RTX 4050 Laptop GPU VRAM	6 GB	nvidia.com 40 Series laptop table: RTX 4050 Laptop GPU ‘Standard Memory Config 6 GB GDDR6’	2026-10-09
GeForce RTX 5060 Laptop GPU VRAM	8 GB	nvidia.com 50 Series laptop page: Standard Memory Configuration ‘8 GB GDDR7’ (5060 Laptop G	2026-10-09
Radeon RX 9060 XT (16 GB model) VRAM	16 GB	amd.com AMD page title ‘AMD Radeon RX 9060 XT (16GB)‘	2026-10-09
GeForce RTX 5050 Laptop GPU VRAM	8 GB	nvidia.com 50 Series laptop table: Standard Memory Configuration row ends with ‘8 GB GDDR7	2026-10-09
Older GPUs without Vulkan (Java 26.2)	Vulkan is not supported by older hardware or drivers; OpenGL is used as a fallback	minecraft.net Vulkan Support: ‘Vulkan is not supported by older hardware or drivers - OpenGL w	2026-10-09
Where to see installed RAM (Windows, per Mojang Help)	Start menu Settings System About	help.minecraft.net ‘Verify the installed RAM on your computer. Windows: Start menu Settings System	2026-10-09
RX 9060 XT 16 GB (ASUS Dual) recommended power supply, per ASUS	550W	asus.com	2026-10-10
RX 9060 XT 16 GB (ASUS Dual) auxiliary power connector	1 x 8-pin	asus.com Tech Specs Power Connectors: 1 x 8-pin	2026-10-10
RX 9060 XT 16 GB (ASUS Dual) dimensions	202 x 120 x 49 mm (2.5 slot)	asus.com Tech Specs Dimensions: 202 x 120 x 49mm / Slot: 2.5 Slot	2026-10-10
RX 9060 XT 16 GB (ASRock Challenger) recommended power supply, per ASRock	550W	asrock.com Specification Recommended PSU: 550W	2026-10-10
RX 9060 XT 16 GB (ASRock Challenger) auxiliary power connector	1 x 8-pin	asrock.com Specification Power Connector: 1 x 8-pin	2026-10-10
RX 9060 XT 16 GB (ASRock Challenger) dimensions	249 x 132 x 41 mm	asrock.com Specification Dimensions: 249 x 132 x 41 mm	2026-10-10
Java bundled with the game from 26.1	The included Java distribution is now the Microsoft build of OpenJDK 25	minecraft.net Published 3/24/26 (26.1 release): ‘The game now requires Java 25’; ‘The included	2026-10-10

Minecraft PC spec check: quick reference

The calculator in this article shows the result for your own settings. Settings can be set in the URL; each link below opens the calculator with those settings.

Settings	Result	Calculator
Java Edition, No mods, Desktop, Integrated graphics, RAM 8 GB	Below the minimum. What to add: RAM to 16 GB or more. Below the minimum the game should still launch, but performance and visual quality aren’t guaranteed (Mojang). Java Edition’s requirements table asks for a Vulkan 1.3 capable GPU. The CPU isn’t part of this check. Compare your CPU model with the CPU row (Minecraft is more CPU-intensive than GPU-intensive)	Open with these settings
Java Edition, No mods, Desktop, Integrated graphics, RAM 16 GB	△ Exactly at recommended (borderline). What to add: Nothing. RAM or VRAM is exactly at the recommendation or minimum, with nothing to spare. Java Edition’s requirements table asks for a Vulkan 1.3 capable GPU. The CPU isn’t part of this check. Compare your CPU model with the CPU row (Minecraft is more CPU-intensive than GPU-intensive)	Open with these settings
Java Edition, No mods, Desktop, Discrete GPU, RAM 8 GB, VRAM 4 GB	△ Exactly at the minimum (borderline). What to add: RAM to 16 GB or more; A graphics card with 6 GB+ VRAM. RAM or VRAM is exactly at the recommendation or minimum, with nothing to spare. Java Edition’s requirements table asks for a Vulkan 1.3 capable GPU. The CPU isn’t part of this check. Compare your CPU model with the CPU row (Minecraft is more CPU-intensive than GPU-intensive)	Open with these settings
Java Edition, No mods, Desktop, Discrete GPU, RAM 8 GB, VRAM 8 GB	△ Exactly at the minimum (borderline). What to add: RAM to 16 GB or more. RAM or VRAM is exactly at the recommendation or minimum, with nothing to spare. Java Edition’s requirements table asks for a Vulkan 1.3 capable GPU. The CPU isn’t part of this check. Compare your CPU model with the CPU row (Minecraft is more CPU-intensive than GPU-intensive)	Open with these settings
Java Edition, No mods, Laptop, Discrete GPU, RAM 16 GB, VRAM 6 GB	△ Exactly at recommended (borderline). What to add: Nothing. RAM or VRAM is exactly at the recommendation or minimum, with nothing to spare. Java Edition’s requirements table asks for a Vulkan 1.3 capable GPU. The CPU isn’t part of this check. Compare your CPU model with the CPU row (Minecraft is more CPU-intensive than GPU-intensive)	Open with these settings
Java Edition, No mods, Desktop, Discrete GPU, RAM 16 GB, VRAM 8 GB	△ Exactly at recommended (borderline). What to add: Nothing. RAM or VRAM is exactly at the recommendation or minimum, with nothing to spare. Java Edition’s requirements table asks for a Vulkan 1.3 capable GPU. The CPU isn’t part of this check. Compare your CPU model with the CPU row (Minecraft is more CPU-intensive than GPU-intensive)	Open with these settings
Java Edition, Heavy modpack, Desktop, Discrete GPU, RAM 16 GB, VRAM 8 GB	△ Exactly at recommended (borderline). What to add: RAM to 24 GB or more. RAM or VRAM is exactly at the recommendation or minimum, with nothing to spare. Java Edition’s requirements table asks for a Vulkan 1.3 capable GPU. The CPU isn’t part of this check. Compare your CPU model with the CPU row (Minecraft is more CPU-intensive than GPU-intensive)	Open with these settings
Java Edition, Heavy modpack, Desktop, Discrete GPU, RAM 32 GB, VRAM 8 GB	✓ Meets recommended. What to add: Nothing. Java Edition’s requirements table asks for a Vulkan 1.3 capable GPU. The CPU isn’t part of this check. Compare your CPU model with the CPU row (Minecraft is more CPU-intensive than GPU-intensive)	Open with these settings
Java Edition, Shaders, Desktop, Discrete GPU, RAM 16 GB, VRAM 8 GB	△ Exactly at recommended (borderline). What to add: A graphics card with 16 GB+ VRAM. RAM or VRAM is exactly at the recommendation or minimum, with nothing to spare. Java Edition’s requirements table asks for a Vulkan 1.3 capable GPU. The CPU isn’t part of this check. Compare your CPU model with the CPU row (Minecraft is more CPU-intensive than GPU-intensive)	Open with these settings
Java Edition, Shaders, Desktop, Discrete GPU, RAM 32 GB, VRAM 16 GB	✓ Meets recommended. What to add: Nothing. Java Edition’s requirements table asks for a Vulkan 1.3 capable GPU. The CPU isn’t part of this check. Compare your CPU model with the CPU row (Minecraft is more CPU-intensive than GPU-intensive)	Open with these settings
Java Edition, While streaming, Desktop, Discrete GPU, RAM 16 GB, VRAM 8 GB	△ Exactly at recommended (borderline). What to add: Nothing. RAM or VRAM is exactly at the recommendation or minimum, with nothing to spare. Java Edition’s requirements table asks for a Vulkan 1.3 capable GPU. The CPU isn’t part of this check. Compare your CPU model with the CPU row (Minecraft is more CPU-intensive than GPU-intensive)	Open with these settings
Bedrock Edition, Desktop, Integrated graphics, RAM 4 GB	△ RAM exactly at the minimum (borderline). What to add: RAM to 8 GB or more. RAM or VRAM is exactly at the recommendation or minimum, with nothing to spare. For Bedrock, compare the GPU and CPU by model. This check covers RAM only. The CPU isn’t part of this check. Compare your CPU model with the CPU row (Minecraft is more CPU-intensive than GPU-intensive)	Open with these settings
Bedrock Edition, Desktop, Integrated graphics, RAM 8 GB	△ RAM exactly at recommended (borderline). What to add: Nothing. RAM or VRAM is exactly at the recommendation or minimum, with nothing to spare. For Bedrock, compare the GPU and CPU by model. This check covers RAM only. The CPU isn’t part of this check. Compare your CPU model with the CPU row (Minecraft is more CPU-intensive than GPU-intensive)	Open with these settings

Sony Walkman NW-A105 review — the Type-C, Android DAP I switched to, revisited from 2026

Sat, 24 Apr 2021 00:00:00 GMT

2026 update: Ported from the old VuePress blog (originally written in 2021, after about half a year of use). As of 2026, the current A-series line is the NW-A300 (NW-A306 in Japan, A307 internationally, released 2023). The NW-A105 itself is discontinued, with new stock only while supplies last. The bones of this review still hold up — just read it alongside the current model when considering a purchase.

I used the Sony Walkman NW-A105 for over five years, for music on the commute and on the road. It was the first Walkman generation to combine USB Type-C, microSD, and Android — the model where the connector finally caught up with the rest of the world.

Looking back from 2026, though, the strengths and weaknesses sit on very different sides of the ledger. This is a long-term impression, compared against the current NW-A300 line and against going phone-only.

The verdict — a first-generation Type-C / microSD / Android DAP; for a new purchase, go A300

Short answer: the NW-A105 was valuable as the first generation of Walkman with Type-C and Android. The sound-tuning depth still holds up. But the CPU and battery let it down, and for a new purchase in 2026, the current NW-A300 line is the easier recommendation.

The reasoning is that the NW-A105’s weak points — sluggish responsiveness and high idle drain — are tied to internal parts. They are not the kind of thing a firmware update can fix at the root. The later battery-saver helped operationally, but the underlying issue is still there.

Who should and should not buy one:

Worth it: anyone who can find a cheap used unit and wants wired earphones, microSD, and deep sound tuning
Skip it: anyone buying new, anyone whose music life is already on a phone, anyone who prioritises battery life

Build and layout

Short answer: a small Android device with a 3.6-inch touchscreen. The physical buttons cluster on the right edge; the bottom carries the 3.5 mm jack, USB Type-C, and microSD slot.

A hi-res-capable, Android-powered Walkman. Most operations happen on the touchscreen.

Spec	Value
Display	3.6 inch
Dimensions	55.9 × 98.9 × 11 mm
Weight	103 g
OS	Android 9 (at launch)
Ports	USB Type-C / 3.5 mm jack / microSD

(rated)

The back is matte, with a Walkman logo. Fingerprints stay out of sight.

The left flank is bare — no buttons.

All the operating buttons live on the right. From left to right: hold / back / play-pause / forward / volume down / volume up / power. That layout lets you control playback blind, with the unit still in a bag or pocket — exactly what a DAP should get right.

The bottom edge holds the 3.5 mm jack, a strap hole, USB Type-C, and the microSD slot. Keeping the 3.5 mm jack matters a lot if you live on wired earphones.

The home screen sticks close to stock Android. You can add apps freely (Google Play is supported).

What worked

Short answer: the Type-C / microSD / Android trio, plus the depth of Sony’s sound-processing options.

USB Type-C — the reason I switched

This was the deciding factor when I bought it. Earlier Walkman models used Sony’s proprietary WM-PORT, which meant carrying a dedicated cable everywhere.

Moving to Type-C means the same cable as the phone and the tablet. The “I forgot the Walkman cable on a business trip” disaster simply stops happening. That alone changed how I packed.

microSD adds storage after the fact

Internal storage is a modest 16 GB. Trying to manage a hi-res library on internal storage alone is not realistic.

Put another way: the device assumes a microSD card. The internal memory is for OS and apps; music lives on the card. With support up to 1 TB microSDXC, there is enough headroom for a serious library.

Sound tuning runs deep

This is where a Sony DAP earns its keep.

The equaliser is roughly ten bands of fine adjustment, with genre presets on top.

Noise cancelling assumes Sony’s compatible earphones (the IER-NW500N and similar). On a commuter train, it knocks the noise floor down hard.

The main sound-processing features:

DSEE HX: upscales compressed audio toward hi-res
DC Phase Linearizer: corrects phase characteristics to mimic an analogue amp
Dynamic Normalizer: evens out volume differences between tracks
Vinyl Processor: shapes the sound to feel like a record
ClearAudio+: Sony’s recommended auto-optimisation preset

What you learn after living with it is that “everything on” is not always best. Toggling DSEE HX on or off per genre, for example, is where the sound changes character if you put in the time.

Android opens it up

It runs Android 9 at the base. Spotify, Apple Music, and YouTube Music install through Google Play. A single device handles a local library and streaming side by side.

Bluetooth pairing for handing audio off from a phone works the way you would expect.

What didn’t

Short answer: battery life and CPU speed. Both come down to hardware; neither is fixable in software.

Battery drain is high

The biggest complaint.

Just leaving it powered on visibly drains the battery. Idle for a day without playback still costs a meaningful chunk of the meter.

Sony quotes around 26 hours of hi-res playback. In real use, what hurts is the consumption during the hours you are not listening. A later firmware update added a battery saver that powers the unit off after a set period of inactivity, and that helped a lot operationally. It is still not a fix at the root.

For commuting, where you charge at home or at work every day, this is fine. For a two-day business trip without an easy charge, it is tight.

A general sluggishness

The CPU is modest, and it shows up as friction when opening apps, switching screens, or scrolling a large library.

It is not a device you tap constantly the way you do a phone, so it never becomes critical — but opening Spotify, tapping the search bar, and typing into it feels heavier than it should.

Software support has ended

As of 2026, the NW-A105 is still on Android 9, with no major updates. From a security standpoint, it is not suitable as a primary phone substitute.

As a dedicated music player, that is acceptable. The long-term compatibility of streaming apps, though, is not guaranteed.

Comparison — NW-A105 vs NW-A300 series vs phone-only

Short answer: buy new and the NW-A300 wins. Want the cheapest path that still works, used NW-A105. Want a single device for everything, the phone is fine.

Angle	NW-A105 (2019)	NW-A300 series (2023–)	Phone-only
OS	Android 9	Android 12 base	Latest per device
Perceived CPU speed	Sluggish	Improved	Depends on device
Battery (rated hi-res)	~26 hours	~36 hours (rated)	Shared with everything else
Internal storage	16 GB	32 GB (rated)	Depends on device
microSD	Yes (up to 1 TB)	Yes	Usually no
Ports	USB Type-C / 3.5 mm	USB Type-C / 3.5 mm	Type-C only on most
Noise cancelling (Sony earphones)	Supported	Supported	Not supported
DSEE HX-style processing	Yes	Yes (now DSEE Ultimate)	No
Price (2026 expectation)	Mostly used	New, ~mid-40,000s yen and up	Already in the phone

The decision splits cleanly:

You care about sound processing and a wired chain, and you do not want music draining the phone → NW-A300
Same requirements, minimum cost → NW-A105 used
Streaming-centric, wireless earphones, one device for everything → phone-only. You do not need a DAP

FAQ

Q. Is the NW-A105 worth buying new in 2026? A. For a new purchase, the current NW-A300 line (NW-A306 in Japan / A307 internationally) is the easier recommendation. The NW-A105 was released in 2019, and its CPU and battery life are both weak. If you only want a DAP with microSD, Type-C, and Android, and you can find a cheap used unit, it is still an option.

Q. Compared with listening on a phone, is there a reason to carry an NW-A105? A. Yes, if you want Sony’s noise cancelling or DSEE HX sound processing, or if you do not want music playback eating your phone’s battery. If your library lives entirely in streaming and you do not use wired earphones, a phone alone is enough.

Q. How long does the battery last? A. Sony lists about 26 hours of hi-res playback. In practice, idle drain is high — leaving it powered on for a day burns through a visible chunk even without playing. A later firmware update added a battery-saver auto-off feature.

Q. What is the maximum microSD capacity? A. Up to 1 TB microSDXC. Internal storage is only 16 GB, so any serious hi-res library has to live on the card — the design assumes you will use one.

Verdict — a device for experiencing the first Android-on-a-DAP generation

The NW-A105 belongs to the generation where “the Walkman finally went Type-C” and “the Walkman runs Android” arrived together. As a reason to buy at the time, that was enough, and after five-plus years it never stopped being a reliable home for the library.

Would I recommend buying one new today? No. The CPU and the battery are tied to the internal design, and the successor generation improved both directly. Either go cheap on a used unit, or step up to the current NW-A300 — those are the two sensible paths.

If your phone, paired with streaming and wireless earphones, already covers your music, there is no reason to add a DAP. Carrying one only earns its keep if you are unwilling to give up one of the three: wired earphones, deep sound processing, or microSD.

HDD vs SSD: Which to Buy, With Lifespan and Capacity Math

Wed, 21 Apr 2021 00:00:00 GMT

Per the makers’ official specs, put the OS, apps and games on an NVMe SSD, and keep large photo and video collections on a hard drive.

Quick table by use

For an internal drive, check your PC’s connection type and supported size first.

Use	Best drive type
OS and apps	NVMe M.2 SSD
PC games	NVMe M.2 SSD
PS5 expansion	PCIe Gen4x4 M.2 SSD with heatsink
Upgrading an older PC	2.5-inch SATA SSD
Photo and video storage	Hard drive (CMR for a NAS)
Backups	External hard drive
Carrying data around	Portable SSD

NVMe, SATA and hard drive types

An SSD stores data on flash chips, and a hard drive (HDD) stores it on spinning disks. NVMe SSDs plug into a slim M.2 slot, and SATA is the connection used by 2.5-inch SSDs and hard drives. External drives connect over USB.

Per SATA-IO and Samsung’s official specs, a SATA 2.5-inch SSD tops out at 600 MB/s (MB/s is how much it can read in a second). Comparing rated maximum read speeds of Samsung 2TB SSDs, the SATA model reads at 560 MB/s and the PCIe 4.0 NVMe model at 7,450 MB/s, a gap of about 13x.

Upgrading an older PC

Per Samsung’s official pages, a PC whose spec sheet lists a 2.5-inch SATA drive can take a 2.5-inch SATA SSD, which also works on older SATA 3Gb/s ports.

CMR and SMR hard drives

SMR is a recording method that overlaps the tracks on the disk, and CMR is the conventional method.

Per Seagate’s and WD’s official specs, for always-on use like a NAS (a disk on your home network), pick a NAS hard drive rated for 180 TB/year. The annual workload is the total amount read and written in a year.

Per Seagate’s official CMR/SMR list, the NAS-oriented IronWolf uses CMR at every capacity, while SMR shows up in desktop BarraCuda drives such as the 4TB and 8TB. The annual workload rating is 180 TB/year for IronWolf and 55 TB/year for BarraCuda.

Check it for your setup

Follow these steps to choose a drive type and capacity.

For an internal drive, check the spec sheet for the connection type, SATA or NVMe/PCIe, and the size, 2.5-inch or M.2
Pick your use from the rows of the table
Check whether each game you play requires an SSD
Add up the data you’ll store, as Windows shows it. Windows 11’s minimum is 64 GB, and for games, add the free space each one lists on its store page
Divide the data by 0.8 (that leaves 20% free), then compare the result with how Windows will show the drive. Windows labels sizes “GB” but counts in units of 1024³ bytes, so a drive sold as 1TB shows up as about 931 GB
For an SSD, estimate how much you write per day for the TBW calculation

HDD or SSD? (use, capacity and how long it lasts)

Pick how you'll use the drive (OS, games, PS5, upgrading an older PC, photo and video storage, backups or carrying files) to see which type fits. From how much data you'll store it works out the capacity, and from how much you write a day it estimates the years to the SSD's TBW rating or compares against a hard drive's annual workload rating. It also shows one product that fits.

SSD lifespan, capacity and long-term storage

How many years does an SSD last?

An SSD’s TBW rating can’t tell you how many years it will last before it fails.

In this calculation, assuming 10 GB to 200 GB written a day, the 5-year warranty ends before the TBW rating is reached. That covers two Samsung drives: a 2TB model rated 1,200 TBW and a 1TB model rated 600 TBW. The 1TB model’s 250GB and 500GB versions are rated lower, at 150 TBW and 300 TBW, and at 200 GB a day they reach the TBW in under 5 years.

Per Samsung’s warranty terms, TBW is the total amount of writing the maker’s warranty covers. It isn’t the point where the drive dies, and the warranty ends at the years or the TBW, whichever comes first.

In this calculation, the warranty (5 years) ends before the TBW is reached in every row. Years are rounded down.

Written per day (assumed)	Years to reach the TBW (2TB, 1,200 TBW)	Years to reach the TBW (1TB, 600 TBW)
10 GB	about 328 years	about 164 years
50 GB	about 65 years	about 32 years
200 GB	about 16 years	about 8 years

How big an SSD do I need?

With 25% left free, a 1TB SSD holds up to 698 GB as Windows shows it.

The steps leave 20% free; this limit uses 25%.

Pick the smallest size that is bigger than your data plus the free space. Leaving 25% free, data up to 698 GB as Windows shows it fits on 1TB, and up to 1,396 GB fits on 2TB. In the makers’ label units, those limits are 750 GB and 1,500 GB.

Per the system requirements on Steam, Starfield, for example, requires an SSD and 125 GB of free space.

Backups and long-term storage

Among the models checked in October 2026, an external hard drive is a candidate for backups. None of the external drives checked state their recording technology.

Per Dell’s support article, which cites JEDEC, an SSD written to its rated limit must hold data for at least three months unpowered, stored at up to 40°C. Dell advises powering a drive that holds data about every 2.5 months for at least 3 weeks. If you leave an SSD sitting, keep a copy on another drive too.

Adding storage to a PS5 or Xbox

PS5 expansion

Per PlayStation’s official support page, the PS5 takes a PCIe Gen4x4 M.2 NVMe SSD from 250 GB to 8 TB, in sizes 2230 to 22110. 5,500 MB/s or faster sequential read is recommended, and it needs a heatsink or similar cooling. With the heatsink on, the SSD can be at most 25 mm wide and 11.25 mm high.

PS5 games moved to USB extended storage have to be copied back to the console or an M.2 SSD before you can play them.

Xbox external drives

Per the Xbox Wire comparison table (September 2020), the Xbox Series X|S can play Xbox One-era games straight from a USB 3.1 drive. Games optimized for Series X|S need the internal SSD or the Storage Expansion Card.

Picks by condition

One listing per use, chosen from Amazon.com listings checked on October 9, 2026 against these conditions:

The maker’s official page lists the specs
Sold by Amazon.com
At least 100 ratings and 4 stars or more

Each pick showed as new that day. Check prices on the store.

OS, apps and games on a PC

It is PCIe 4.0 without a heatsink.

PS5 expansion

The maker states PS5 compatibility, and it reads at 5,500 MB/s or faster.

Carrying data around

It is a USB 3.2 Gen 2 portable SSD that weighs 98 g and passed a 3 m drop in Samsung’s own test.

Photos and videos inside a PC or NAS

It is a 3.5-inch internal hard drive listed as CMR, rated 180 TB/year. For USB, pick an external drive.

Backups

The maker leaves recording technology blank in its spec table, so it is not a NAS pick.

FAQ

Should I buy an HDD or an SSD?

Per the makers’ official specs, put the OS, apps and games on an NVMe SSD, and keep large photo and video collections on a hard drive.

How many years does an SSD last?

An SSD’s TBW rating can’t tell you how many years it will last before it fails. In this calculation, assuming 10 GB to 200 GB written a day, the 5-year warranty ends before the TBW rating is reached.

That covers two Samsung drives: a 2TB model rated 1,200 TBW and a 1TB model rated 600 TBW. The 1TB model’s 250GB and 500GB versions are rated lower, at 150 TBW and 300 TBW, and at 200 GB a day they reach the TBW in under 5 years.

How big an SSD do I need?

In this calculation, leaving 25% free, data up to 698 GB as Windows shows it fits on 1TB, and up to 1,396 GB fits on 2TB. In the makers’ label units, those limits are 750 GB and 1,500 GB.

How do I add storage to a PS5?

With the heatsink on, the SSD can be at most 25 mm wide and 11.25 mm high.

Does an SSD lose data if it sits unplugged?

If you leave an SSD sitting, keep a copy on another drive too.

Data and formulas used (for verification)

How this was checked: Worked out from the makers’ official pages as of October 2026. No drive was tested in person. The Amazon.com heatsink 990 PRO is the C model (MZ-V9P2T0CW), and Samsung’s US page for it shows PS5 Compatible.

Sources: SATA-IO ／ Samsung 990 PRO ／ Samsung 870 EVO ／ PlayStation M.2 SSD ／ Xbox Wire ／ Seagate CMR/SMR list ／ Dell article on SSD data retention

Method

Years until an SSD reaches its TBW = TBW × 1000 ÷ (GB written per day × 365). Daily writing is an assumption, checked across 10 GB to 200 GB a day
A hard drive’s yearly reads and writes (TB) = GB read and written per day × 365 ÷ 1000. Assumed 20 GB to 400 GB a day, compared with the annual workload rating (reads plus writes combined)
Capacity needed = data to store ÷ (1 − share of space left free). Data is measured the way Windows shows it (1 GB = 1024³ bytes), so it is converted to the label’s units (1 GB = 1 billion bytes) before dividing. The 1TB cutoff is 698 GB as Windows shows it = floor(1000 × 0.75 ÷ 1.073741824), and the 2TB cutoff is 1,396 GB = floor(2000 × 0.75 ÷ 1.073741824); in label units they are 750 GB and 1,500 GB. Free space was checked across 10% to 25%, and the smallest capacity at or above the result was picked
Not examined: hard drive speed and lifespan, PCIe 5.0 or real-world SSD speed gains and game load times, how long a hard drive keeps data unpowered, and whether recording TV needs CMR. The Amazon.com heatsink 990 PRO is the C model (MZ-V9P2T0CW), and Samsung’s US page for that model carries the PS5 Compatible label. Its TBW is not used because Samsung’s pages disagree
Speed ratio = 990 PRO 2TB read ÷ 870 EVO 2TB read (both are the makers’ rated maximums, rounded down)
No drive was tested in person. Values come from published pages and documents of the makers, PlayStation, Xbox Wire, Steam, Microsoft and Dell (checked 2026-10-09)

Assumptions

How much is written to the SSD per day. The makers publish no standard figure, so this is an assumption of this article: 10 GB is office and browsing, 50 GB is typical use including game updates and photo sorting, and 200 GB is heavy use such as video editing or swapping large games: 50 (10–200)
A drive filled to the brim is said to slow down and wear faster, but no official figure was found. This article assumes 20% of the space is left free (checked from 10% to 25%). It is only a margin for the capacity math, not a maker recommendation: 0.2 (0.1–0.25)
Assumed GB a hard drive reads and writes per day, to compare with its annual workload rating (reads plus writes). There is no official standard figure, so this is an assumption of this article, modeled on a household using an external drive for backups, photos and recordings: 100 (20–400)

Official values used

Item	Value	Source	Checked
SATA 3.0 link speed (bits)	6 Gb/s	sata-io.org	2026-10-09
SATA 6Gb/s ceiling	600 MB/s	sata-io.org	2026-10-09
Samsung 990 PRO 2TB sequential read (max)	7450 MB/s	samsung.com	2026-10-09
Samsung 990 PRO 2TB sequential write (max)	6900 MB/s	samsung.com Sequential Write Up to 6,900 MB/s	2026-10-09
Samsung 990 PRO 2TB warranted endurance	1200 TBW	samsung.com Warranty: 5-Year Limited Warranty or 1200 TBW Limited Warranty（2TB）	2026-10-09
Samsung 990 PRO warranty length	5 years	samsung.com 5-Year Limited Warranty or 1200 TBW	2026-10-09
Samsung 990 PRO MTBF	1,500,000 hours	samsung.com Reliability (MTBF): 1.5 Million Hours	2026-10-09
Samsung 990 PRO 2TB average power	5.5 W	samsung.com Average Power Consumption (system level): Average 5.5 W, Maximum 8.5 W (Burst mo	2026-10-09
Samsung 990 PRO 2TB weight (max)	9 g	samsung.com Weight: Max 9.0g	2026-10-09
Samsung 990 PRO shock rating (0.5 ms)	1500 G	samsung.com Shock: 1,500 G & 0.5 ms (Half sine)	2026-10-09
990 PRO with heatsink 2TB (MZ-V9P2T0CW) weight (max)	28 g	samsung.com 990 PRO with Heatsink 2TB MZ-V9P2T0CW: Weight Max 28.0g, 80 x 24.3 x 8.2 mm	2026-10-09
990 PRO heatsink: G and C model pages disagree on TBW	Samsung US page for MZ-V9P2T0GW (990 PRO with Heatsink 2TB) lists 5-Year Limited Warranty or 600 TBW and 1GB cache, while the Singapore MZ-V9P2T0CW page lists 1200 TBW and the US MZ-V9P2T0CW page lists 600 TBW; the pages disagree	samsung.com	2026-10-09
990 PRO with heatsink (G model): PS5 support and speed	MZ-V9P2T0GW (990 PRO w/ Heatsink 2TB): PS5 Compatible; 7,450MB/s Read, 6,900MB/s Write; PCIe 4.0 NVMe M.2	samsung.com	2026-10-09
990 PRO with heatsink C model (MZ-V9P2T0CW, US page): PS5 support and speed	MZ-V9P2T0CW (990 PRO w/ Heatsink 2TB, Samsung US page): PS5 Compatible; 7,450MB/s Read, 6,900MB/s Write; PCIe 4.0 NVMe M.2, 2280	samsung.com	2026-10-09
990 PRO with heatsink C model (MZ-V9P2T0CW) thickness	8.2 mm	samsung.com MZ-V9P2T0CW: 80 x 24.3 x 8.2 mm (with heatsink)	2026-10-09
Samsung 990 EVO Plus 2TB sequential read (max)	7250 MB/s	samsung.com Sequential Read Up to 7,250 MB/s（2TB）	2026-10-09
Samsung 990 EVO Plus 2TB sequential write (max)	6300 MB/s	samsung.com Sequential Write Up to 6,300 MB/s（2TB）	2026-10-09
Samsung 990 EVO Plus 2TB warranted endurance	1200 TBW	samsung.com 5-year Limited Warranty or 1200 TBW Limited Warranty（2TB）	2026-10-09
Samsung 990 EVO Plus interface	PCIe 4.0 x4 / 5.0 x2, NVMe 2.0	samsung.com Interface: PCIe® 4.0 x4 / 5.0 x2 NVMe™ 2.0	2026-10-09
WD_BLACK SN850X 2TB sequential read (max)	7300 MB/s	sandisk.com	2026-10-09
WD_BLACK SN850X 2TB sequential write (max)	6600 MB/s	sandisk.com Sequential Write 6600MB/s	2026-10-09
WD_BLACK SN850X 2TB TBW	1200 TBW	sandisk.com Endurance (TBW) 1200、Warranty 5-Year Limited Warranty、Weight 7.5gms	2026-10-09
WD_BLACK SN8100 2TB sequential read (max)	14900 MB/s	sandisk.com	2026-10-09
WD_BLACK SN8100 2TB sequential write (max)	14000 MB/s	sandisk.com Sequential Write 14000MB/s	2026-10-09
WD_BLACK SN8100 2TB TBW	1200 TBW	sandisk.com Endurance (TBW) 1200、Warranty 5-Year Limited Warranty	2026-10-09
WD_BLACK SN8100 2TB with heatsink weight	31.2 g	sandisk.com	2026-10-09
Samsung 9100 PRO 2TB sequential read (max)	14700 MB/s	samsung.com	2026-10-09
Samsung 9100 PRO 2TB sequential write (max)	13400 MB/s	samsung.com Sequential Write Up to 13,400 MB/s	2026-10-09
Samsung 9100 PRO 2TB warranted endurance	1200 TBW	samsung.com 5-year Limited Warranty or 1,200TBW Limited Warranty	2026-10-09
Samsung 9100 PRO 2TB average power (read)	8.1 W	samsung.com Average Power Consumption (system level): Read 8.1 W / Write 7.9 W	2026-10-09
Samsung 870 EVO 2TB sequential read (max)	560 MB/s	samsung.com Sequential Read Up to 560 MB/s	2026-10-09
Samsung 870 EVO 2TB sequential write (max)	530 MB/s	samsung.com	2026-10-09
Samsung 870 EVO 2TB warranted endurance	1200 TBW	samsung.com	2026-10-09
Samsung 870 EVO 1TB warranted endurance	600 TBW	samsung.com	2026-10-09
Samsung 870 EVO warranty length	5 years	samsung.com 5 Years Limited Warranty or 1,200 TBW	2026-10-09
Samsung 870 EVO MTBF	1,500,000 hours	samsung.com Reliability (MTBF): 1.5 Million Hours	2026-10-09
Samsung 870 EVO 2TB average power	2.5 W	samsung.com	2026-10-09
Samsung 870 EVO 2TB weight	46 g	samsung.com Weight: Approx. 46.0g	2026-10-09
Samsung 870 EVO shock rating (0.5 ms)	1500 G	samsung.com Shock: 1,500 G & 0.5 ms (Half sine)	2026-10-09
Samsung T7 Shield 2TB sequential read (max)	1050 MB/s	samsung.com Sequential Read: Up to 1,050MB/s, Sequential Write: Up to 1,000 MB/s	2026-10-09
Samsung T7 Shield 2TB sequential write (max)	1000 MB/s	samsung.com Sequential Write: Up to 1,000 MB/s	2026-10-09
Samsung T7 Shield warranty length (US page)	3 years	samsung.com Warranty: Limited 3-year	2026-10-09
Samsung T7 Shield weight	98 g	samsung.com Weight: 98g (The exact weight of the product may vary by capacity)、59 x 88 x 13m	2026-10-09
Samsung T7 Shield drop test height	3 m	samsung.com	2026-10-09
SanDisk Extreme Portable SSD 2TB (SDSSDE61) read (max)	1050 MB/s	sandisk.com	2026-10-09
SanDisk Extreme Portable SSD warranty length	5 years	sandisk.com 5-Year Limited Warranty	2026-10-09
WD Red Plus 4TB recording technology	CMR	westerndigital.com WD40EFPX: Recording Technology CMR、5400 RPM、Cache 256MB	2026-10-09
WD Red Plus annual workload rating	180 TB/year	westerndigital.com Rated for 180TB/year workload and 1M hours MTBF	2026-10-09
WD Red Plus MTBF	1,000,000 hours	westerndigital.com 1M hours MTBF	2026-10-09
WD Red Plus warranty length	3 years	westerndigital.com 3-Year Limited Warranty	2026-10-09
WD Red Plus 4TB (WD40EFPX) weight	570 g	westerndigital.com WD40EFPX: Weight 570gms	2026-10-09
WD Red Plus 4TB model numbers (cache variants)	4TB 256MB: WD40EFPX; 4TB 128MB: WD40EFZZ (both CMR, 5400 RPM class)	westerndigital.com Model Numbers: 4TB, 256MB: WD40EFPX / 4TB, 128MB: WD40EFZZ	2026-10-09
WD Red Plus 8TB (WD80EFPX) specs	WD80EFPX (8TB): CMR, workload rate 180 TB/year, 3-year limited warranty, MTBF 1,000,000 hours	documents.westerndigital.com	2026-10-09
WD Elements Portable 4TB (WDBU6Y0040BBK-WESN) specs	WDBU6Y0040BBK-WESN (WD Elements Portable 4TB): USB 3.2 Gen 1 / USB 3.0 / USB 2.0, 2-Year Limited Warranty, 0.51 lbs; recording technology left blank on the spec table	westerndigital.com	2026-10-09
Samsung T7 Shield interface	USB 3.2 Gen.2 (10Gbps)	samsung.com	2026-10-09
Seagate IronWolf 4TB (ST4000VN006) recording technology	CMR	seagate.com	2026-10-09
Seagate IronWolf annual workload rating	180 TB/year	seagate.com	2026-10-09
Seagate IronWolf warranty length (Japan page)	3 years	seagate.com	2026-10-09
Seagate IronWolf MTBF	1,000,000 hours	seagate.com	2026-10-09
Seagate BarraCuda 4TB (ST4000DM004) recording technology	SMR	seagate.com	2026-10-09
Seagate BarraCuda annual workload rating	55 TB/year	seagate.com Rated workload: Average annualized workload rating: 55 TB/year	2026-10-09
Seagate capacities that use SMR (BarraCuda, FireCuda)	3.5-inch BarraCuda: SMR 4TB and 8TB (CMR 12/16/20/24TB); 2.5-inch BarraCuda: SMR 2TB and 4TB; 2.5-inch FireCuda: SMR 1TB and 2TB	seagate.com	2026-10-09
Seagate largest CMR HDD (Exos, HAMR)	32 TB	seagate.com	2026-10-09
Seagate Expansion Desktop product weight	1.176 kg	seagate.com	2026-10-09
Seagate Expansion Desktop 4TB warranty by region	Limited warranty by region: AMER 1 year, EMEA 2 years, APAC 3 years (model STKP4000400)	seagate.com	2026-10-09
Buffalo HD-AD4U3 weight (approx.)	900 g	buffalo.jp	2026-10-09
Buffalo HD-AD4U3 warranty length	1 year	buffalo.jp	2026-10-09
Buffalo HD-AD4U3 TV recording support listing	Buffalo official page (HD-AD4U3, Japan-only model): works with compatible TVs and recorders for recording and playback; recording technology (CMR/SMR) is not stated	buffalo.jp	2026-10-09
PS5 M.2 SSD recommended sequential read	5500 MB/s	playstation.com Sequential read speed 5,500MB/s or faster is recommended	2026-10-09
PS5 M.2 SSD supported capacity	250GB to 8TB	playstation.com Capacity 250 GB - 8 TB	2026-10-09
PS5 M.2 SSD requirements (interface, size, cooling)	PCI-Express Gen4x4 M.2 NVMe SSD (Key M); sizes 2230, 2242, 2260, 2280, 22110; heatsink or equivalent cooling required; total size up to 25mm wide and 11.25mm high	playstation.com PCI-Express Gen4x4 supported M.2 NVMe SSD (Key M)、effective heat dissipation wit	2026-10-09
PS5 games on USB extended storage	PS5 games stored on USB extended storage must be copied back to console storage or M.2 SSD storage to be played; PS5 games can be played directly from an M.2 SSD	playstation.com USB Extended Storage vs M.2 SSD Storage	2026-10-09
Xbox external USB storage and games	External USB 3.1 HDD/SSD: play Xbox One games including backward-compatible Xbox 360 and original Xbox games directly from it; Xbox Series X／S optimized games can be stored on USB 3.1 storage for quick transfer but not played from it (comparison table: Plays games optimized for Xbox Series X／S - No); they need the internal SSD or the Storage Expansion Card	news.xbox.com	2026-10-09
Windows 11 minimum storage	64 GB	microsoft.com Storage: 64 GB or larger storage device	2026-10-09
DirectStorage requirement	DirectStorage requires an NVMe SSD to store and run games that use the Standard NVM Express Controller driver and a DirectX12 GPU with Shader Model 6.0 support	microsoft.com	2026-10-09
Starfield required free space (SSD required)	125 GB	store.steampowered.com	2026-10-09
Baldur’s Gate 3 required free space (SSD required)	150 GB	store.steampowered.com Storage 150 GB available space, Additional Notes: SSD required	2026-10-09
Cyberpunk 2077 required free space (SSD required)	70 GB	store.steampowered.com Storage 70 GB available space, Additional Notes: SSD required	2026-10-09
Monster Hunter Wilds required free space (SSD required)	75 GB	store.steampowered.com Storage 75 GB available space, Additional Notes: SSD required	2026-10-09
Hogwarts Legacy required free space (HDD supported at minimum)	85 GB	store.steampowered.com	2026-10-09
How capacity is labeled (1 GB = 1 billion bytes)	1GB = 1,000,000,000 bytes and 1TB = 1,000,000,000,000 bytes on the label; the capacity the OS shows is less	samsung.com Capacity: 2,000GB (1GB=1 Billion byte by IDEMA) * Actual usable capacity may be	2026-10-09
Unpowered SSD retention (Dell article citing JEDEC)	Dell’s support article, citing JEDEC JESD218/JESD219, says SSD/NVMe drives must retain data for a minimum of three months powered off at the maximum rated endurance, stored at no more than 40C; Dell advises powering drives that hold user data about every 2.5 months for at least 3 weeks	dell.com	2026-10-09
Crucial consumer business exit announcement	Micron announced on 2025-12-03 that it will exit the Crucial consumer business; Crucial consumer products ship through the consumer channel until the end of fiscal Q2 (February 2026), with warranty service and support continuing	investors.micron.com	2026-10-09

HDD or SSD? (use, capacity and how long it lasts): quick reference

The calculator in this article shows the result for your own settings. Settings can be set in the URL; each link below opens the calculator with those settings.

Settings	Result	Calculator
OS and apps, 250GB, 20GB	✓ NVMe M.2 SSD, 500GB. A PCIe 4.0 NVMe SSD reads at up to 7,450 MB/s (Samsung 990 PRO 2TB), far past the SATA ceiling of 600 MB/s. Windows 11 needs at least 64 GB, and updates can need more over time. Capacity needed: 250GB of data (as Windows shows it), converted to label units, ÷ (1 − 20% free) = 336GB → 500GB. A 500GB label shows up as about 465 GB when counted in units of 1024³ bytes. A 2TB SSD rated for 1,200 TBW, written at 20 GB a day, reaches its TBW in about 164.3 years. The 5-year warranty runs out before the TBW does (it ends at the years or the TBW, whichever comes first). TBW is the total writing the maker’s warranty covers, not the point where the drive dies. Smaller drives have lower TBW (the Samsung 870 EVO is rated 600 TBW at 1TB and 300 TBW at 500GB). The years above use the 2TB figure, so redo it with the TBW of the size you buy	Open with these settings
PC games, 500GB, 50GB	✓ NVMe M.2 SSD, 1TB. A PCIe 4.0 NVMe SSD reads at up to 7,450 MB/s (Samsung 990 PRO 2TB), far past the SATA ceiling of 600 MB/s. Some big games list an SSD as required on Steam (Starfield, Baldur’s Gate 3, Cyberpunk 2077 and others). Capacity needed: 500GB of data (as Windows shows it), converted to label units, ÷ (1 − 20% free) = 672GB → 1TB. A 1TB label shows up as about 931 GB when counted in units of 1024³ bytes. A 2TB SSD rated for 1,200 TBW, written at 50 GB a day, reaches its TBW in about 65.7 years. The 5-year warranty runs out before the TBW does (it ends at the years or the TBW, whichever comes first). TBW is the total writing the maker’s warranty covers, not the point where the drive dies. Smaller drives have lower TBW (the Samsung 870 EVO is rated 600 TBW at 1TB and 300 TBW at 500GB). The years above use the 2TB figure, so redo it with the TBW of the size you buy	Open with these settings
PC games, 1TB, 200GB	✓ NVMe M.2 SSD, 2TB. A PCIe 4.0 NVMe SSD reads at up to 7,450 MB/s (Samsung 990 PRO 2TB), far past the SATA ceiling of 600 MB/s. Some big games list an SSD as required on Steam (Starfield, Baldur’s Gate 3, Cyberpunk 2077 and others). Capacity needed: 1TB of data (as Windows shows it), converted to label units, ÷ (1 − 20% free) = 1.343TB → 2TB. A 2TB label shows up as about 1862 GB when counted in units of 1024³ bytes. A 2TB SSD rated for 1,200 TBW, written at 200 GB a day, reaches its TBW in about 16.4 years. The 5-year warranty runs out before the TBW does (it ends at the years or the TBW, whichever comes first). TBW is the total writing the maker’s warranty covers, not the point where the drive dies	Open with these settings
PC games, 1.5TB, 50GB	✓ NVMe M.2 SSD, 4TB. A PCIe 4.0 NVMe SSD reads at up to 7,450 MB/s (Samsung 990 PRO 2TB), far past the SATA ceiling of 600 MB/s. Some big games list an SSD as required on Steam (Starfield, Baldur’s Gate 3, Cyberpunk 2077 and others). Capacity needed: 1.5TB of data (as Windows shows it), converted to label units, ÷ (1 − 20% free) = 2.014TB → 4TB. A 4TB label shows up as about 3725 GB when counted in units of 1024³ bytes. A 2TB SSD rated for 1,200 TBW, written at 50 GB a day, reaches its TBW in about 65.7 years. The 5-year warranty runs out before the TBW does (it ends at the years or the TBW, whichever comes first). TBW is the total writing the maker’s warranty covers, not the point where the drive dies	Open with these settings
PS5 expansion, 1TB	✓ PCIe Gen4x4 M.2 SSD with heatsink, 2TB. The PS5 takes a PCIe Gen4x4 M.2 NVMe SSD, with 5,500 MB/s or faster read recommended. It needs a heatsink or similar cooling, and the SSD with its heatsink can be at most 25 mm wide and 11.25 mm high. The PS5 supports 250GB to 8TB. PS5 games moved to USB storage have to be copied back to the console or an M.2 SSD before you can play them. Capacity needed: 1TB of data (as Windows shows it), converted to label units, ÷ (1 − 20% free) = 1.343TB → 2TB. A 2TB label shows up as about 1862 GB when counted in units of 1024³ bytes	Open with these settings
PS5 expansion, 1.5TB	✓ PCIe Gen4x4 M.2 SSD with heatsink, 4TB. The PS5 takes a PCIe Gen4x4 M.2 NVMe SSD, with 5,500 MB/s or faster read recommended. It needs a heatsink or similar cooling, and the SSD with its heatsink can be at most 25 mm wide and 11.25 mm high. The PS5 supports 250GB to 8TB. PS5 games moved to USB storage have to be copied back to the console or an M.2 SSD before you can play them. Capacity needed: 1.5TB of data (as Windows shows it), converted to label units, ÷ (1 − 20% free) = 2.014TB → 4TB. A 4TB label shows up as about 3725 GB when counted in units of 1024³ bytes	Open with these settings
Older PC upgrade, 250GB, 10GB	✓ 2.5-inch SATA SSD, 500GB. First check the PC’s spec sheet for the supported interface (SATA or NVMe/PCIe) and size (2.5-inch, or the M.2 length). If it lists a 2.5-inch SATA drive, a SATA SSD tops out at the SATA ceiling of 600 MB/s, and it works on older SATA 3Gb/s ports too. Capacity needed: 250GB of data (as Windows shows it), converted to label units, ÷ (1 − 20% free) = 336GB → 500GB. A 500GB label shows up as about 465 GB when counted in units of 1024³ bytes. A 2.5-inch SATA Samsung 870 EVO 500GB (300 TBW), written at 10 GB a day, reaches its TBW in about 82.1 years. The 5-year warranty runs out before the TBW does (it ends at the years or the TBW, whichever comes first). TBW is the total writing the maker’s warranty covers, not the point where the drive dies	Open with these settings
Older PC upgrade, 750GB, 50GB	✓ 2.5-inch SATA SSD, 2TB. First check the PC’s spec sheet for the supported interface (SATA or NVMe/PCIe) and size (2.5-inch, or the M.2 length). If it lists a 2.5-inch SATA drive, a SATA SSD tops out at the SATA ceiling of 600 MB/s, and it works on older SATA 3Gb/s ports too. Capacity needed: 750GB of data (as Windows shows it), converted to label units, ÷ (1 − 20% free) = 1.007TB → 2TB. A 2TB label shows up as about 1862 GB when counted in units of 1024³ bytes. A 2.5-inch SATA Samsung 870 EVO 2TB (1,200 TBW), written at 50 GB a day, reaches its TBW in about 65.7 years. The 5-year warranty runs out before the TBW does (it ends at the years or the TBW, whichever comes first). TBW is the total writing the maker’s warranty covers, not the point where the drive dies	Open with these settings
Photos and video, 2TB, 50GB	✓ Hard drive (CMR for a NAS), 4TB. Hard drives suit large piles of photos and video. For a NAS that reads and writes all the time, pick a NAS drive rated for 180 TB/year. In Seagate’s list, the NAS IronWolf line is CMR at every size, while SMR shows up in desktop BarraCuda drives such as the 4TB and 8TB. Capacity needed: 2TB of data (as Windows shows it), converted to label units, ÷ (1 − 20% free) = 2.685TB → 4TB. A 4TB label shows up as about 3725 GB when counted in units of 1024³ bytes. 50 GB read and written a day is 18.25 TB a year. That’s inside the 180 TB/year workload rating of NAS CMR drives (WD Red Plus, IronWolf)	Open with these settings
Photos and video, 2.5TB, 400GB	✓ Hard drive (CMR for a NAS), 4TB. Hard drives suit large piles of photos and video. For a NAS that reads and writes all the time, pick a NAS drive rated for 180 TB/year. In Seagate’s list, the NAS IronWolf line is CMR at every size, while SMR shows up in desktop BarraCuda drives such as the 4TB and 8TB. Capacity needed: 2.5TB of data (as Windows shows it), converted to label units, ÷ (1 − 20% free) = 3.356TB → 4TB. A 4TB label shows up as about 3725 GB when counted in units of 1024³ bytes. 400 GB read and written a day is 146 TB a year. That’s inside the 180 TB/year workload rating of NAS CMR drives (WD Red Plus, IronWolf). It’s over the 55 TB/year rating of the SMR BarraCuda, so pick a NAS drive with a higher workload rating	Open with these settings
Backups, 1TB	✓ External hard drive, 2TB. None of the external hard drives checked state their recording technology. For a NAS that reads and writes all the time, use an internal NAS drive rated for 180 TB/year. Per JEDEC’s requirement (as cited by Dell), an SSD written to its rated limit only has to hold data unpowered for 3 months, stored at up to 40°C. If you leave an SSD sitting, keep a copy on another drive too. Capacity needed: 1TB of data (as Windows shows it), converted to label units, ÷ (1 − 20% free) = 1.343TB → 2TB. A 2TB label shows up as about 1862 GB when counted in units of 1024³ bytes	Open with these settings
Carry around, 500GB	✓ Portable SSD, 1TB. The Samsung T7 Shield is a USB 3.2 Gen 2 portable SSD. It weighs 98 g and passed a 3 m drop in Samsung’s own test. Capacity needed: 500GB of data (as Windows shows it), converted to label units, ÷ (1 − 20% free) = 672GB → 1TB. A 1TB label shows up as about 931 GB when counted in units of 1024³ bytes	Open with these settings

HDMI vs DisplayPort vs USB-C: Which Port Do You Need?

Sat, 17 Apr 2021 00:00:00 GMT

In this calculation, HDMI 2.0 and DisplayPort 1.4 have enough bandwidth for 1080p at 60Hz. For USB-C, first check that your device supports DisplayPort Alt Mode.

For 4K at 144Hz in 10-bit HDR, HDMI 2.1 and DisplayPort 2.1 at UHBR13.5 or UHBR20 cover the full calculated range uncompressed. DisplayPort 1.4 needs DSC (compression that’s visually lossless).

The numbers come from standards bodies’ documents as of October 2026. Color is not subsampled (4:4:4, meaning no color detail is dropped), and no hardware was tested.

Which port fits each resolution and refresh rate

In this calculation, for 4K at 60Hz in 10-bit HDR, HDMI 2.1 and DisplayPort 1.4 have enough bandwidth.

To check your setup:

Look up the video ports on your device and your monitor in their spec sheets
Decide on a resolution and refresh rate. On a console, look in the display menu for a setting such as “4K at 120Hz”
Find your row in the table below, or enter it in the calculator

The table compares each port’s maximum bandwidth. Check which resolutions and refresh rates your device and monitor actually support.

The table has no UHBR10 column. In this calculation (4:4:4 color), DisplayPort 2.1 at UHBR10 carries 4K at 144Hz in 8-bit uncompressed. In 10-bit HDR it depends on the signal: some fit and some don’t.

Assumptions: uncompressed data rates, as a range between PC and TV signal timings, with full 4:4:4 color and HDR counted as 10-bit. HDMI 2.1 at 4K 120Hz and up and 8K follow HDMI.org’s video table

Video format	Video data needed	HDMI 2.0	HDMI 2.1	DisplayPort 1.4	DisplayPort 2.1 UHBR13.5	DisplayPort 2.1 UHBR20
1080p at 60Hz, 8-bit	3.19–3.57 Gbps	✓ Uncompressed	✓ Uncompressed	✓ Uncompressed	✓ Uncompressed	✓ Uncompressed
1440p at 144Hz, 8-bit	14.07–14.2 Gbps	△ Uncompressed (borderline)	✓ Uncompressed	✓ Uncompressed	✓ Uncompressed	✓ Uncompressed
1440p at 240Hz, 10-bit (HDR)	30.78 Gbps	✕ Not enough	✓ Uncompressed	△ Needs DSC	✓ Uncompressed	✓ Uncompressed
4K at 60Hz, 8-bit	12.54–14.26 Gbps	△ Uncompressed (borderline)	✓ Uncompressed	✓ Uncompressed	✓ Uncompressed	✓ Uncompressed
4K at 60Hz, 10-bit (HDR)	15.67–17.82 Gbps	✕ Not enough	✓ Uncompressed	✓ Uncompressed	✓ Uncompressed	✓ Uncompressed
4K at 120Hz, 10-bit (HDR)	32.24–35.64 Gbps	✕ Not enough	✓ Uncompressed	△ Needs DSC	✓ Uncompressed	✓ Uncompressed
4K at 144Hz, 10-bit (HDR)	37.5–39.19 Gbps	✕ Not enough	✓ Uncompressed	△ Needs DSC	✓ Uncompressed	✓ Uncompressed
4K at 240Hz, 10-bit (HDR)	68.56 Gbps	✕ Not enough	△ Needs DSC	△ Needs DSC	△ Needs DSC	✓ Uncompressed
8K at 60Hz, 10-bit (HDR)	62.05–71.28 Gbps	✕ Not enough	△ Needs DSC	△ Needs DSC	△ Needs DSC	✓ Uncompressed

“Uncompressed” means the video goes through as is
“Uncompressed (borderline)” means less than 3% of the port’s capacity is left over
“Needs DSC” means that in this calculation the video fits only if it is compressed with DSC, which is nearly lossless. Both the sending and the receiving device must support it

Check your own setup with the calculator

Pick the ports, resolution, refresh rate and color depth from steps 1 and 2. It shows the verdict and the cable.

Video port and cable check

Pick your device and its port, the monitor's port, and a resolution, refresh rate and color depth to see which port and version carry it, whether it fits uncompressed, and which cable or adapter you need, plus a certified cable or adapter that fits.

Ports and cables for PS5, Xbox and Switch 2

Console	Cable requirement
PS5	Per the PlayStation Blog, the PS5’s included HDMI cable is recommended. For another cable, use an Ultra High Speed one that supports HDMI 2.1
Xbox Series X and S	Per the Xbox website, both the Xbox Series X and Series S come with an Ultra High Speed HDMI cable
Switch 2	Per Nintendo Australia’s support page, connecting a Switch 2 to a TV takes an Ultra High Speed HDMI cable, and one comes with the Switch 2

Pick cables by this table. It is separate from the bandwidth math.

Per the makers’ official pages, the PS5, Xbox Series X|S and Switch 2 all output video over HDMI. The PS5 and Xbox use HDMI 2.1, and the Switch 2’s specs don’t state an HDMI version.

For 4K at 120Hz from a PS5 or Xbox Series X, the monitor also needs an HDMI 2.1 input. An HDMI 2.0 input can’t carry 4K at 120Hz.

The Switch 2’s specs don’t name an HDMI version, so its 4K at 60Hz is counted as HDMI 2.0. It is borderline in 8-bit, and falls short for 10-bit HDR in full 4:4:4 color.

Bandwidth by port and version

What separates the ports is how much video data they can send per second.

Port and version	Max bandwidth	Usable for video	With DSC (before compression)
HDMI 2.0	18 Gbps	14.4 Gbps	None
HDMI 2.1	48 Gbps	42.66 Gbps	127.98 Gbps
DisplayPort 1.4	32.4 Gbps	25.92 Gbps	77.76 Gbps
DisplayPort 2.1 UHBR10	40 Gbps	38.68 Gbps	116.04 Gbps
DisplayPort 2.1 UHBR13.5	54 Gbps	52.22 Gbps	156.66 Gbps
DisplayPort 2.1 UHBR20	80 Gbps	77.37 Gbps	232.11 Gbps

UHBR is the name for DisplayPort 2.1’s faster link speeds, and a bigger number is faster. The DSC column assumes DSC compresses the video to one third, so it shows the uncompressed amount that fits.

Per HDMI.org, every HDMI 2.1 feature is optional, so an “HDMI 2.1” label doesn’t guarantee 4K at 120Hz.

HDMI 2.2 isn’t in the table or the calculator, because compatible devices weren’t researched. It goes up to 96 Gbps, and the Ultra96 cable is the only one that supports all of its features.

Why is USB-C not showing video on a monitor?

Your USB-C port may not support video output. Check in this order:

Per the DisplayPort FAQ, whether a USB-C port outputs video depends on the device; check the specs for “DisplayPort Alt Mode” or a DP logo next to the port
Per VESA, USB-C DP Alt Mode carries up to 80 Gbps when all four lanes carry video, and up to 40 Gbps when USB data shares the cable. If you can’t tell which setup you have, judge by the USB-sharing side. The calculator lets you pick the connection
Per the DisplayPort FAQ, from a USB-C port that supports DP Alt Mode, adapters and cables connect to DisplayPort, HDMI, DVI and VGA inputs. For 4K at 60Hz, check the cable’s supported resolution and refresh rate in its specs
For a hub or dock, check its specs for video output and for the resolutions and refresh rates it supports

USB-C and Thunderbolt are not the same thing. Per Intel, USB Type-C is a connector shape and Thunderbolt 5 is a standard. Thunderbolt 5 uses the USB Type-C port, and computers, accessories and cables all require certification.

Per USB-IF, USB4’s 80 Gbps runs over 80 Gbps-certified cables.

DisplayPort cables and how long they can be

Pick a DisplayPort cable by its logo. Per VESA, DisplayPort cables are told apart by VESA’s DP40 and DP80 logos. DP40 is rated for UHBR10 and is backward compatible with DisplayPort 1.4 speeds.

Cable	Speed it carries	Length guidance
DP40	Up to UHBR10	Needs improvements beyond 2 m
DP54	Up to UHBR13.5	Up to 2 m
DP80	Up to UHBR20	Needs improvements beyond 1 m
DP80LL (active)	Up to UHBR20	Up to 3 m

Can you convert DisplayPort to HDMI?

Check the converter’s direction and its supported resolution and refresh rate first. The converter and both devices must support your target video format.

Per Dell’s support article, DisplayPort to HDMI works through an adapter. HDMI to DisplayPort takes an active converter.

A passive DisplayPort-to-HDMI adapter only works on a DP++ (Dual-Mode) output. Active DisplayPort-to-HDMI adapters work with both regular DP and DP++ outputs. Conversion is one-way; HDMI to DisplayPort needs a converter made for that direction.

Check your PC’s or graphics card’s spec sheet to see whether it supports DP++.

For bandwidth, the smaller side of the chain sets the limit. Going from DisplayPort 1.4 to HDMI 2.1, 4K at 120Hz in 10-bit needs DSC.

Picks by condition

Check your monitor’s input port and the speed you need against the table or the calculator before you pick. One pick per condition where a listing qualifies. The criteria:

The maker’s official page states the certification or specs. VESA’s certified-product database did not confirm a registration for the lengths picked
Sold new, with at least 100 ratings and 4 stars or more. The 100 is judged on the count pooled across lengths. For DP80 the pooling is confirmed. Whether the DP40 count is pooled is not confirmed. The USB-C to HDMI count is likely pooled across lengths too
Not a marked-up third-party import

The picks come from the Amazon.com listings compared, and only when they were sold by the maker and shipped by Amazon. Check prices on the store.

For DisplayPort 1.4 or UHBR10

It is 2 m (6.6 ft) long.

Sold by Cable Matters (the maker). Ships from Amazon.

For UHBR13.5 or faster

A passive UHBR20 cable needs improvements beyond 1 m, so the pick is 1 m (3.3 ft).

USB-C to HDMI for 4K at 60Hz

The maker lists 4K at 60Hz or higher, and it is 1.8 m (6 ft) long. It is one-way, USB-C to HDMI, and needs a USB-C port with DisplayPort Alt Mode.

FAQ

Should I use HDMI or DisplayPort?

For a game console or TV, use HDMI. For a PC monitor with a DisplayPort input, use whichever port the “Which port fits each resolution and refresh rate” table says is enough.

In this calculation, HDMI 2.0 and DisplayPort 1.4 have enough bandwidth for 1080p at 60Hz. For USB-C, first check that your device supports DisplayPort Alt Mode.

This assumes both your device and your monitor have that port.

Can you convert DisplayPort to HDMI, or HDMI to DisplayPort?

Per Dell’s support article, DisplayPort to HDMI works through an adapter. HDMI to DisplayPort takes an active converter.

A passive DisplayPort-to-HDMI adapter only works on a DP++ (Dual-Mode) output. Active DisplayPort-to-HDMI adapters work with both regular DP and DP++ outputs.

Conversion is one-way; HDMI to DisplayPort needs a converter made for that direction.

Which port do I need for 4K at 240Hz with 10-bit HDR?

In this calculation (4:4:4 color), among the ports compared here, only DisplayPort 2.1 UHBR20 carries 4K at 240Hz in 10-bit HDR uncompressed. HDMI 2.1 and DisplayPort 1.4 need DSC.

Data and formulas used (for verification)

How this was checked: Worked out from the standards bodies’ and makers’ official pages as of October 2026. No hardware was tested. Standards bodies: HDMI.org HDMI 2.2, VESA DisplayPort 2.1, VESA DisplayPort FAQ, USB-IF USB4. Makers: Intel Thunderbolt 5, PlayStation Blog PS5 FAQ, Nintendo Switch 2 tech specs. Adapters: Dell’s adapter guide.

Method

Video data rate = pixel clock × bits per color × 3 ÷ 1000, in Gbps, uncompressed. Color is not subsampled (4:4:4) and HDR is counted as 10-bit. The pixel clock is a range: the VESA CVT-RB2 formula (minimum blanking used by PCs) against CTA-861 (TV and console timing) and Dell’s published timings. Shown to two decimals, the low end rounded down and the high end rounded up.
Usable video capacity per port: HDMI 2.0 is 18 Gbps × 8/10 (TMDS) = 14.4 Gbps, HDMI 2.1 is 48 Gbps × 16/18 (FRL) = 42.66 Gbps, DisplayPort 1.4 uses VESA’s maximum payload 25.92 Gbps, and DisplayPort 2.1 UHBR20 is 77.37 Gbps. UHBR10 and UHBR13.5 scale the UHBR20 figure by per-lane rate (38.68 Gbps, 52.22 Gbps). Rounded down to two decimals.
DSC is counted as compressing up to one third on DisplayPort 1.4 and 2.1 and HDMI 2.1. HDMI 2.0 has no DSC. DSC needs support on both the sending and receiving device. HDMI 2.1 at 4K 120Hz and up, and 8K, follow HDMI.org’s video table. HDMI 2.2 is left out because compatible devices were not checked.
Verdict: if the high end fits, uncompressed. With less than 3% to spare, borderline. If only the low end fits, it depends on the signal. If DSC makes it fit, DSC is needed. The Switch 2 states no HDMI version, so it is judged at HDMI 2.0 capacity. Cable requirements (such as a console’s included cable) are given separately from the bandwidth math, from each maker’s official page.
USB-C is judged by the DisplayPort version in its specs and by the connection. Per VESA, DP Alt Mode carries up to 80 Gbps when all four lanes carry video and up to 40 Gbps (two lanes) when USB data is sent at the same time. The four-lane case uses the version’s usable video capacity as is; the two-lane case with USB data uses half of it. When the connection is unknown, it is judged on the 40 Gbps (two-lane) side: if that fits, it works either way. If the 40 Gbps side falls short but four lanes fit, no firm check mark is shown and the result is ‘depends on the connection’. For USB-C to HDMI, the connection applies only to the DisplayPort leg.
Color-depth check: 4K at 240Hz in 12-bit needs up to 82.27 Gbps at the high end, above UHBR20’s usable video capacity of 77.37 Gbps. The article’s claim that only UHBR20 carries 4K at 240Hz uncompressed covers up to 10-bit (HDR).
Cable length uses only VESA’s statements: the lengths beyond which passive cables need improvements (DP40 beyond 2 m, DP80 beyond 1 m), DP80LL up to 3 m, and DP54 up to 2 m. No hardware was tested. Picks come from products whose maker pages state the certification and specs, among new Amazon listings with 100+ ratings (judged on the count pooled across lengths), 4+ stars, no Sakura Checker warning and no lower-than-Amazon score, and no import, the cheapest per condition including shipping.

Official values used

Item	Value	Source	Checked
HDMI 2.1 / Ultra High Speed cable bandwidth	48 Gbps	hdmi.org	2026-10-09
Video formats targeted by Ultra High Speed cable	8K@60 and 4K@120	hdmi.org Ultra High Speed: 8K@60 and 4K@120 formats	2026-10-09
HDMI 2.2 / Ultra96 cable bandwidth	96 Gbps	hdmi.org Up to 96Gbps bandwidth and next-gen HDMI Fixed Rate Link technology; The new Ult	2026-10-09
HDMI 2.2 supported formats	12K@120 and 16K@60; uncompressed 8K@60/4:4:4 and 4K@240/4:4:4 at 10-bit and 12-bit	hdmi.org Supported resolutions include 4K240 and 8K60, both uncompressed and with full ch	2026-10-09
Cable needed for all HDMI 2.2 applications	Ultra96 HDMI Cable is the only cable that supports all HDMI 2.2 applications	hdmi.org The new Ultra96 HDMI Cable supports up to 96Gbps bandwidth and is the only cable	2026-10-09
Ultra96 feature name (devices)	Ultra96 feature name: 64Gbps, 80Gbps or 96Gbps	hdmi.org “Ultra96” is a feature name that manufacturers are encouraged to use to indicate	2026-10-09
Ultra cable certification being mandatory	mandatory certification program (Ultra96 and Ultra High Speed)	hdmi.org Both the Ultra96 HDMI Cable and Ultra High Speed HDMI Cable are part of the Ultr	2026-10-09
HDMI 2.1 features being optional	optional	hdmi.org	2026-10-09
DisplayPort 1.4a maximum link bandwidth	32.4 Gbps	vesa.org The previous version of DisplayPort, v1.4a, provided a maximum link bandwidth of	2026-10-09
DisplayPort 1.4a maximum payload	25.92 Gbps	vesa.org With 8b/10b channel coding, that equates to a maximum payload of 25.92 Gbps.	2026-10-09
DisplayPort 2.1 maximum bandwidth (UHBR20, four lanes)	80 Gbps	vesa.org VESA certified DP80 cables support up to the UHBR20 link rate (20 Gbps), with fo	2026-10-09
DisplayPort 2.1b maximum payload	77.37 Gbps	displayport.org DisplayPort 2.1b … maximum link rate to up to 20 Gbps/lane and features more e	2026-10-09
DP40 cable definition	UHBR10, 40Gbps (DP40)	vesa.org VESA certified DP40 cables support up to the UHBR10 link rate (10 Gbps), with fo	2026-10-09
DisplayPort cable certification logo	DP40 / DP80 logo from VESA	vesa.org if your cable has the DP40 or DP80 logo from VESA, it will meet the specs for th	2026-10-09
DP40 / DP80 cable backward compatibility	DP40 and DP80 cables are backward compatible with RBR, HBR, HBR2 and HBR3	vesa.org fully backward compatible with devices supporting DisplayPort link rates previou	2026-10-09
DP40 / DP80 cable length guidance	DP40: improvements needed beyond 2 m; DP80 passive: beyond 1 m; DP80LL active: up to 3 m	vesa.org improved connectivity and longer cable lengths (beyond two meters for DP40 cable	2026-10-09
DSC support required in DisplayPort 2.1	DisplayPort 2.1 mandates DSC support; can reduce transport bandwidth in excess of 67 percent	vesa.org	2026-10-09
DisplayPort 1.4 DSC (up to 3:1)	DSC 1.2 up to 3:1, visually lossless	vesa.org	2026-10-09
DisplayPort 2.1 uncompressed configurations	8K 60Hz 4:4:4 with 30 bits per pixel (HDR-10); two 4K 144Hz 24bpp uncompressed	displayport.org First standard to support 8K resolution (7680 x 4320) at 60 Hz refresh rate with	2026-10-09
How to tell whether a USB-C port outputs video (DP Alt Mode)	check device documentation or the DP logo; USB Type-C connectors that support DisplayPort over USB-C normally carry the DP logo	displayport.org How do I know if a device supports DisplayPort Alt Mode? This information should	2026-10-09
USB-C DP Alt Mode bandwidth	up to 80Gbps of DisplayPort video using all four lanes, or up to 40Gbps with simultaneous SuperSpeed USB data	vesa.org	2026-10-09
USB4 Version 2 maximum bandwidth	80 Gbps	usb.org	2026-10-09
USB4 compatibility	compatible with USB 3.2, USB 2.0 and Thunderbolt 3 hosts and devices	usb.org Compatibility with existing USB 3.2, USB 2.0, and Thunderbolt 3 hosts and device	2026-10-09
Thunderbolt 5 bandwidth (each direction)	80 Gbps	download.intel.com Thunderbolt 5 will deliver 80 gigabits per second (Gbps) of bi-directional bandw	2026-10-09
Thunderbolt 5 Bandwidth Boost (transmit)	120 Gbps	download.intel.com with Bandwidth Boost it will provide up to 120 Gbps for the best display experie	2026-10-09
Thunderbolt 5 display configurations (Intel)	dual 8K (60Hz) or three 4K (144Hz)	thunderbolttechnology.net	2026-10-09
Thunderbolt 5 foundations and compatibility	built on USB4 V2, DisplayPort 2.1 and PCI Express Gen 4; fully compatible with previous versions	download.intel.com Built on industry standards including USB4 V2, DisplayPort 2.1 and PCI Express G	2026-10-09
Thunderbolt 5 passive cable length (Intel announcement)	passive cables up to 1 meter	download.intel.com Utilizes a new signaling technology, PAM-3, to deliver these significant increas	2026-10-09
Thunderbolt certification being mandatory	mandatory certification for all shipping computers, accessories, and cables	thunderbolttechnology.net Maximized Performance: Mandatory certification for all shipping computers, acces	2026-10-09
Thunderbolt port shape (USB Type-C)	One universal computer port (USB Type-C)	thunderbolttechnology.net Thunderbolt 5 Leaps Ahead: One universal computer port (USB Type-C)	2026-10-09
PS5 HDMI 2.1 and 4K 120Hz	HDMI 2.1, 4K 120Hz	blog.playstation.com	2026-10-09
PS5 HDMI cable recommendation	included cable recommended; otherwise Ultra High Speed	blog.playstation.com	2026-10-09
PS5 Pro 8K output condition	PS5 Pro 8K output requires a screen supporting 8K at 60Hz with DSC	playstation.com PS5 Pro console 8K output requires a screen that supports 8K at 60Hz with Displa	2026-10-09
Xbox Series X HDMI 2.1 and included cable	1x HDMI 2.1 port; up to 120 FPS; Ultra High Speed HDMI cable included	xbox.com	2026-10-09
Xbox Series S HDMI 2.1 and included cable	1x HDMI 2.1 port; 1440p up to 120 FPS; Ultra High Speed HDMI cable included	xbox.com	2026-10-09
Switch 2 TV output	HDMI output in TV mode; max 4K at 60 fps; 120 fps at 1080p/1440p; HDR10; VRR up to 120Hz	nintendo.com	2026-10-09
HDMI cable needed for Switch 2 TV mode	Ultra High Speed HDMI cable required (meets HDMI 2.1 specifications); one is included with Switch 2	nintendo.com	2026-10-09
When a passive DisplayPort-to-HDMI adapter works	passive DisplayPort-to-HDMI adapters only work on DP++ (Dual-Mode) ports; active adapters work with a pure DisplayPort or DP++ output	dell.com	2026-10-09
HDMI-to-DisplayPort conversion and direction	HDMI to DisplayPort always requires an active adapter; an adapter designed for DisplayPort to HDMI cannot work in reverse	dell.com	2026-10-09
USB-C DP Alt Mode adapters to other video ports	adapters and adapter cables defined in the DisplayPort Alt Mode for USB Type-C Standard connect DisplayPort over USB-C to DisplayPort, HDMI, DVI and VGA monitor inputs	displayport.org Adapters and Adapter cables, as defined in the DisplayPort Alt Mode for USB Type	2026-10-09
HDMI 2.0 maximum bandwidth	18 Gbps	hdmi.org	2026-10-09
HDMI 2.0 encoding efficiency (8b/10b)	0.8	xilinx.com AMD/Xilinx PG351: TMDS up to 6 Gb/s @ 3 lanes	2026-10-07
HDMI 2.1 encoding efficiency (16b/18b)	0.8888888888888888	xilinx.com AMD/Xilinx PG351: FRL 12 Gb/s @ 4 lanes, decode the incoming 16-bit/18-bit encod	2026-10-07
DSC maximum compression ratio	3	vesa.org	2026-10-09
CVT-RB2 horizontal blanking	80 px	github.com CVT_RB_V2_H_BLANK 80 pixels	2026-10-07
CVT-RB2 vertical blanking time	460 µs	github.com CVT_RB_MIN_V_BLANK 460 us	2026-10-07
CVT-RB2 vertical sync lines	8 lines	github.com RB v2 is detected when vsync == 8	2026-10-07
CVT-RB2 minimum back porch lines	6 lines	github.com CVT_RB_V_BPORCH 6	2026-10-07
CVT-RB2 minimum front porch lines	1 lines	github.com CVT_RB_V2_MIN_V_FPORCH 1	2026-10-07
CTA-861 1080p60 pixel clock	148.5 MHz	github.com edid_cea_modes: /* 16 - 1920x1080@60Hz */ 148500 kHz	2026-10-08
CTA-861 1080p120 pixel clock	297 MHz	github.com edid_cea_modes: /* 63 - 1920x1080@120Hz */ 297000 kHz	2026-10-08
CTA-861 4K60 pixel clock	594 MHz	github.com edid_cea_modes: /* 97 - 3840x2160@60Hz */ 594000 kHz	2026-10-08
CTA-861 4K120 pixel clock	1188 MHz	github.com edid_cea_modes: /* 118 - 3840x2160@120Hz */ 1188000 kHz	2026-10-08
CTA-861 8K60 pixel clock	2376 MHz	github.com edid_cea_modes: /* 199 - 7680x4320@60Hz */ 2376000 kHz	2026-10-08
Dell 4K60 pixel clock (CVT)	533.25 MHz	dl.dell.com UHD 3840x2160 (DP) 60Hz 533.25 MHz PDF p.16	2026-10-07
Dell 4K120 pixel clock (low)	1074.73 MHz	dl.dell.com VESA 3840x2160 119.88Hz 1074.73 MHz PDF p.14	2026-10-07
Dell 4K144 pixel clock	1250 MHz	dl.dell.com VESA 3840x2160 144Hz 1250 MHz PDF p.13-14	2026-10-07
Dell 1440p144 pixel clock	591.31 MHz	dl.dell.com 2560x1440 144Hz 591.31 MHz PDF p.14	2026-10-07
HDMI.org video table (cable per format)	HDMI Video Table: 4K100/120 all formats = UHS; 4K144 12-bit RGB/4:4:4 = UHS+DSC or Ultra96, others = UHS; 4K200/240 and 8K48/50/60 RGB/4:4:4/4:2:2 = UHS+DSC or Ultra96, 4:2:0 = UHS	hdmi.org HDMI Video Table - All Formats. BLACK = Support with UHS HDMI Cables or Ultra96	2026-10-08
Cable required for higher formats	Ultra High Speed or Ultra96 required	hdmi.org Will some resolutions and refresh rates require a specific cable? In order to en	2026-10-08
HDMI compression (DSC) support	uncompressed and compressed (DSC 1.2a) modes; manufacturers can implement either or both	hdmi.org The specification supports both uncompressed and compressed modes. Manufacturers	2026-10-08
UHBR10 per-lane rate	10 Gbps	vesa.org	2026-10-09
UHBR13.5 per-lane rate	13.5 Gbps	vesa.org	2026-10-09
DP54 cable (for UHBR13.5)	DP54: UHBR10 and UHBR13.5 up to 2 m passive; UHBR13.5 previously only through a DP80 cable	vesa.org	2026-10-09

Video port and cable check: quick reference

The calculator in this article shows the result for your own settings. Settings can be set in the URL; each link below opens the calculator with those settings.

Settings	Result	Calculator
DisplayPort 1.4, DisplayPort 1.4, 4K, 144Hz, HDR (10-bit)	△ Works with DSC compression. DisplayPort 1.4 carries 25.92 Gbps as video. DSC only works when both the source and the monitor support it	Open with these settings
HDMI 2.1, HDMI 2.1, 4K, 144Hz, HDR (10-bit)	✓ Fits uncompressed. HDMI 2.1 carries 42.66 Gbps as video. Every HDMI 2.1 feature is optional, so check that the monitor takes this resolution and refresh rate over HDMI	Open with these settings
DisplayPort 2.1 (UHBR20), DisplayPort 2.1 (UHBR20), 4K, 240Hz, HDR (10-bit)	✓ Fits uncompressed. DisplayPort 2.1 (UHBR20) carries 77.37 Gbps as video. UHBR20 over a passive DP80 cable has been limited to 1 m. The active DP80LL goes to 3 m (VESA)	Open with these settings
DisplayPort 2.1 (UHBR13.5), DisplayPort 2.1 (UHBR13.5), 4K, 240Hz, HDR (10-bit)	△ Works with DSC compression. DisplayPort 2.1 (UHBR13.5) carries 52.22 Gbps as video. DSC only works when both the source and the monitor support it. For UHBR13.5, a passive DP54 cable works up to 2 m (VESA)	Open with these settings
DisplayPort 1.4, DisplayPort 1.4, 1440p, 165Hz, SDR (8-bit)	✓ Fits uncompressed. DisplayPort 1.4 carries 25.92 Gbps as video	Open with these settings
DisplayPort 1.4, HDMI 2.1, 4K, 120Hz, HDR (10-bit)	△ Works with DSC compression. DisplayPort 1.4 carries 25.92 Gbps as video. HDMI 2.1 carries 42.66 Gbps as video. DSC only works when both the source and the monitor support it. On a DP++ output a passive adapter works too. An active adapter works with or without DP++ (per Dell). The HDMI side needs HDMI 2.1 bandwidth, so pick an adapter or cable with HDMI 2.1 output. Every HDMI 2.1 feature is optional, so check that the monitor takes this resolution and refresh rate over HDMI	Open with these settings
HDMI 2.1, DisplayPort 1.4, 1080p, 60Hz, SDR (8-bit)	△ Works if the adapter supports the format. HDMI 2.1 carries 42.66 Gbps as video. DisplayPort 1.4 carries 25.92 Gbps as video. HDMI to DisplayPort takes an active adapter, and a DisplayPort-to-HDMI adapter won’t work in reverse (per Dell). Check the adapter’s specs for the formats it supports. Every HDMI 2.1 feature is optional, so check that the monitor takes this resolution and refresh rate over HDMI	Open with these settings
HDMI 2.1, HDMI 2.1, 8K, 60Hz, HDR (10-bit)	△ Works with DSC compression. HDMI 2.1 carries 42.66 Gbps as video. DSC only works when both the source and the monitor support it. Every HDMI 2.1 feature is optional, so check that the monitor takes this resolution and refresh rate over HDMI	Open with these settings
USB-C (DP Alt Mode, DisplayPort 1.4), Not sure, DisplayPort 1.4, 4K, 60Hz, HDR (10-bit)	△ Depends on the connection. DisplayPort 1.4 carries 25.92 Gbps as video. DSC only works when both the source and the monitor support it. Check the device’s specs for “DisplayPort Alt Mode”, or a DP logo next to the port, to see if that USB-C outputs video. Since the connection is unknown, this can’t be confirmed. With all four lanes for video: ✓ Fits uncompressed. With two lanes shared with USB data: △ Works with DSC compression	Open with these settings
USB-C (DP Alt Mode, DisplayPort 1.4), Not sure, HDMI 2.0, 4K, 60Hz, SDR (8-bit)	△ Depends on the connection. DisplayPort 1.4 carries 25.92 Gbps as video. HDMI 2.0 carries 14.4 Gbps as video. HDMI 2.0 has 1% to spare. It fits with PC timings (VESA CVT-RB2) but not with TV timings. The HDMI side fits within HDMI 2.0’s capacity. Check the device’s specs for “DisplayPort Alt Mode”, or a DP logo next to the port, to see if that USB-C outputs video. Since the connection is unknown, this can’t be confirmed. With all four lanes for video: △ Just fits uncompressed. With two lanes shared with USB data: △ Depends on the signal	Open with these settings
USB-C (DP Alt Mode, DisplayPort 2.1 UHBR20), Not sure, DisplayPort 2.1 (UHBR20), 4K, 240Hz, HDR (10-bit)	△ Depends on the connection. DisplayPort 2.1 (UHBR20) carries 77.37 Gbps as video. DSC only works when both the source and the monitor support it. Check the device’s specs for “DisplayPort Alt Mode”, or a DP logo next to the port, to see if that USB-C outputs video. Since the connection is unknown, this can’t be confirmed. With all four lanes for video: ✓ Fits uncompressed. With two lanes shared with USB data: △ Works with DSC compression. UHBR20 over a passive DP80 cable has been limited to 1 m. The active DP80LL goes to 3 m (VESA)	Open with these settings
USB-C (DP Alt Mode, DisplayPort 2.1 UHBR20), All four lanes carry video (up to 80 Gbps), DisplayPort 2.1 (UHBR20), 4K, 240Hz, HDR (10-bit)	✓ Fits uncompressed. DisplayPort 2.1 (UHBR20) carries 77.37 Gbps as video. Check the device’s specs for “DisplayPort Alt Mode”, or a DP logo next to the port, to see if that USB-C outputs video. Calculated as a connection where all four lanes carry video (up to 80 Gbps). UHBR20 over a passive DP80 cable has been limited to 1 m. The active DP80LL goes to 3 m (VESA)	Open with these settings
HDMI 2.0, HDMI 2.0, 4K, 60Hz, SDR (8-bit)	△ Just fits uncompressed. HDMI 2.0 carries 14.4 Gbps as video. HDMI 2.0 has 1% to spare	Open with these settings
PS5 (4K at 120Hz), HDMI 2.1, HDR (10-bit)	✓ Fits uncompressed. HDMI 2.1 carries 42.66 Gbps as video. Every HDMI 2.1 feature is optional, so check that the monitor takes this resolution and refresh rate over HDMI	Open with these settings
PS5 (4K at 120Hz), HDMI 2.0, HDR (10-bit)	✕ This port and version fall short (by this math). HDMI 2.0 carries 14.4 Gbps as video. With 4:2:0 chroma subsampling: ✕ This port and version fall short (by this math)	Open with these settings
Switch 2 (4K at 60Hz), HDMI 2.0, SDR (8-bit)	△ Just fits uncompressed. HDMI 2.0 carries 14.4 Gbps as video. HDMI 2.0 has 1% to spare. Nintendo’s specs don’t state an HDMI version, so this uses HDMI 2.0’s capacity	Open with these settings
Xbox Series S (1440p at 120Hz), HDMI 2.1, HDR (10-bit)	✓ Fits uncompressed. HDMI 2.1 carries 42.66 Gbps as video. Every HDMI 2.1 feature is optional, so check that the monitor takes this resolution and refresh rate over HDMI	Open with these settings

Peak Design Capture V3 and ProPad review — the clip that locks a DSLR onto your hip belt

Fri, 16 Apr 2021 00:00:00 GMT

2026 update: Ported from the old VuePress blog. The Peak Design Capture is still in the current line-up in 2026, with successive minor revisions of the V3 generation on sale. The review angles here — comfort on a hip belt, lock robustness, the metallic click from plate tolerance — still apply.

Take a DSLR out for the day and every shot has the same tax: dig the camera out of the bag, miss the moment, repeat. Lay it flat inside the pack and by the time it is out, whatever you wanted to shoot has already moved on.

To deal with this at the physical layer, I bought the Peak Design Capture V3 (CC-BK-3) and ProPad (PP-2) and have lived with them for a few years. The flow becomes: clip the camera directly onto a hip belt or backpack shoulder strap, lift it off one-handed, raise it to the eye.

The verdict — worth the money if you raise a DSLR often

Short answer: if you want out of strap-based carry, buy it. The lock is solid, the camera does not come loose while running, and after several years the button has never broken.

The reason is that the number of actions before the camera is at your eye drops sharply. Three actions — put the bag down, open the zip, lift the camera out — collapse to one: press the button and lift. With moving subjects that will not wait — kids, animals at a zoo, sports day — that difference is the gap between getting the shot and not.

A few cautions before buying:

Hands-free types who only shoot occasionally: a strap for a few thousand yen is enough
Full-frame + f/2.8 trinity users: hanging that much weight off the hip swings while you walk. Routing the load through the shoulder is more realistic
People shooting in environments that need to be quiet: tolerance between the plate and the clamp can produce a metallic click (more below)

Look and feel

Short answer: machined aluminium, single matte black, edges rounded but the surface feels hard. The build quality sits above most camera accessories.

The front is plain. A lock-release button sits in the centre, with hex-socket clamp screws to either side.

Spec	Capture V3
Size	about 13 × 13 × 3 cm
Weight	about 144g (rated)
Material	machined aluminium
Strap thickness	up to ~6.4mm (rated)

The back is flat. This is the face that meets the belt, and it is designed to have nothing catching. On a hip belt the surface ends up close to your skin and hip bone, so the cushion (ProPad) earns its place.

From the side the depth is shallow. On a hip belt it is small enough that the hem of a jacket can cover it.

The underside is the female slot for the plate. The trapezoidal quick-release plate on the camera slides into here from above.

ProPad PP-2

The cushion that sits between the Capture and a hip belt to spread the load.

Spec	ProPad PP-2
Size	about 17 × 9 × 3 cm
Weight	about 36g (rated)
Material	cushion + hook-and-loop

The cushion is not overly thick, and once the belt is tightened it sits flush against the hip. Even when I worked up a sweat the hook-and-loop did not loosen.

What worked

Short answer: the motion to bring the camera to your eye is at its shortest / the lock is genuinely robust / nothing has broken over several years.

One-handed, one-action to ready

The biggest single win. Hold the lock-release button while you pull the camera up — done.

Unlike pulling it out of a bag, by the time it is up it is already in shooting position. With kids or animals that move unpredictably, the two or three seconds saved decide whether the shot exists. The effect has been largest for travel, sports day, the zoo, and hiking.

The lock does not open unless you mean it

If it is one-handed in one motion, the obvious worry is “will it pop open while I am walking?” Over several years of use, that has never happened.

The lock geometry will not allow vertical removal without the button being pressed. Run, jump — as long as nothing touches the button, the camera stays put. Press it intentionally and it slides out cleanly. The on/off distinction is unambiguous, with no nagging worry about accidental release.

The button is built into the body, so it cannot be lost

A generic camera clip I had used before put the lock-release button on a removable cap. Within six months it had gone missing and the clip was unusable.

On the Peak Design side the button is fully integrated into the body — you cannot pop it off even if you try. Serviceability suffers in exchange, but the field-loss risk drops to zero, and in real use that matters.

Arca-Swiss plates work too

The standard Peak Design plate has an Arca-Swiss-compatible cross section, so if your tripod head already takes Arca-Swiss clamps, the same plate goes straight onto the tripod with no swapping.

Third-party Arca-Swiss plates have also worked in the slot. If you are buying a new kit it is safest to stay all Peak Design, but if you have already standardised your tripod side on Arca, the cost of moving over is low.

What didn’t

Short answer: third-party plates can rattle metallically / heavy rigs swing on a hip belt / the standard plate slides on its own bottom face.

Tolerance against third-party plates causes a metallic rattle

The plate slot is built with some clearance, presumably so that it will accept a range of compatible plates.

With the genuine Peak Design plate there is almost no play, but a third-party Arca-Swiss plate can produce a light metallic click-clack, more obvious the faster you walk. For environments where silence matters — bird photography, say — the realistic options are to standardise on Peak Design plates, or to take up the slack with a thin strip of PTFE tape.

Heavy lenses swing on a hip belt

Mount a full-frame body with a 70–200mm f/2.8-class lens on the hip and it bounces up and down with every step. This is not really a problem with Capture; it is the physical limit of hanging that much weight off a hip belt at all.

In practice, for rigs that heavy, move the Capture onto a backpack shoulder strap, or pair it with a regular strap to share the load. Hip-belt use stays comfortable up to about mirrorless plus a prime or a standard zoom.

The standard plate slides on a tripod head

The base of the standard plate is flat, with no cork or rubber anti-slip layer (may vary by unit).

If you set the camera down on its bottom directly, or onto a slick tripod head, there is a slight slip. Clamped down it is fine, but a casual drop onto a desk can start sliding, so worth knowing.

How to mount and use it

Short answer: open the clamp, thread the belt, tighten — 30 seconds. The plate stays on the tripod thread once you put it on and basically never comes off.

First, loosen the two clamp screws with a hex key. You can work from either side. With the screws loose, the upper half rotates open and clears a path for the belt.

For hip-belt use, open the ProPad and sandwich the Capture in along the printed guide on the inside. Thread the whole sandwich onto the belt.

Close the ProPad, swing the Capture’s upper half back into place, and retighten the screws. Done — 30 seconds once you know the moves.

Fitting the camera plate

On the camera side, screw the quick-release plate into the tripod thread with the 1/4-inch screw. Once it is on, it basically stays on.

Seen from the side, the base is trapezoidal (Arca-Swiss compatible). That dovetail slides into the matching female profile inside the Capture.

Press the black cylindrical button on the side of the Capture, then drop the plate in from above.

Push it all the way down, release the button, and the lock engages. Removing it uses the same button in reverse.

A GoPro will also fit, given an adapter mount with a 1/4-inch thread. In practice, though, the main use case is mirrorless or DSLR carry on a hip belt, not action cams.

Compared with camera belts and sling harnesses

Short answer: Capture locks the camera flush to your body, a harness slings it off your shoulder for a fast lift, a neck strap just hangs it from your neck — they are answers to different questions.

Angle	Peak Design Capture V3	Sling harness (general)	Neck strap (general)
Mount position	Hip belt or shoulder strap	Shoulder / under arm	Neck
Bring-up motion	Press button, lift out	Lift the dangling camera	Lift the dangling camera
Hands fully free?	Yes — locked in place	No — still swings	No — still swings
Stability while running or jumping	High (no swing)	Medium (strap rebound)	Low (camera flails)
Heavy lenses	Swings on a hip belt	Good — shoulder absorbs it	Average
Noise	Possible plate-tolerance click	Quiet	Quiet
Price band (at the time of writing)	~18,000–20,000 yen	~5,000–15,000 yen	~2,000–5,000 yen

“I want the absolute shortest path to my eye” and “I want both hands genuinely free” both point at Capture. “I want to walk with the camera hanging off me all day” leans toward a harness, which is less tiring over long stretches.

For an adjacent piece in the same camera-accessory category, I also wrote up the JOBY GorillaPod JB01542-PKK, which is a small flexible tripod for wrapping around a pillar or shelf. That sits on the desk-or-pillar layer rather than the body-mounted layer, so it is a separate choice from Capture.

FAQ

Q. What cameras and packs does it fit? A. The clamp is rated for straps and belts up to about 6.4mm thick. That covers the shoulder straps on most camera backpacks and hip belts up to roughly 5cm wide. On the camera side, anything with a 1/4-inch tripod thread will work — mirrorless, DSLR, or compact. Heavy rigs (a full-frame body with a telephoto, for example) put real load on the shoulder or hip, so pairing with a shoulder strap is the practical call.

Q. Is it better than a harness or a sling strap? A. They solve different problems. A harness or sling strap is for “hang it off the shoulder and bring it up fast”; the camera always dangles from your body. Capture is for “lock it flush to your body” and is strong when you want both hands fully free — hiking, shooting with kids. The way the weight is routed is also different: a harness rides on the shoulder, Capture distributes load through the hip (via the ProPad). People who shoot a lot on a tripod, or who swap lenses often, frequently end up using both.

Q. Do I actually need the ProPad? A. For hip-belt use, effectively yes. With Capture alone, the back of the clamp digs straight into the belt, and after a long walk it presses on the hip bone and hurts. ProPad sandwiches the belt in cushioning so the load spreads out around the waist. For backpack-shoulder use the cushion is not needed and Capture on its own is enough. The rule of thumb is “ProPad for hip, Capture alone for shoulder.”

Q. Does it hold up over years of use? A. After several years of daily use, the lock mechanism has never failed. The clamp screws, the lock button, the aluminium body — none of them show meaningful wear. By way of comparison, a generic clip from another brand I had used before had a removable lock button, and within six months the button was lost or broken. On the Peak Design side the button is built into the body and cannot be removed, so the loss risk is structurally zero.

Verdict — if you raise a DSLR often, the day-to-day changes

It is the kind of accessory that picks its users, but for the right use case it is hard to go back to strap-only carry.

It is an easy recommendation for kids, travel, the zoo, hiking — anywhere you want both hands free with a DSLR ready to come up in one motion. Conversely, if the camera only comes out a few times a month, a neck strap for a few thousand yen will do the job.

The two cautions are the metallic click from plate tolerance and hip-belt swing with heavy lenses. The first is fixed by standardising on Peak Design plates; the second by moving over to a shoulder-strap mount.

Vue.js basics — v-if and v-for

Thu, 15 Apr 2021 00:00:00 GMT

2026 update: this post was first written for Vue 2 / Options API. In 2026, Vue 3 with the Composition API (<script setup>) is the default, but the template syntax for v-if / v-else / v-for hasn’t changed, so the shape of the examples below still maps over.

A short reference for the Vue.js directives that handle conditional rendering and iteration.

The mustache syntax and v-bind are covered in a separate post.

The short answer

v-if: renders the element only when the condition is truthy. Pairs with v-else / v-else-if
v-show: toggles visibility through CSS display. Suited to elements that flip on and off often
v-for: iterates over an array. Always pass :key

v-if

v-if renders an element only when the expression is truthy.

It’s common in real work — UI that needs to differ between mobile and desktop, or between logged-in and logged-out states, is the typical case.

Basic usage

v-if is written as an attribute on the target element, with the condition as its value.

The snippet below decides inside the template whether the block renders.

In practice this displays This is a sample.

<template>
  <div v-if="sampleNum === 1">
    サンプルです。
  </div>
</template>

<script>
export default {
  data() {
    return {
      sampleNum: 1
    };
  },
};
</script>

v-else

v-if accepts an else branch in the same way a regular if statement does.

Just add v-else as an attribute on the next element.

This snippet displays else is shown.

<template>
  <div>
    <div v-if="sampleNum === 0">
      サンプルです。
    </div>
    <div v-else>
      elseが表示です。
    </div>
  </div>
</template>

<script>
export default {
  data() {
    return {
      sampleNum: 1
    };
  },
};
</script>

One gotcha: v-else has to sit on the element directly after the v-if. Anything in between breaks the pairing.

v-else-if

else-if works alongside else.

The form is the same — add v-else-if as the attribute.

This snippet displays sampleNum is 1.

<template>
  <div>
    <div v-if="sampleNum === 0">
      サンプルです。
    </div>
    <div v-else-if="sampleNum === 1">
      sampleNumは1です。
    </div>
  </div>
</template>

<script>
export default {
  data() {
    return {
      sampleNum: 1
    };
  },
};
</script>

v-show

v-show looks similar to v-if from the outside, but the mechanics are different.

The syntax matches v-if — write v-show="..." as an attribute.

The difference between v-if and v-show is whether the DOM node is added and removed, or just hidden with display.

v-if creates the element when the condition turns truthy and removes it from the DOM when the condition turns falsy. v-show keeps the element in the DOM from the start and only toggles display: none. As a result, the toggle cost is lower with v-show, while the initial-render cost is lower with v-if.

A rough rule of thumb:

Element that’s likely falsy at first paint and rarely shown after — v-if
Toggle UI that flips on and off frequently — v-show

Note that v-show leans on the display property, so it doesn’t compose with <template> elements or with v-else. In those cases, use v-if.

v-for

v-for iterates over an array or object and renders the template once per entry.

It maps closely to a JavaScript for loop — Vue lays out the same template for each item in the list.

The form is v-for="(item, index) in list". Pass a unique value to :key so Vue can diff the list correctly.

This snippet displays num is 1, index is 0. num is 2, index is 1.

<template>
  <div>
    <div v-for="(num, i) in numList" :key="i">
      中身は{{ num }}, indexは{{ i }}。
    </div>
  </div>
</template>

<script>
export default {
  data() {
    return {
      numList: [1, 2]
    };
  },
};
</script>

Wrapping up

v-if creates or removes the element based on the condition; v-show keeps the element and hides it with CSS. v-for needs :key to iterate cleanly. The template syntax carries over to Vue 3 unchanged, so this same shape works once you move to the Composition API.

Related posts:

Getting started with Vue.js — v-bind basics, explained

Wed, 14 Apr 2021 00:00:00 GMT

2026-05-16 update: An older post. The bones still hold up, but the tool UI and versions are from the original writing — newer releases may differ.

A short intro to Vue.js’s own HTML notation, written for people picking up Vue for the first time.

The takeaway

The mustache tag ({{ }}) outputs script-side variables into HTML
v-bind is a directive that binds attributes to dynamic expressions
: is the shorthand prefix for v-bind:
Array notation lets you apply multiple classes or styles at once

Mustache tag

The mustache tag is for surfacing variables or function results — defined in script — inside the HTML.

You use it by writing {{ }} somewhere in the template.

The code below defines displayText in script and prints it from HTML.

When you run it, the output reads 文字列です。.

<template>
  {{ displayText }}
</template>

<script>
export default {
  data() {
    return {
      displayText: '文字列です。'
    };
  },
};
</script>

That’s the simplest possible form.

The example uses a string variable, but the same syntax accepts an object.

<template>
  {{ displayObj }}
</template>

<script>
export default {
  data() {
    return {
      displayObj: {
        text: '文字列です。'
      }
    };
  },
};
</script>

Worth noting: running this prints { text: '文字列です。' } literally.

To render 文字列です。 on its own, write displayObj.text.

v-bind

v-bind is the main thing this post is about.

v-bind lets you bind attributes to dynamic logic with almost no ceremony.

Toggling class dynamically opens up a lot of expressive room on the page, so this is where it gets used most.

Basic v-bind usage

The code below applies the activeClass class based on the boolean isActive.

When you run it, 文字列です。 appears in white text.

<template>
  <div v-bind:class="{ activeClass: isActive }">
    サンプルです。
  </div>
</template>

<script>
export default {
  data() {
    return {
      isActive: true
    };
  },
};
</script>

<style lang="css">
  .activeClass {
    color: #fff;
  }
</style>

isActive is defined in script, and the HTML’s v-bind:class evaluates it.

Here isActive is true, so the class applies. Flip it to false and the text renders in black.

v-bind shorthand

v-bind has a shorthand.

Prefix the attribute with : and v-bind kicks in.

The code below applies the activeClass class based on the boolean isActive.

When you run it, 文字列です。 appears in black text.

The reason it’s black: the expression uses ! to negate the value.

<template>
  <div :class="{ activeClass: !isActive }">
    サンプルです。
  </div>
</template>

<script>
export default {
  data() {
    return {
      isActive: true
    };
  },
};
</script>

<style lang="css">
  .activeClass {
    color: #fff;
  }
</style>

Multiple v-bind

To bind more than one class condition on a single attribute, wrap the expressions in an array.

The code below applies both activeClass and sampleClass.

When you run it, 文字列です。 appears in white text.

<template>
  <div :class="[{ activeClass: isActive }, { sampleClass: sampleNum === 1 }]">
    サンプルです。
  </div>
</template>

<script>
export default {
  data() {
    return {
      isActive: true,
      sampleNum: 1
    };
  },
};
</script>

<style lang="css">
  .activeClass {
    color: #fff;
  }

  .sampleClass {
    width: 100%;
  }
</style>

Each entry is a normal JS expression, so comparison operators are fair game.

v-bind works on essentially any attribute. The ones it ends up on most often are class, style, and src.

Closing

That’s the mustache tag and v-bind in Vue.js — the foundation everything else builds on.

Worth getting comfortable with before moving on.

Sennheiser HD 598 SR review — open-back spaciousness and a fit that lasts for hours

Tue, 13 Apr 2021 00:00:00 GMT

2026 update: Ported from the old VuePress blog. The HD 598 SR is an Amazon-exclusive black variant, and as of 2026 new-stock circulation has all but ended. On the open-back side, the HD 599 and HD 560S remain in the current lineup as natural successors. The review angles here — open-back soundstage, fit, and how it compares against closed-backs — should still be useful, so I kept the bones and only cleaned up the wording and structure.

I have used the Sennheiser HD 598 SR open-back headphones for a long stretch of work and personal listening, and this is the write-up.

For anyone who has not owned an open-back before, the first wall is the same: “sound leaks out, outside sound leaks in.” Whether you can live with that splits the verdict cleanly in two. If you can — at home, in your own room, doing remote work — the HD 598 SR is a more honest way to enjoy music than most closed-backs at the same price.

The verdict — still a working pick for a fixed home setup

Short answer: the strengths are the open-back spaciousness and the lightness over long sessions. The price for that is zero isolation and audible leakage, so it does not belong outside or in a shared room.

Three reasons.

Wide soundstage: as an open-back, the “inside-the-skull” feeling is muted. It pays off for film, streaming, and live recordings
Light fit: the clamp is moderate, and the over-ear cups fully surround the ear, so multi-hour use does not tire you out
Removable ear pads: cleaning and replacement are easy, which suits long-term ownership

On the other hand, look elsewhere if any of these apply:

You also want to use them outside or on a commute: open-backs leak heavily in both directions and are unusable in public spaces
You share a room and listen late at night: someone seated next to you will faintly hear what you are playing
You need wireless: the HD 598 SR is wired only, with no Bluetooth support

Build and accessories

Short answer: the all-black finish reads as calm and restrained. As an Amazon-exclusive color, it is the one to pick if the cream of the standard HD 598 is not your taste.

The HD 598 SR is the Amazon-exclusive variant of the HD 598, with the cream finish swapped for black. For most home and remote-work setups, the SR is the easier color to live with.

The housings are vertically elongated ovals. If you are used to round earcups it looks unusual at first, but the shape follows the outline of the ear and settles in cleanly once worn.

The cable is detachable and exits from the bottom of the right cup. Being able to swap only the cable when it eventually breaks is a real plus for long-term ownership.

The bundled accessories are as follows.

In the box

3 m cable with a 6.3 mm stereo plug (for desktop amps and audio interfaces)

1.2 m cable with a 3.5 mm straight plug (for PCs and phones)

6.3 mm to 3.5 mm adapter

The inside of the headband is cushioned. The L / R marks live on the inside as well, so you can tell the sides apart by feel in a dark room.

The length adjustment is nearly continuous in feel, so larger heads fit without issue (the photo shows the headband at full extension).

What worked

Short answer: the soundstage, the long-session comfort, and the ease of maintenance — those three carry the headphone.

Open-back soundstage and an honest balance

As an open-back design, the HD 598 SR avoids the “stuffy, inside-the-skull” feeling closed-backs tend to have. There is a sense of air, closer to listening through speakers.

Low end on an open-back will always sit below an equivalent closed-back by construction, but the HD 598 SR puts out what the music needs. Bass and kick do not vanish, and no genre falls apart.

If your daily listening is EDM or hip-hop where the low end has to push the song, this is not the right fit. For most other listening — vocals, acoustic music, film, games — it stays enjoyable without forcing anything.

Over-ear comfort that holds up over long sessions

The combination of fully over-ear cups, an open-back chamber, and a moderate clamp means very little listening fatigue.

Unlike in-ears or closed-backs, heat does not build up around or inside the ear, so it stays comfortable even in summer. Half a day of wearing them during remote work is genuinely low stress.

Tool-less, removable ear pads

To remove a pad, orient the logo downward and pull it straight off (it comes free with a soft click). Cleaning and replacement are easy, and that pays back the longer you own the headphone.

After a few years of use, swapping only the pads extends the life of the rest of the headphone. Sennheiser has supplied replacement pads for the HD 598 family for a long time.

What didn’t

Short answer: zero isolation, wired-only operation, and the practical difficulty of buying one new today.

Zero isolation — not for outside, not for late nights

By construction, leakage in and out happens essentially without filtering.

People next to you can faintly hear what you are playing
Voices and household sounds come straight in
Noise cancellation is, of course, not on the table

This is a design characteristic rather than a flaw — but it is the line item that decides whether the headphone is “wrong tool for the job.” For listening alone in your own room at home, it is not an issue. For an office, a cafe, or a bedroom that shares a wall with family, this is not the headphone to choose.

Wired only, no Bluetooth

By 2026 the default is wireless. The HD 598 SR is wired only, designed to plug straight into a PC, a desktop amp, or a phone.

Phones no longer carry a 3.5 mm jack in most cases, so a USB-C or Lightning DAC dongle is an extra step in the chain.

Price band and current availability

At the time of writing, the street price was around 20,000 yen. That is not entry-level.

On top of that, as of 2026 new-stock circulation has effectively ended, which makes “buy new” hard to recommend. Looking at the successor open-backs — the HD 599 and the HD 560S — is the more realistic move.

Comparison: open-back HD 598 SR vs same-price closed-back / wireless

Short answer: the HD 598 SR wins on “at home, fixed setup, wired, soundstage-first.” Outside that, a different category is more comfortable.

Angle	HD 598 SR (open-back, wired)	Same-price closed-back	Same-price wireless (with ANC)
Soundstage	Wide	Narrower	Depends on model (typically narrower)
Bass quantity	Enough for the music	Stronger	Strong (DSP-assisted)
Isolation	None (open-back)	High	High (with ANC)
Sound leakage	Large	Small	Small
Long-session fatigue	Low	Heat builds up	Battery weight adds up
Cable	Wired only	Mostly wired	Mostly wireless
Outdoor / commute use	No	Yes	Best fit
Home remote work	Best fit	Yes	Yes (needs charging)

The HD 598 SR’s winning lane is narrow: at home, in your own room, wired is fine, soundstage matters. For anyone who lands in that lane, the listening experience is clearly better than a same-price closed-back or a budget wireless.

If there is any chance you will want to use them outside or on a commute, buying a closed-back or a wireless ANC pair from the start tends to have fewer regrets.

FAQ

Q. Can I plug them straight into a phone? A. If the phone still has a 3.5 mm jack, the 1.2 m cable plugs in directly. Most current phones have dropped the jack, so a USB-C or Lightning DAC dongle is needed separately. By impedance, the HD 598 SR is generally easy enough to drive from a phone.

Q. How does it differ from a closed-back? A. The biggest gaps are soundstage and isolation. An open-back like the HD 598 SR has speaker-like air around the sound, without the boxed-in feeling. The trade is heavy leakage in both directions. A closed-back does the opposite — strong isolation and strong low end, at the cost of heat and pressure around the ear over long sessions. Open-back for listening alone in your own room, closed-back for outside or shared rooms — that is the simple split.

Q. How does it compare to wireless headphones? A. Setting tuning aside, the operational gap is the big one. Wireless asks you to manage charging, pairing, and latency, in exchange for the freedom of no cable. The HD 598 SR has no charging, no latency, and a stable connection by virtue of being wired, but the cable is always there. For a desk-bound remote-work setup, wired is often the easier choice.

Q. Are they comfortable over long sessions? A. The combination of an over-ear cup, an open-back chamber, and a light clamp puts fatigue on the low end of the same-price range, in my experience. Unlike in-ears that push into the canal, these only surround the ear, so the pressure is mild. Comfort still varies by head shape, so trying a pair for ten or fifteen minutes in a store before committing is the safe move.

Verdict — for someone happy to keep them at home

The HD 598 SR is a still-useful open-back if you are clear about the use case.

It suits home listening — remote work, music, film, games in your own room — where the wide soundstage, the long-session comfort, and the easy maintenance all show up in daily use.

It does not suit using them outside, sharing time with family in the same room, or operating wireless-first. For those uses, the design loses by construction.

If you are buying new in 2026, new stock has thinned out, so it is worth looking at the successor lineup (HD 599 / HD 560S) at the same time. On the secondhand or clearance market, the HD 598 SR is “the black HD 598” — the sound is the same as the standard version, so picking on color preference is fine.

Elecom EX-G Bluetooth M-XGL10BBBK review — the budget mouse where grip differences show up most

Mon, 12 Apr 2021 00:00:00 GMT

2026 update: Ported from the old VuePress blog. The M-XGL10BBBK is still in circulation as a current budget Bluetooth mouse in 2026. The review angles here (size options, grip comfort, wireless stability) still apply. Spec values quoted are the manufacturer’s figures from the time of writing, so check the latest product page before buying.

I used the Elecom EX-G Bluetooth M-XGL10BBBK at work for about a year. This is a review combining what it felt like then with a 2026 re-read of the same points.

In the “I don’t want to spend much on a mouse, but I do want to graduate from the one that came in the box” price band, this is the model where the grip difference is most obvious. Being able to pick S, M, or L is also rare at this price.

The verdict — the best “first step up” if you want size options

Short answer: if you’re moving up from a bundled cheap mouse, this is the one in its price band where the difference in grip is most clearly felt. The S/M/L sizing makes it easy to recommend to people with smaller hands too.

The reason is that the ergonomic shape and the side buttons are both there — you can feel the design intent in your fingers the moment you pick it up. Coming from a generic bundled mouse, the way your hand stiffens during long sessions visibly improves.

That said, look elsewhere if any of the following apply:

You prioritize wireless stability: it is Bluetooth-only, so it will not be as steady as a USB-receiver model
You want fine button customization: the bundled software is more modest than the Logicool Options family
Gaming use: response speed and sensor performance are tuned for office work

Build — a mid-height ergonomic shape

Short answer: a shape that sits between flat mice and tall sculpted mice. An asymmetric design that drops away on the right-click side, made to be palmed rather than clawed.

The shape from above. Photographed in L size. You pick S, M, or L to match your hand.

The underside holds the power switch, pairing button, and battery cover. It runs on a single AA cell, and the manufacturer’s rated battery life is about 402 days (check the latest spec before buying).

Seen from the front, the right-click side drops away noticeably. Not flat, not fully sculpted — a mid-height ergonomic shape.

From the back. The hump is modest — it sits in your whole palm. If you only ever claw-grip, the MX Master family will suit you better.

The right flank is wrapped in rubber, and the ring and little fingers stop there without sliding. Even with sweaty hands during long sessions, it does not slip.

The left flank has a thumb rest and two side buttons. The thumb area is shaped with a gentle convex curve that matches the natural shape of the thumb, so the thumb settles there without thinking about it.

What worked — grip and side buttons that punch above the price

Short answer: a deliberate grip, side buttons, S/M/L sizing, and a light weight. The basics of a budget mouse, all present.

A grip designed to be held

The defining trait of this mouse is that it pushes its budget into the way it feels in the hand.

Moving over from a bundled mouse or a thousand-yen plain one, the difference is obvious the moment you pick it up. The rubber on the right flank, the convex thumb rest, the drop on the right-click side — none of them look dramatic on their own, but together they change how your hand feels at the end of a long day.

Side buttons make “back” a one-finger action

Side buttons connect directly to workflow.

The main use is browser back/forward. Web browsing, file managers, IDEs — all of them rely on it constantly, so having the buttons or not having them is a real gap. That alone is enough reason to leave a bundled mouse behind.

S / M / L sizing

Rare in this price band: three sizes to pick from.

L suits adult-male-sized hands, S works for children and smaller hands, M sits in between. A mouse that does not fit your hand will tire you out no matter how much it cost, so size choices at this price are quietly useful.

About 87 g without the battery

The rated weight is about 87 g without the battery.

A mouse over 100 g changes how tired your arm gets across a full day, so for office work, light is right. Even with a AA inserted it lands on the lighter side by feel.

What didn’t — the Bluetooth-only trade-off

Short answer: connection is Bluetooth-only, so the wireless link is less stable than a USB-receiver model. The lag right after waking from sleep is the most common case.

Pickup delay when resuming from sleep

The biggest complaint.

For the first few seconds after the PC wakes, the mouse sometimes does not respond. That is a Bluetooth-general behavior rather than a specific defect of this mouse, but coming from a USB-receiver model you will notice it.

If you sleep and resume often (a laptop-and-dock setup, for example), it is safer to also consider models that ship with a USB receiver, such as the Logicool Unifying-compatible ones.

Modest button-remapping software

It cannot do the fine-grained gestures the Logicool Options family allows. If you want to map the side buttons to anything beyond simple back/forward, you will end up leaning on a third-party Windows utility.

That said, the basics (DPI, button presses) work reliably, so for office use without elaborate remapping, there is no real inconvenience.

Comparison — where it sits against the same and higher tiers

Short answer: at the budget end, it stands out on grip. Compared to higher-tier mice like the Logicool MX line, the side wheel and customization are where the gap shows up.

Angle	M-XGL10BBBK	Logicool mid-range (e.g. M650)	Sub-1,500-yen budget mice
Price	Around 2,000 yen	Around 4,000–6,000 yen	Around 1,000 yen
Connection	Bluetooth only	Bluetooth + USB receiver	Bluetooth or wired
Sizes	S / M / L	M / L (model-dependent)	Usually none
Grip	Ergonomic shape, rubber grip	Ergonomic-leaning, smoother	Mostly flat
Side buttons	2 buttons	2 buttons + custom	None to 2
Side wheel	No	Yes on some models	No
Battery	1 × AA (about 402 days rated)	1 × AA or rechargeable	AA or USB-powered
Remapping software	Yes (modest)	Logi Options+ (full-featured)	Mostly none

Put briefly: pick the M-XGL10BBBK for the “one step up from a cheap mouse” use case, and the Logicool MX line when you want another step up that includes Excel horizontal scrolling. For the comparison with higher-tier MX Master mice, see the Logitech MX MASTER 2S review.

FAQ

Q. Does it work with phones and tablets, not just PCs? A. It connects over Bluetooth, so any OS with Bluetooth-mouse support will work — Windows, macOS, iPadOS 13.4 and later, parts of Android. Without a USB receiver, compatibility depends on the host OS’s Bluetooth support.

Q. How far can the bundled software remap buttons? A. Elecom provides “Elecom Mouse Assistant,” which remaps the side buttons and similar (Windows only — check the latest supported OS list). It is not as fine-grained as the Logicool Options family — things like “different actions per gesture direction” are not on offer. For simple back/forward/copy/paste-level remapping, it is enough.

Q. Is it less tiring for smaller hands? A. Pick the S size and it is less tiring. The L size, for smaller hands, actually makes the fingers stretch and tires the hand faster, so size choice matters. If you can, holding one in a store is the surest way to avoid regret.

Q. Can it be used for gaming? A. Response speed and sensor performance are tuned for office work, so it is not a good fit for FPS or anything that needs precise input. For browser games or low-intensity MMOs, it is fine.

Verdict — a fitting upgrade for your first “real” mouse

On the price-to-grip ratio, this is one of the standouts in the budget band.

Especially for “I want to move on from a bundled mouse or a flat thousand-yen one” and “I want to try a mouse that fits my hand size without spending much,” this is a safe first pick.

The two things to keep in mind are the Bluetooth-only stability and the modest software. If you live in Excel horizontal scrolling or lean on custom gestures, you’ll be happier in the long run with the Logicool MX line.

How to Choose an iPad for Illustration: Paper Size and Pressure

Sun, 11 Apr 2021 00:00:00 GMT

Among current iPads, to draw A5 at actual size with Apple Pencil Pro (the pressure-sensing pen), you need an 11-inch-class screen or larger; at that size it’s borderline. On 13-inch-class models, both the screen’s long and short sides exceed A5’s by at least 48.5 mm, counting the room on both edges together.

If A6 is enough and portability comes first, the 8.3 in model, though that screen is below CLIP STUDIO PAINT’s recommended 10.5 inches.

No current iPad shows a whole A4 page (210 × 297 mm) at actual size. Worked out from Apple’s diagonal and resolution, even the largest 13-inch screen is only about 264.1 × 198.1 mm.

Screen size vs. paper size at actual size

A5 (148 × 210 mm) is shown whole at actual size on every model except the iPad mini. On the 11-inch iPad Air and iPad Pro and the base iPad (A16 chip), though, the short side has about 12 mm or less to spare, so it’s borderline.

Assumptions: the whole screen is compared with the paper’s long and short sides, before app toolbars are subtracted. Fits means shown whole at actual size. Turning the iPad to portrait gives the same result.

Model	Screen size	A6	A5	A4
13-inch iPad Pro	about 264.1 × 198.1 mm	✓ Fits at actual size	✓ Fits at actual size	Fits scaled to 88%
13-inch iPad Air	about 262.1 × 196.5 mm	✓ Fits at actual size	✓ Fits at actual size	Fits scaled to 88%
11-inch iPad Pro	about 232.1 × 160.0 mm	✓ Fits at actual size	△ Fits at actual size (borderline)	Fits scaled to 76%
11-inch iPad Air	about 226.5 × 157.4 mm	✓ Fits at actual size	△ Fits at actual size (borderline)	Fits scaled to 74%
Base iPad (A16 chip)	about 226.5 × 157.4 mm	✓ Fits at actual size	△ Fits at actual size (borderline)	Fits scaled to 74%
iPad mini	about 176.2 × 115.7 mm	✓ Fits at actual size	Fits scaled to 78%	Fits scaled to 55%

“Borderline” means the spare room is under 10% of the paper’s side.

Apple doesn’t publish the screens’ width and height, so they’re calculated from the diagonal and resolution. Using Apple’s official ppi instead changes the result by less than 0.7 mm and doesn’t change whether a page fits at actual size.

A4 has to be scaled down to fit whole. The largest scale is 88% on the 13-inch iPad Air and iPad Pro, and 55% on the iPad mini. At the largest scale in 1% steps, the tighter side of every model has at most 2.8 mm to spare, so it’s borderline.

The iPad mini holds a whole page at actual size only up to A6 (105 × 148 mm). A5 fits when scaled to 78%, with only 0.2 mm to spare, so it’s borderline.

Which iPad fits the paper size you draw at

Pick the paper size you draw at, whether you need pressure or hover, and your app to see the smallest current iPad that holds a whole page at actual size. If none does, it shows the scale that fits, plus one iPad you can buy that fits.

11-inch or 13-inch: room vs. weight

For A5, pick the 13-inch if you want more room, or the 11-inch if weight matters more. Per Apple’s spec pages, the 13-inch models are heavier than the 11-inch ones, by 152 g for the iPad Air and 135 g for the iPad Pro (Wi-Fi models).

Apple Pencil pressure and hover

In Apple’s compatibility page and comparison table, Apple Pencil Pro works with the iPad Pro, iPad Air and iPad mini. It doesn’t work with the base iPad (A16 chip).

Only Apple Pencil Pro is listed with pressure sensing; Apple Pencil (USB-C) doesn’t have it, but it does sense tilt.

The base iPad (A16 chip) works with the USB-C Apple Pencil. It can also use the 1st-generation Apple Pencil with an adapter. Neither is listed with pressure, so the base iPad has no pressure-sensing pen here.

Hover, which shows where the pen tip is before it touches the screen, works on the iPad Pro, iPad Air and iPad mini. The base iPad (A16 chip) isn’t on the list.

What the drawing apps need

Per Procreate’s Help Center, the current version, 5.4.14, needs iPadOS 16.3 or later and has no list of models. Every current iPad runs iPadOS 27, so all of them qualify.

Per CLIP STUDIO PAINT’s requirements and device pages, the list names Apple Pencil Pro and the 1st and 2nd generations for pressure, not the USB-C one. Current iPads can use Pro, and the base iPad (A16 chip) can use the 1st generation with an adapter. No current iPad takes the 2nd generation.

The current model on the list for high-resolution illustration and manga is the iPad Pro. The base iPad (A16 chip) and iPad mini are on the list for standard-size illustration.

The iPad Air (M4) is not on the supported-device page. The page doesn’t say it won’t run; it just isn’t listed.

The supported OS versions are iPadOS 18 and iPadOS 26; iPadOS 27 isn’t mentioned.

Current models compared (screen, weight, memory, battery, color)

Model	Screen diagonal	Weight	Memory	Battery (Apple’s tests)	ProMotion	Color and lamination
13-inch iPad Pro	13 in	579 g	12 GB for 256GB and 512GB models, 16 GB for 1TB and 2TB models (unified memory)	up to 10 hours	Yes (10-120 Hz)	P3, fully laminated
13-inch iPad Air	12.9 in	616 g	12 GB (unified memory)	up to 10 hours	Not listed	P3, fully laminated
11-inch iPad Pro	11.1 in	444 g	12 GB for 256GB and 512GB models, 16 GB for 1TB and 2TB models (unified memory)	up to 10 hours	Yes (10-120 Hz)	P3, fully laminated
11-inch iPad Air	10.86 in	464 g	12 GB (unified memory)	up to 10 hours	Not listed	P3, fully laminated
Base iPad (A16 chip)	10.86 in	477 g	Not listed by Apple	up to 10 hours	Not listed	P3 and full lamination not listed (comparison page says sRGB)
iPad mini	8.3 in	293 g	Not listed by Apple	up to 10 hours	Not listed	P3, fully laminated

Weights are Wi-Fi models. Battery isn’t drawing time.

Per Apple’s spec pages, memory is listed at 12 GB or 16 GB for the iPad Pro and 12 GB for the iPad Air. Both meet CLIP STUDIO PAINT’s high-resolution recommendation of 6 GB or more. The base iPad (A16 chip) and iPad mini list none, so that can’t be confirmed.

ProMotion is a display that adjusts its refresh rate to what’s on screen. Its range, 10Hz to 120Hz, is listed only for the iPad Pro. The iPad Air, the base iPad (A16 chip) and iPad mini specs give neither ProMotion nor a refresh-rate figure.

Wide color (P3) is a range wider than sRGB, and full lamination means no air gap between glass and display. Both are listed for the iPad Pro, Air and mini. The base iPad (A16 chip) lists neither.

Picks by condition

One pick per condition, from Amazon.com, if a listing meets the criteria. If several qualify, the pick is the cheapest of those checked.

Sold new by Amazon.com
Configuration matches Apple’s spec
At least 100 ratings and 4 stars or more

Check prices on the retailer’s page.

In the checks here, on Amazon.com, every pick but the ProMotion one is 128 GB, the cheapest qualifying tier, not a judgment that it’s enough. The ProMotion pick is 256 GB. CLIP STUDIO PAINT needs 6 GB of free space; no official storage guide for a given use was found.

Pressure with Apple Pencil Pro, smallest screen that holds A5 at actual size

The A5 margin is under 10% of the paper’s side, so it’s borderline.

Color variants are listed; price may vary by color. The rating count is shared with other models on the same listing.

Pressure with Apple Pencil Pro, smallest and lightest screen, A6 at actual size

In this calculation, A6 fits at actual size and A5 only scaled to 78%, borderline.

The rating count is shared with other models on the same listing.

No pressure sensing needed, A5 at actual size

A5 fits at actual size with about 12 mm or less to spare on the short side, so it’s borderline.

The rating count is shared with other models on the same listing.

13-inch screen with room around A5, A4 scaled to 88%

In this calculation, A4 fits only scaled to 88%, borderline.

The rating count is shared with other models on the same listing.

ProMotion, iPad Pro only

How ProMotion feels to draw on isn’t judged.

The rating count is shared with other models on the same listing.

FAQ

Which iPad is best for drawing and illustration?

Among current iPads, to draw A5 at actual size with Apple Pencil Pro (the pressure-sensing pen), you need an 11-inch-class screen or larger; at that size it’s borderline. If A6 is enough and portability comes first, the 8.3 in model, though that screen is below CLIP STUDIO PAINT’s recommended 10.5 inches.

Models that fit are in the picks section.

Can you draw A4 at actual size on an iPad?

No current iPad shows a whole A4 page (210 × 297 mm) at actual size. Worked out from Apple’s diagonal and resolution, even the largest 13-inch screen is only about 264.1 × 198.1 mm.

Can you draw with the Apple Pencil (USB-C)?

Yes. In Apple’s compatibility page and comparison table, only Apple Pencil Pro is listed with pressure sensing; Apple Pencil (USB-C) doesn’t have it, but it does sense tilt.

Is the base iPad (A16 chip) good enough for illustration?

Yes, it can draw. In Apple’s compatibility page and comparison table, the A16-chip model works with the USB-C Apple Pencil.

It can also use the 1st-generation Apple Pencil with an adapter. Neither is listed with pressure, so that model has no pressure-sensing pen here.

Which iPads run Procreate?

Per Procreate’s Help Center, the current version, 5.4.14, needs iPadOS 16.3 or later and has no list of models. Every current iPad runs iPadOS 27, so all of them qualify.

Which iPads run CLIP STUDIO PAINT?

Per CLIP STUDIO PAINT’s requirements and device pages, the current models on the list for high-resolution illustration and manga have the M5 chip. The A16 and A17 Pro models are on the list for standard-size illustration.

The current M4 models are not on the supported-device page. The page doesn’t say they won’t run; they just aren’t listed.

The supported OS versions are iPadOS 18 and iPadOS 26; iPadOS 27 isn’t mentioned. The recommended display size is 10.5 inches or larger, and the 8.3 in model is smaller than that.

Data and formulas used (for verification)

How this was checked: In the checks here, only screen size, the type of Apple Pencil and the apps’ official requirements are used. Drawing feel and speed weren’t tested on real hardware and aren’t judged. Sources as of October 2026: Apple’s Compare iPad, iPad Pro specs and Apple Pencil compatibility.

Sources: CLIP STUDIO PAINT system requirements, supported-device FAQ, Procreate FAQ and the ISO 216 table

Method

Current lineup: the six models in ‘Current iPad models’ on Apple’s compare page: 13- and 11-inch iPad Pro (M5), 13- and 11-inch iPad Air (M4), iPad (A16) and iPad mini (A17 Pro). Older models are not covered.
Specs: Apple’s spec page for each model (Japan and US) and the compare page. Memory is quoted only where Apple states it (iPad Pro and iPad Air); otherwise ‘not stated’ and no figure is guessed.
Battery life is Apple’s test figure on pre-production units, not drawing time.
Apple Pencil support: Apple’s ‘Apple Pencil compatibility’ page (published 2026-03-31) and the Apple Pencil page. The page states pressure sensing is on Apple Pencil Pro but not Apple Pencil (USB-C).
App requirements: Procreate 5.4.14 (current per its Help Center) needs iPadOS 16.3 or later; CLIP STUDIO PAINT for iPad per its official requirements (iPadOS 18 / 26, 3GB required, 4GB recommended, 6GB for high-resolution work) and its supported-device FAQ.
Products: new Wi-Fi models with standard glass on Amazon.co.jp and Amazon.com. A candidate needs 100+ ratings, 4.0+ stars, Amazon as seller, a configuration matching Apple’s spec, and (Japan) no strong sakura warning.
Drawing area = screen size. Apple does not publish the screen’s width and height, so they are calculated from the official diagonal (measured as a rectangle) and the pixel resolution: width = diagonal x 25.4 x horizontal px / sqrt(horizontal px^2 + vertical px^2), height = diagonal x 25.4 x vertical px / sqrt(horizontal px^2 + vertical px^2). Cross-checked against the official ppi (264, 326) as px / ppi x 25.4: the difference is under 1 mm, whether a page fits at actual size is the same either way, and the scale from the diagonal is equal to or 1% below the ppi-based one (the safe side). Apple notes the corners are rounded and the actual viewable area is smaller. Space taken by a drawing app’s on-screen controls is not subtracted.
Paper is ISO 216 A-series (A6 to A3), with the same sizes and fit rule as the pen-display guide (paper long side vs screen long side, short vs short; portrait gives the same result). For paper that doesn’t fit at actual size, the scale that fits the whole page = min(screen long / paper long, screen short / paper short), rounded down to 1%. Room to spare (screen minus paper) under 10% of the paper side is called borderline.
ProMotion (10Hz to 120Hz) is listed only for iPad Pro, as on Apple’s spec pages. Drawing feel and speed weren’t tested on real hardware and aren’t judged. Prices are not given.

Official values used

Item	Value	Source	Checked
13-inch iPad Pro (M5): screen diagonal (measured as a rectangle)	13 inch	apple.com	2026-10-08
13-inch iPad Pro (M5): resolution (long side)	2752 px	apple.com	2026-10-08
13-inch iPad Pro (M5): resolution (short side)	2064 px	apple.com	2026-10-08
13-inch iPad Pro (M5): pixel density	264 ppi	apple.com	2026-10-08
13-inch iPad Pro (M5): weight (Wi-Fi model)	579 g	apple.com	2026-10-08
13-inch iPad Pro (M5): weight (Wi-Fi + Cellular model)	582 g	apple.com	2026-10-08
13-inch iPad Pro (M5): body height	281.6 mm	apple.com	2026-10-08
13-inch iPad Pro (M5): body width	215.5 mm	apple.com	2026-10-08
13-inch iPad Pro (M5): body thickness	5.1 mm	apple.com	2026-10-08
13-inch iPad Pro (M5): chip	M5 text	apple.com	2026-10-08
13-inch iPad Pro (M5): storage options	256GB, 512GB, 1TB, 2TB	apple.com	2026-10-08
13-inch iPad Pro (M5): memory	12GB on 256GB/512GB models, 16GB on 1TB/2TB models (unified memory)	apple.com	2026-10-08
13-inch iPad Pro (M5): port	Thunderbolt / USB 4 port (USB-C connector). USB 4 (up to 40Gb/s)	apple.com	2026-10-08
13-inch iPad Pro (M5): battery capacity	38.99 Wh	apple.com	2026-10-08
13-inch iPad Pro (M5): battery life (Apple test figure, web browsing and video over Wi-Fi)	up to 10 hours	apple.com	2026-10-08
13-inch iPad Pro (M5): refresh rate	ProMotion technology with an adaptive refresh rate of 10Hz to 120Hz	apple.com	2026-10-08
13-inch iPad Pro (M5): display type, lamination and color gamut	Ultra Retina XDR (tandem OLED). Fully laminated. Wide color (P3)	apple.com	2026-10-08
13-inch iPad Pro (M5): Nano-texture glass	Nano-texture display glass is an option on 1TB and 2TB models	apple.com	2026-10-08
13-inch iPad Pro (M5): supported Apple Pencil	Apple Pencil Pro, Apple Pencil (USB-C)	support.apple.com	2026-10-08
13-inch iPad Pro (M5): Apple Pencil hover	Supported	apple.com	2026-10-08
11-inch iPad Pro (M5): screen diagonal (measured as a rectangle)	11.1 inch	apple.com	2026-10-08
11-inch iPad Pro (M5): resolution (long side)	2420 px	apple.com	2026-10-08
11-inch iPad Pro (M5): resolution (short side)	1668 px	apple.com	2026-10-08
11-inch iPad Pro (M5): pixel density	264 ppi	apple.com	2026-10-08
11-inch iPad Pro (M5): weight (Wi-Fi model)	444 g	apple.com	2026-10-08
11-inch iPad Pro (M5): weight (Wi-Fi + Cellular model)	446 g	apple.com	2026-10-08
11-inch iPad Pro (M5): body height	249.7 mm	apple.com	2026-10-08
11-inch iPad Pro (M5): body width	177.5 mm	apple.com	2026-10-08
11-inch iPad Pro (M5): body thickness	5.3 mm	apple.com	2026-10-08
11-inch iPad Pro (M5): chip	M5 text	apple.com	2026-10-08
11-inch iPad Pro (M5): storage options	256GB, 512GB, 1TB, 2TB	apple.com	2026-10-08
11-inch iPad Pro (M5): memory	12GB on 256GB/512GB models, 16GB on 1TB/2TB models (unified memory)	apple.com	2026-10-08
11-inch iPad Pro (M5): port	Thunderbolt / USB 4 port (USB-C connector). USB 4 (up to 40Gb/s)	apple.com	2026-10-08
11-inch iPad Pro (M5): battery capacity	31.29 Wh	apple.com	2026-10-08
11-inch iPad Pro (M5): battery life (Apple test figure, web browsing and video over Wi-Fi)	up to 10 hours	apple.com	2026-10-08
11-inch iPad Pro (M5): refresh rate	ProMotion technology with an adaptive refresh rate of 10Hz to 120Hz	apple.com	2026-10-08
11-inch iPad Pro (M5): display type, lamination and color gamut	Ultra Retina XDR (tandem OLED). Fully laminated. Wide color (P3)	apple.com	2026-10-08
11-inch iPad Pro (M5): Nano-texture glass	Nano-texture display glass is an option on 1TB and 2TB models	apple.com	2026-10-08
11-inch iPad Pro (M5): supported Apple Pencil	Apple Pencil Pro, Apple Pencil (USB-C)	support.apple.com	2026-10-08
11-inch iPad Pro (M5): Apple Pencil hover	Supported	apple.com	2026-10-08
13-inch iPad Air (M4): screen diagonal (measured as a rectangle)	12.9 inch	apple.com	2026-10-08
13-inch iPad Air (M4): resolution (long side)	2732 px	apple.com	2026-10-08
13-inch iPad Air (M4): resolution (short side)	2048 px	apple.com	2026-10-08
13-inch iPad Air (M4): pixel density	264 ppi	apple.com	2026-10-08
13-inch iPad Air (M4): weight (Wi-Fi model)	616 g	apple.com	2026-10-08
13-inch iPad Air (M4): weight (Wi-Fi + Cellular model)	617 g	apple.com	2026-10-08
13-inch iPad Air (M4): body height	280.6 mm	apple.com	2026-10-08
13-inch iPad Air (M4): body width	214.9 mm	apple.com	2026-10-08
13-inch iPad Air (M4): body thickness	6.1 mm	apple.com	2026-10-08
13-inch iPad Air (M4): chip	M4 text	apple.com	2026-10-08
13-inch iPad Air (M4): storage options	128GB, 256GB, 512GB, 1TB	apple.com	2026-10-08
13-inch iPad Air (M4): memory	12GB (unified memory)	apple.com	2026-10-08
13-inch iPad Air (M4): port	USB-C port. USB 3 (up to 10Gb/s)	apple.com	2026-10-08
13-inch iPad Air (M4): battery capacity	36.59 Wh	apple.com	2026-10-08
13-inch iPad Air (M4): battery life (Apple test figure, web browsing and video over Wi-Fi)	up to 10 hours	apple.com	2026-10-08
13-inch iPad Air (M4): refresh rate (not stated by Apple)	No ProMotion or refresh-rate figure stated by Apple	apple.com	2026-10-08
13-inch iPad Air (M4): display type, lamination and color gamut	Liquid Retina (IPS). Fully laminated. Wide color (P3)	apple.com	2026-10-08
13-inch iPad Air (M4): supported Apple Pencil	Apple Pencil Pro, Apple Pencil (USB-C)	support.apple.com	2026-10-08
13-inch iPad Air (M4): Apple Pencil hover	Supported	apple.com	2026-10-08
11-inch iPad Air (M4): screen diagonal (measured as a rectangle)	10.86 inch	apple.com	2026-10-08
11-inch iPad Air (M4): resolution (long side)	2360 px	apple.com	2026-10-08
11-inch iPad Air (M4): resolution (short side)	1640 px	apple.com	2026-10-08
11-inch iPad Air (M4): pixel density	264 ppi	apple.com	2026-10-08
11-inch iPad Air (M4): weight (Wi-Fi model)	464 g	apple.com	2026-10-08
11-inch iPad Air (M4): weight (Wi-Fi + Cellular model)	465 g	apple.com	2026-10-08
11-inch iPad Air (M4): body height	247.6 mm	apple.com	2026-10-08
11-inch iPad Air (M4): body width	178.5 mm	apple.com	2026-10-08
11-inch iPad Air (M4): body thickness	6.1 mm	apple.com	2026-10-08
11-inch iPad Air (M4): chip	M4 text	apple.com	2026-10-08
11-inch iPad Air (M4): storage options	128GB, 256GB, 512GB, 1TB	apple.com	2026-10-08
11-inch iPad Air (M4): memory	12GB (unified memory)	apple.com	2026-10-08
11-inch iPad Air (M4): port	USB-C port. USB 3 (up to 10Gb/s)	apple.com	2026-10-08
11-inch iPad Air (M4): battery capacity	28.93 Wh	apple.com	2026-10-08
11-inch iPad Air (M4): battery life (Apple test figure, web browsing and video over Wi-Fi)	up to 10 hours	apple.com	2026-10-08
11-inch iPad Air (M4): refresh rate (not stated by Apple)	No ProMotion or refresh-rate figure stated by Apple	apple.com	2026-10-08
11-inch iPad Air (M4): display type, lamination and color gamut	Liquid Retina (IPS). Fully laminated. Wide color (P3)	apple.com	2026-10-08
11-inch iPad Air (M4): supported Apple Pencil	Apple Pencil Pro, Apple Pencil (USB-C)	support.apple.com	2026-10-08
11-inch iPad Air (M4): Apple Pencil hover	Supported	apple.com	2026-10-08
iPad (A16): screen diagonal (measured as a rectangle)	10.86 inch	apple.com	2026-10-08
iPad (A16): resolution (long side)	2360 px	apple.com	2026-10-08
iPad (A16): resolution (short side)	1640 px	apple.com	2026-10-08
iPad (A16): pixel density	264 ppi	apple.com	2026-10-08
iPad (A16): weight (Wi-Fi model)	477 g	apple.com	2026-10-08
iPad (A16): weight (Wi-Fi + Cellular model)	481 g	apple.com	2026-10-08
iPad (A16): body height	248.6 mm	apple.com	2026-10-08
iPad (A16): body width	179.5 mm	apple.com	2026-10-08
iPad (A16): body thickness	7 mm	apple.com	2026-10-08
iPad (A16): chip	A16 text	apple.com	2026-10-08
iPad (A16): storage options	128GB, 256GB, 512GB	apple.com	2026-10-08
iPad (A16): memory (not stated by Apple)	not stated by Apple	apple.com	2026-10-08
iPad (A16): port	USB-C port. USB 2.0 (up to 480Mb/s)	apple.com	2026-10-08
iPad (A16): battery capacity	28.93 Wh	apple.com	2026-10-08
iPad (A16): battery life (Apple test figure, web browsing and video over Wi-Fi)	up to 10 hours	apple.com	2026-10-08
iPad (A16): refresh rate (not stated by Apple)	No ProMotion or refresh-rate figure stated by Apple	apple.com	2026-10-08
iPad (A16): display type, lamination and color gamut	Liquid Retina (IPS). No full lamination stated. sRGB color (no P3 stated)	apple.com	2026-10-08
iPad (A16): supported Apple Pencil	Apple Pencil (USB-C), Apple Pencil (1st generation) (needs the USB-C to Apple Pencil Adapter)	support.apple.com	2026-10-08
iPad (A16): Apple Pencil hover	Not listed as supported	apple.com	2026-10-08
iPad mini (A17 Pro): screen diagonal (measured as a rectangle)	8.3 inch	apple.com	2026-10-08
iPad mini (A17 Pro): resolution (long side)	2266 px	apple.com	2026-10-08
iPad mini (A17 Pro): resolution (short side)	1488 px	apple.com	2026-10-08
iPad mini (A17 Pro): pixel density	326 ppi	apple.com	2026-10-08
iPad mini (A17 Pro): weight (Wi-Fi model)	293 g	apple.com	2026-10-08
iPad mini (A17 Pro): weight (Wi-Fi + Cellular model)	297 g	apple.com	2026-10-08
iPad mini (A17 Pro): body height	195.4 mm	apple.com	2026-10-08
iPad mini (A17 Pro): body width	134.8 mm	apple.com	2026-10-08
iPad mini (A17 Pro): body thickness	6.3 mm	apple.com	2026-10-08
iPad mini (A17 Pro): chip	A17 Pro text	apple.com	2026-10-08
iPad mini (A17 Pro): storage options	128GB, 256GB, 512GB	apple.com	2026-10-08
iPad mini (A17 Pro): memory (not stated by Apple)	not stated by Apple	apple.com	2026-10-08
iPad mini (A17 Pro): port	USB-C port. USB 3 (up to 10Gb/s)	apple.com	2026-10-08
iPad mini (A17 Pro): battery capacity	19.3 Wh	apple.com	2026-10-08
iPad mini (A17 Pro): battery life (Apple test figure, web browsing and video over Wi-Fi)	up to 10 hours	apple.com	2026-10-08
iPad mini (A17 Pro): refresh rate (not stated by Apple)	No ProMotion or refresh-rate figure stated by Apple	apple.com	2026-10-08
iPad mini (A17 Pro): display type, lamination and color gamut	Liquid Retina (IPS). Fully laminated. Wide color (P3)	apple.com	2026-10-08
iPad mini (A17 Pro): supported Apple Pencil	Apple Pencil Pro, Apple Pencil (USB-C)	support.apple.com	2026-10-08
iPad mini (A17 Pro): Apple Pencil hover	Supported	apple.com	2026-10-08
Apple Pencil Pro: pressure sensing	Supported	apple.com	2026-10-08
Apple Pencil (USB-C): pressure sensing	Not supported	apple.com	2026-10-08
Apple Pencil Pro: tilt sensing	Supported	apple.com	2026-10-08
Apple Pencil (USB-C): tilt sensing	Supported	apple.com	2026-10-08
Features only Apple Pencil Pro has	Squeeze, barrel roll, haptic feedback, double tap and Find My (not on Apple Pencil (USB-C))	apple.com	2026-10-08
iPad models with Apple Pencil hover	Supported on iPad Pro (M4, M5), iPad Air (M2, M3, M4) and iPad mini (A17 Pro), among others; iPad (A16) is not listed	apple.com	2026-10-08
iPadOS needed for Apple Pencil (USB-C)	iPadOS 17.1.1 or later	support.apple.com	2026-10-08
iPad (A16) and Apple Pencil (1st generation)	iPad (A16) also works with Apple Pencil (1st generation) when paired and charged through the USB-C to Apple Pencil Adapter	support.apple.com	2026-10-08
Apple Pencil (2nd generation) support	None of the six current iPads supports Apple Pencil (2nd generation)	support.apple.com	2026-10-08
iPad models Apple currently sells	13-inch iPad Pro (M5), 11-inch iPad Pro (M5), 13-inch iPad Air (M4), 11-inch iPad Air (M4), iPad (A16), iPad mini (A17 Pro)	apple.com	2026-10-08
Latest iPadOS version on Apple’s iPadOS page	iPadOS 27 text	apple.com	2026-10-08
Procreate for iPad: current version	5.4.14 text	help.procreate.com Procreate FAQ: The current version of Procreate for iPad is 5.4.14	2026-10-08
iPadOS required by Procreate 5.4.14 (or later)	16.3 text	help.procreate.com Procreate FAQ: Procreate 5.4.14 requires an iPad running iPadOS 16.3 or newer	2026-10-08
Procreate: current iPads that qualify	Procreate lists no model names, only iPadOS 16.3 or later. All six current iPads run iPadOS 27, so all meet it	help.procreate.com	2026-10-08
Procreate: maximum canvas by device	134 megapixels (11,585px x 11,585px) on iPad Pro (M4, M5) and iPad Air (M2, M3); 67 megapixels (8,192px x 8,192px) on iPad mini (6th and 7th generation). iPad (A16) and iPad Air (M4) are not listed	help.procreate.com	2026-10-08
CLIP STUDIO PAINT for iPad: App Store version	5.1.4 text	apps.apple.com	2026-10-08
CLIP STUDIO PAINT for iPad: OS in the official requirements	iPadOS 18 / iPadOS 26 (iPadOS 27 is not mentioned)	clipstudio.net	2026-10-08
CLIP STUDIO PAINT for iPad: memory requirement	3GB or more required, 4GB or more recommended; 6GB or more recommended for high-resolution illustration and manga	clipstudio.net	2026-10-08
CLIP STUDIO PAINT for iPad: recommended display size (or larger)	10.5 inch	clipstudio.net	2026-10-08
CLIP STUDIO PAINT for iPad: free storage required (or more)	6 GB	clipstudio.net	2026-10-08
CLIP STUDIO PAINT for iPad: pressure-sensitive pens in the official requirements	Apple Pencil (1st generation), Apple Pencil (2nd generation), Apple Pencil Pro (Apple Pencil (USB-C) is not listed)	clipstudio.net	2026-10-08
CLIP STUDIO PAINT supported devices: high-resolution list	High-resolution illustration and manga list: 13- and 11-inch iPad Pro (M5), (M4), 13- and 11-inch iPad Air (M3, M2) and others. iPad Air (M4) is not on the page	support.clip-studio.com	2026-10-08
CLIP STUDIO PAINT supported devices: standard-size list	Standard-size illustration and web manga list: iPad (A16), iPad mini (A17 Pro) and others	support.clip-studio.com	2026-10-08
iPad Air (M4) and CLIP STUDIO PAINT	iPad Air (M4) is not on CLIP STUDIO PAINT’s supported-device page (FAQ updated 2025-12-02)	support.clip-studio.com	2026-10-08
A3 paper short side	297 mm	cl.cam.ac.uk	2026-10-08
A3 paper long side	420 mm	cl.cam.ac.uk	2026-10-08
A4 paper short side	210 mm	cl.cam.ac.uk	2026-10-08
A4 paper long side	297 mm	cl.cam.ac.uk	2026-10-08
A5 paper short side	148 mm	cl.cam.ac.uk	2026-10-08
A5 paper long side	210 mm	cl.cam.ac.uk	2026-10-08
A6 paper short side	105 mm	cl.cam.ac.uk	2026-10-08
A6 paper long side	148 mm	cl.cam.ac.uk	2026-10-08

Which iPad fits the paper size you draw at: quick reference

The calculator in this article shows the result and a matching product for your own settings. Settings can be set in the URL; each link below opens the calculator with those settings.

Settings	Result	Calculator
A6, Pressure needed, Hover optional, No app chosen	✓ Actual size from the iPad mini (A17 Pro). A6: 105 × 148 mm. The iPad mini (A17 Pro) holds a whole A6 page at actual size. At least 10.7 mm to spare. Other models that fit at actual size: 11-inch iPad Air (M4), 11-inch iPad Pro (M5), 13-inch iPad Air (M4), 13-inch iPad Pro (M5). Apple lists pressure sensing only for Apple Pencil Pro, which doesn’t work with the iPad (A16)	Open with these settings
A6, No pressure needed, Hover optional, No app chosen	✓ Actual size from the iPad mini (A17 Pro). A6: 105 × 148 mm. The iPad mini (A17 Pro) holds a whole A6 page at actual size. At least 10.7 mm to spare. Other models that fit at actual size: iPad (A16), 11-inch iPad Air (M4), 11-inch iPad Pro (M5), 13-inch iPad Air (M4), 13-inch iPad Pro (M5). Apple Pencil (USB-C) has no pressure sensing (Apple’s feature comparison)	Open with these settings
A5, Pressure needed, Hover optional, No app chosen	△ Actual size on the 11-inch iPad Air (M4) (borderline). A5: 148 × 210 mm. The 11-inch iPad Air (M4) holds a whole A5 page at actual size. Only 9.4 mm to spare (6% of the paper side), so it’s borderline. Other models that fit at actual size: 11-inch iPad Pro (M5), 13-inch iPad Air (M4), 13-inch iPad Pro (M5). Apple lists pressure sensing only for Apple Pencil Pro, which doesn’t work with the iPad (A16)	Open with these settings
A5, No pressure needed, Hover optional, No app chosen	△ Actual size on the iPad (A16) (borderline). A5: 148 × 210 mm. The iPad (A16) holds a whole A5 page at actual size. Only 9.4 mm to spare (6% of the paper side), so it’s borderline. Other models that fit at actual size: 11-inch iPad Air (M4), 11-inch iPad Pro (M5), 13-inch iPad Air (M4), 13-inch iPad Pro (M5). Apple Pencil (USB-C) has no pressure sensing (Apple’s feature comparison)	Open with these settings
A4, Pressure needed, Hover optional, No app chosen	△ Not at actual size (scale to 88%). A4: 210 × 297 mm. No current iPad holds a whole A4 page at actual size. On the 13-inch iPad Air (M4), 13-inch iPad Pro (M5), it fits whole when scaled to 88%, with only 0.8 mm to spare, so it’s borderline. Apple lists pressure sensing only for Apple Pencil Pro, which doesn’t work with the iPad (A16)	Open with these settings
A4, No pressure needed, Hover optional, No app chosen	△ Not at actual size (scale to 88%). A4: 210 × 297 mm. No current iPad holds a whole A4 page at actual size. On the 13-inch iPad Air (M4), 13-inch iPad Pro (M5), it fits whole when scaled to 88%, with only 0.8 mm to spare, so it’s borderline. Apple Pencil (USB-C) has no pressure sensing (Apple’s feature comparison)	Open with these settings
A3, Pressure needed, Hover optional, No app chosen	△ Not at actual size (scale to 62%). A3: 297 × 420 mm. No current iPad holds a whole A3 page at actual size. On the 13-inch iPad Air (M4), 13-inch iPad Pro (M5), it fits whole when scaled to 62%, with only 1.7 mm to spare, so it’s borderline. Apple lists pressure sensing only for Apple Pencil Pro, which doesn’t work with the iPad (A16)	Open with these settings
A3, No pressure needed, Hover optional, No app chosen	△ Not at actual size (scale to 62%). A3: 297 × 420 mm. No current iPad holds a whole A3 page at actual size. On the 13-inch iPad Air (M4), 13-inch iPad Pro (M5), it fits whole when scaled to 62%, with only 1.7 mm to spare, so it’s borderline. Apple Pencil (USB-C) has no pressure sensing (Apple’s feature comparison)	Open with these settings
A5, Pressure needed, Hover needed, CLIP STUDIO PAINT	△ Actual size on the 11-inch iPad Air (M4) (borderline). A5: 148 × 210 mm. The 11-inch iPad Air (M4) holds a whole A5 page at actual size. Only 9.4 mm to spare (6% of the paper side), so it’s borderline. Other models that fit at actual size: 11-inch iPad Pro (M5), 13-inch iPad Air (M4), 13-inch iPad Pro (M5). Apple lists pressure sensing only for Apple Pencil Pro, which doesn’t work with the iPad (A16). The iPad (A16) isn’t on Apple’s list of models with hover. The 11-inch iPad Air (M4) isn’t on CLIP STUDIO PAINT’s supported-device page. The page doesn’t say it won’t run; it just isn’t listed. The 11-inch iPad Air (M4) has 12 GB of memory (per Apple), meeting CLIP STUDIO PAINT’s high-resolution recommendation of 6 GB or more. CLIP STUDIO PAINT lists iPadOS 18 and iPadOS 26; iPadOS 27 isn’t mentioned	Open with these settings
A5, No pressure needed, Hover optional, CLIP STUDIO PAINT	△ Actual size on the iPad (A16) (borderline). A5: 148 × 210 mm. The iPad (A16) holds a whole A5 page at actual size. Only 9.4 mm to spare (6% of the paper side), so it’s borderline. Other models that fit at actual size: 11-inch iPad Air (M4), 11-inch iPad Pro (M5), 13-inch iPad Air (M4), 13-inch iPad Pro (M5). Apple Pencil (USB-C) has no pressure sensing (Apple’s feature comparison). Apple doesn’t list memory for the iPad (A16), so CLIP STUDIO PAINT’s recommendation can’t be confirmed. CLIP STUDIO PAINT’s list of pressure-sensitive pens doesn’t include Apple Pencil (USB-C). CLIP STUDIO PAINT lists iPadOS 18 and iPadOS 26; iPadOS 27 isn’t mentioned	Open with these settings
A6, Pressure needed, Hover optional, CLIP STUDIO PAINT	✓ Actual size from the iPad mini (A17 Pro). A6: 105 × 148 mm. The iPad mini (A17 Pro) holds a whole A6 page at actual size. At least 10.7 mm to spare. Other models that fit at actual size: 11-inch iPad Air (M4), 11-inch iPad Pro (M5), 13-inch iPad Air (M4), 13-inch iPad Pro (M5). Apple lists pressure sensing only for Apple Pencil Pro, which doesn’t work with the iPad (A16). Apple doesn’t list memory for the iPad mini (A17 Pro), so CLIP STUDIO PAINT’s recommendation can’t be confirmed. The iPad mini (A17 Pro)‘s screen is smaller than CLIP STUDIO PAINT’s recommended 10.5 inches. CLIP STUDIO PAINT lists iPadOS 18 and iPadOS 26; iPadOS 27 isn’t mentioned	Open with these settings
A4, Pressure needed, Hover optional, CLIP STUDIO PAINT	△ Not at actual size (scale to 88%). A4: 210 × 297 mm. No current iPad holds a whole A4 page at actual size. On the 13-inch iPad Air (M4), 13-inch iPad Pro (M5), it fits whole when scaled to 88%, with only 0.8 mm to spare, so it’s borderline. Apple lists pressure sensing only for Apple Pencil Pro, which doesn’t work with the iPad (A16). The 13-inch iPad Air (M4) isn’t on CLIP STUDIO PAINT’s supported-device page. The page doesn’t say it won’t run; it just isn’t listed. The 13-inch iPad Air (M4) has 12 GB of memory (per Apple), meeting CLIP STUDIO PAINT’s high-resolution recommendation of 6 GB or more. The 13-inch iPad Pro (M5) has 12 GB or 16 GB of memory (per Apple), meeting CLIP STUDIO PAINT’s high-resolution recommendation of 6 GB or more. CLIP STUDIO PAINT lists iPadOS 18 and iPadOS 26; iPadOS 27 isn’t mentioned	Open with these settings
A5, No pressure needed, Hover needed, Procreate	△ Actual size on the 11-inch iPad Air (M4) (borderline). A5: 148 × 210 mm. The 11-inch iPad Air (M4) holds a whole A5 page at actual size. Only 9.4 mm to spare (6% of the paper side), so it’s borderline. Other models that fit at actual size: 11-inch iPad Pro (M5), 13-inch iPad Air (M4), 13-inch iPad Pro (M5). Apple Pencil (USB-C) has no pressure sensing (Apple’s feature comparison). The iPad (A16) isn’t on Apple’s list of models with hover. Procreate 5.4.14 needs iPadOS 16.3 or later. Every current iPad runs iPadOS 27	Open with these settings

How to Choose a Numeric Keypad for Windows: Connection and Keys

Sat, 10 Apr 2021 00:00:00 GMT

Per the makers’ specs, if you also want = and BS keys for Excel, go wired. If Tab and 00 are enough, any connection type works. This guide compares keypads by the makers’ specs from ELECOM, Sanwa Supply and BUFFALO, and ends with picks for US readers on Amazon.com.

The spec guidance applies to any keypad. According to ELECOM’s product pages, the 2.4 GHz models plug a USB-A receiver into the PC, and the Bluetooth models use the PC’s own Bluetooth.

In this guide, typing feel, noise and speed weren’t tested on real hardware and aren’t judged; only the makers’ specs are compared.

Wired, 2.4 GHz or Bluetooth: choosing a connection

Per the makers’ specs, choose wired to avoid batteries, 2.4 GHz for a USB receiver that needs no PC Bluetooth, or Bluetooth for no receiver if the PC has it. Both wireless types are cable-free and run on batteries, and the wired models list no battery.

According to ELECOM’s product pages, the Bluetooth models need a PC with Bluetooth support, and a PC without it can use a separately sold Bluetooth adapter.

ELECOM’s wired and 2.4 GHz models all use a USB-A connector. Sanwa Supply and BUFFALO connectors weren’t checked.

Keypads by condition: what the table shows

Three terms first:

BS is Backspace, which deletes the last character.
Pantograph and scissor keys are the low, flat laptop-style switches that make a keypad slim. The Japanese models in the table are pantograph; the US pick is scissor and its maker calls it slim.
“NumLock handling stated” means the maker says how the keypad’s NumLock works with the PC, which matters if numbers don’t type.

To choose, pick the connection first, then the keys you need, then slim or not if you carry it, then whether NumLock handling should be stated. The table shows whether the makers’ specs list each feature.

The picks at the end come from Amazon.com listings checked in October 2026. The criteria were:

New, and sold by Amazon.com or the maker’s official store
Model number matches the maker’s specs
At least 100 ratings and a 4-star average
For the Japanese picks only: no warning from Sakura Checker, a fake-review checker
Windows listed in the maker’s supported OS (7 or higher for the wired US slot; Windows 11 for the Japanese slots). Tab and 00 keys are not required for the US slot | Connection | Tab and 00 | Slim (pantograph or scissor keys) | BS and = | NumLock handling stated | Connector | | :— | :— | :— | :— | :— | :— | | Wired USB | Listed | Listed | Listed | Listed | USB-A | | 2.4 GHz | Listed | Listed | Not listed | Listed | USB-A | | Bluetooth | Listed | Listed | Not listed | Listed | None (uses the PC’s Bluetooth) |

The connector column describes the ELECOM models.

In this guide, each column is judged on its own, so no single model necessarily meets every condition in a row.

Per the makers’ specs, for example, no slim wired model lists both BS and =. Use the finder below to check combinations.

In the calculator, pick a connection, the keys you want, whether it should be slim, and whether NumLock handling must be stated. It lists models whose official specs match your choices. The US picks are in the last section.

Find a numeric keypad by connection and keys

Pick a connection (wired USB, 2.4 GHz or Bluetooth), the keys you want (Tab, 00, BS, =), whether it should be slim, and whether NumLock handling must be stated. It lists the models whose official specs match and suggests one keypad you can buy.

Excel keys: Tab, 00, BS and =

On the makers’ product pages, BUFFALO describes Tab as handy for moving between Excel cells and selecting functions, and 00 for large amounts. ELECOM’s wired model with BS and = calls = and parentheses handy for functions, and 00 and BS handy for numbers.

In this guide, Tab moves between cells, and 00 helps enter large amounts. The = and parentheses keys help enter formulas. BS works as Backspace and deletes the last character.

Per the makers’ specs, Tab and 00 keys are listed on models of every connection type. Only wired models list both BS and = keys; no wireless model does.

Keys are judged by whether the maker lists them. A key missing from the list may still be on the keypad, so check the layout diagram.

According to ELECOM’s product pages, for the wired model with BS and = keys, the maker notes there’s no kana input mode. Special keys such as = and parentheses need romaji input.

The NumLock problem and what makers say

According to BUFFALO’s product pages, typing numbers on a USB keypad requires turning its NumLock on. On most laptops, that turns part of the built-in keyboard into number keys.

Per the makers’ specs, each maker describes its fix in its own words. ELECOM says some wired models solve it in hardware, and BUFFALO says its keypads are designed not to sync the PC’s NumLock with the keypad’s. Sanwa Supply calls its keypads NumLock-independent.

Thickness and weight

Per the makers’ specs, heights run 17–21.5 mm for pantograph models and 20–26.6 mm for membrane models, so membrane models can be slim too.

Weights run 70–143 g.

For carrying, look for a slim wireless keypad with pantograph keys. The makers list such models in both 2.4 GHz and Bluetooth.

Wireless battery life and paired devices

Per the makers’ specs, ELECOM states an estimated battery life of 10 months to 1.5 years with alkaline batteries. Sanwa Supply states 70 to 800 hours of continuous operation. The measures differ, so they can’t be compared.

Maker, connection and keys	Value and condition
ELECOM, 2.4 GHz, membrane keys, AA	1.3 years (estimated battery life, with alkaline batteries, maker’s own test)
ELECOM, 2.4 GHz, membrane keys, AAA	1.5 years (estimated battery life, with alkaline batteries, assuming 8 hours of PC use a day)
ELECOM, 2.4 GHz, pantograph keys	1 year (estimated battery life, with alkaline batteries, maker’s own test)
ELECOM, Bluetooth, membrane keys	10 months (estimated battery life, with alkaline batteries, assuming 8 hours of PC use a day)
ELECOM, Bluetooth, pantograph keys	1 year (estimated battery life, with alkaline batteries, maker’s own test)
Sanwa Supply, Bluetooth, Tab and 00	150 hours (continuous operating time)
Sanwa Supply, 2.4 GHz, Tab and 00	70 hours (continuous operating time, with alkaline batteries)
Sanwa Supply, Bluetooth, Tab and =	240 hours (continuous operating time)
Sanwa Supply, 2.4 GHz, Tab and =	800 hours (continuous operating time)

ELECOM’s Bluetooth models connect to one device only, and Sanwa Supply’s product pages give no device count. Microsoft’s support page says the Microsoft Number Pad pairs with up to 3 devices, but its availability wasn’t checked.

Picks for US readers

On Amazon.com as checked in October 2026, one wired slim keypad met the listing criteria: the Perixx PERIPAD-202C. Its maker lists Windows 7 or higher but not Windows 11. The maker’s page lists no Tab or 00 key; only the Amazon product title says Tab.

Wired slim keypad

It has scissor keys and two USB-A ports on the side.

Sold by Perixx USA and shipped by Amazon. Free delivery depends on Prime membership or order size. The Amazon listing is the USB-C model with two USB-A hub ports; the USB-A-only model is a separate listing.

The Japanese version picks ELECOM models sold in Japan. Check the specs above against any keypad before buying.

Sources: ELECOM product pages for TK-TCM011BK, TK-TCM015BK, TK-TCP018BK, TK-TDM017BK, TK-TDP019BK and TK-TBM023SKBK
Sources: Sanwa Supply product page and BUFFALO product page
Sources: Microsoft Number Pad support and Windows 2000 registry reference (legacy)

FAQ

Is wired, 2.4 GHz or Bluetooth better for a numeric keypad?

Per the makers’ specs, choose wired to avoid batteries, 2.4 GHz for a USB receiver that needs no PC Bluetooth, or Bluetooth for no receiver if the PC has it.

How long do wireless keypad batteries last, and how many devices can they pair with?

Per the makers’ specs, battery life is 10 months to 1.5 years by ELECOM’s estimate and 70 to 800 hours of continuous use for Sanwa Supply. Those measures can’t be compared, and each uses one AA or AAA battery.

ELECOM’s Bluetooth models connect to one device only.

Do you need Tab, 00, BS and = keys for Excel?

As of October 2026, per the makers’ specs, Tab moves between cells, and 00 helps enter large amounts. The = and parentheses keys help enter formulas, and BS deletes the last character.

Tab and 00 appear on every connection type, and only wired models list both BS and =.

What is the NumLock problem, and why won’t a keypad type numbers?

According to BUFFALO’s product pages, the keypad’s NumLock must be on to type numbers. On most laptops, that also turns part of the built-in keyboard into number keys.

On the makers’ product pages, some of ELECOM’s wired models and Sanwa Supply’s and BUFFALO’s keypads state how they handle it. Per a legacy Microsoft document (the Windows 2000 registry reference), Windows saves the NumLock state at logoff or shutdown and restores it at the next logon.

No current official page on how Windows 11 handles this was found, so this article doesn’t judge it.

Which keypad is best to carry with a laptop?

For carrying, look for a slim wireless keypad with pantograph keys, since the makers list such models in both 2.4 GHz and Bluetooth. The picks section has a slim wired keypad that met this guide’s Amazon.com criteria.

Do numeric keypads work with a Mac?

As of October 2026, per the makers’ specs, for a Mac, pick an ELECOM model that lists macOS. The supported-OS field shows macOS High Sierra 10.13 for pantograph models and macOS Big Sur 11 for models ending in SKBK.

Check that your macOS is on the list.

Data and formulas used (for verification)

Method

Covers current models whose official pages were read on October 8, 2026 (ELECOM, Sanwa Supply, BUFFALO), plus the Perixx PERIPAD-202C, whose maker page was read on October 11, 2026 for the US pick. Specs (connection, keys, key switch, size, weight, battery, supported OS, NumLock handling) come only from the makers’ official pages.
Keys are judged by whether the maker lists them on the product page (e.g. TK-TCM011BK: Tab and 00; TK-TCM015BK: Esc, Tab, =, (, ), 00 and BS keys, plus three hotkeys that launch Excel, a calculator app and a browser (2-port USB 2.0 hub); NT-27UBK: Esc, F2, F4, Tab, =, (, ), comma, copy, paste and find (no 00 or BS key listed)). A key that isn’t listed is not treated as missing.
Slim means a low-profile key switch that the maker lists: pantograph for the Japanese models, and scissor for the US pick, which the maker also calls a slim keypad (a grouping set for this article). Heights: pantograph 17–21.5 mm, membrane 20–26.6 mm (makers’ figures).
NumLock uses each maker’s own statement (ELECOM TK-TCM011BK: Solves the NumLock problem in hardware; no software needed; BUFFALO BSTK130BK: Designed so the keypad’s NumLock does not sync with the PC’s NumLock (maker statement); Sanwa Supply NT-BT21BK: NumLock-independent (maker statement)). For Windows itself, the only official source found is a legacy Microsoft Learn page (Windows 2000 registry reference): Windows stores the NumLock state at logoff/shutdown in HKCU\Control Panel\Keyboard\InitialKeyboardIndicators (0 = off, 2 = on) and restores it at the next logon. Current Windows 11 behavior isn’t judged.
Battery figures are the makers’ values with their stated conditions, not converted. Examples: TK-TDM017BK 1.3 years (estimated battery life, with alkaline batteries, maker’s own test), TK-TBM023SKBK 10 months (estimated battery life, with alkaline batteries, assuming 8 hours of PC use a day), NT-WL21BK 70 hours (continuous operating time, with alkaline batteries), NT-BT21BK 150 hours (continuous operating time). Bluetooth device counts: ELECOM TK-TBM023SKBK One device at a time (paired); Microsoft Number Pad 3 devices.
Picks are models that match the conditions, then new listings sold by Amazon (or a confirmed maker store) with 100+ ratings, 4+ stars and a model number matching the maker’s, cheapest including shipping. The Japanese slots also require Windows 11 in the maker’s supported OS list and no Sakura Checker warning (Japan listings only). The US slot is a wired slim keypad whose maker lists Windows 7 or higher; no Tab or 00 key is required. The shipping cost of the US pick is treated as 0 because Amazon ships it, and the non-Prime shipping cost on a small order wasn’t checked. Typing feel, noise and speed weren’t tested on real hardware and aren’t judged.

Official values used

Item	Value	Source	Checked
ELECOM TK-TCM011BK connection	(stated in a Japanese-language source)	elecom.co.jp	2026-10-08
ELECOM TK-TCM011BK key count	20 keys	elecom.co.jp	2026-10-08
ELECOM TK-TCM011BK extra keys the maker lists	(stated in a Japanese-language source)	elecom.co.jp	2026-10-08
ELECOM TK-TCM011BK key switch	(stated in a Japanese-language source)	elecom.co.jp	2026-10-08
ELECOM TK-TCM011BK width	90 mm	elecom.co.jp	2026-10-08
ELECOM TK-TCM011BK depth	140 mm	elecom.co.jp	2026-10-08
ELECOM TK-TCM011BK height	23.5 mm	elecom.co.jp	2026-10-08
ELECOM TK-TCM011BK weight	110 g	elecom.co.jp	2026-10-08
ELECOM TK-TCM011BK supported OS	Windows 11/10/8/8.1/RT/RT8.1/7(SP1)/Vista(SP2)/XP(SP3)	elecom.co.jp	2026-10-08
ELECOM TK-TCM011BK connector	USB-A	elecom.co.jp	2026-10-08
ELECOM TK-TCM011BK NumLock handling (maker statement)	(stated in a Japanese-language source)	elecom.co.jp	2026-10-08
ELECOM TK-TCM012BK connection	(stated in a Japanese-language source)	elecom.co.jp	2026-10-08
ELECOM TK-TCM012BK key count	20 keys	elecom.co.jp	2026-10-08
ELECOM TK-TCM012BK extra keys the maker lists	(stated in a Japanese-language source)	elecom.co.jp	2026-10-08
ELECOM TK-TCM012BK key switch	(stated in a Japanese-language source)	elecom.co.jp	2026-10-08
ELECOM TK-TCM012BK width	90 mm	elecom.co.jp	2026-10-08
ELECOM TK-TCM012BK depth	140 mm	elecom.co.jp	2026-10-08
ELECOM TK-TCM012BK height	23.5 mm	elecom.co.jp	2026-10-08
ELECOM TK-TCM012BK weight	114 g	elecom.co.jp	2026-10-08
ELECOM TK-TCM012BK supported OS	Windows 11/10/8/8.1/RT/RT8.1/7(SP1)/Vista(SP2)/XP(SP3)	elecom.co.jp	2026-10-08
ELECOM TK-TCM012BK connector	USB-A	elecom.co.jp	2026-10-08
ELECOM TK-TCM012BK NumLock handling (maker statement)	(stated in a Japanese-language source)	elecom.co.jp	2026-10-08
ELECOM TK-TCM013BK connection	(stated in a Japanese-language source)	elecom.co.jp	2026-10-08
ELECOM TK-TCM013BK key count	20 keys	elecom.co.jp	2026-10-08
ELECOM TK-TCM013BK extra keys the maker lists	(stated in a Japanese-language source)	elecom.co.jp	2026-10-08
ELECOM TK-TCM013BK key switch	(stated in a Japanese-language source)	elecom.co.jp	2026-10-08
ELECOM TK-TCM013BK width	78 mm	elecom.co.jp	2026-10-08
ELECOM TK-TCM013BK depth	127.5 mm	elecom.co.jp	2026-10-08
ELECOM TK-TCM013BK height	24 mm	elecom.co.jp	2026-10-08
ELECOM TK-TCM013BK weight	97 g	elecom.co.jp	2026-10-08
ELECOM TK-TCM013BK supported OS	Windows 11/10/8.1/RT8.1/7(SP1)/Vista(SP2)	elecom.co.jp	2026-10-08
ELECOM TK-TCM013BK connector	USB-A	elecom.co.jp	2026-10-08
ELECOM TK-TCM014BK connection	(stated in a Japanese-language source)	elecom.co.jp	2026-10-08
ELECOM TK-TCM014BK key count	20 keys	elecom.co.jp	2026-10-08
ELECOM TK-TCM014BK extra keys the maker lists	(stated in a Japanese-language source)	elecom.co.jp	2026-10-08
ELECOM TK-TCM014BK key switch	(stated in a Japanese-language source)	elecom.co.jp	2026-10-08
ELECOM TK-TCM014BK width	78 mm	elecom.co.jp	2026-10-08
ELECOM TK-TCM014BK depth	127.5 mm	elecom.co.jp	2026-10-08
ELECOM TK-TCM014BK height	24 mm	elecom.co.jp	2026-10-08
ELECOM TK-TCM014BK weight	97 g	elecom.co.jp	2026-10-08
ELECOM TK-TCM014BK supported OS	Windows 11/10/8.1/RT8.1/7(SP1)/Vista(SP2)	elecom.co.jp	2026-10-08
ELECOM TK-TCM014BK connector	USB-A	elecom.co.jp	2026-10-08
ELECOM TK-TCM015BK connection	(stated in a Japanese-language source)	elecom.co.jp	2026-10-08
ELECOM TK-TCM015BK key count	27 keys	elecom.co.jp	2026-10-08
ELECOM TK-TCM015BK extra keys the maker lists	(stated in a Japanese-language source)	elecom.co.jp	2026-10-08
ELECOM TK-TCM015BK key switch	(stated in a Japanese-language source)	elecom.co.jp	2026-10-08
ELECOM TK-TCM015BK width	90 mm	elecom.co.jp	2026-10-08
ELECOM TK-TCM015BK depth	170 mm	elecom.co.jp	2026-10-08
ELECOM TK-TCM015BK height	24.5 mm	elecom.co.jp	2026-10-08
ELECOM TK-TCM015BK weight	143 g	elecom.co.jp	2026-10-08
ELECOM TK-TCM015BK supported OS	(stated in a Japanese-language source)	elecom.co.jp	2026-10-08
ELECOM TK-TCM015BK connector	USB-A	elecom.co.jp	2026-10-08
ELECOM TK-TCM015BK input-mode note	(stated in a Japanese-language source)	elecom.co.jp	2026-10-08
ELECOM TK-TCM021SKBK connection	(stated in a Japanese-language source)	elecom.co.jp	2026-10-08
ELECOM TK-TCM021SKBK key count	20 keys	elecom.co.jp	2026-10-08
ELECOM TK-TCM021SKBK extra keys the maker lists	(stated in a Japanese-language source)	elecom.co.jp	2026-10-08
ELECOM TK-TCM021SKBK key switch	(stated in a Japanese-language source)	elecom.co.jp	2026-10-08
ELECOM TK-TCM021SKBK width	89.3 mm	elecom.co.jp	2026-10-08
ELECOM TK-TCM021SKBK depth	135.3 mm	elecom.co.jp	2026-10-08
ELECOM TK-TCM021SKBK height	26.6 mm	elecom.co.jp	2026-10-08
ELECOM TK-TCM021SKBK weight	90 g	elecom.co.jp	2026-10-08
ELECOM TK-TCM021SKBK supported OS	(stated in a Japanese-language source)	elecom.co.jp	2026-10-08
ELECOM TK-TCM021SKBK connector	USB-A	elecom.co.jp	2026-10-08
ELECOM TK-TCP018BK connection	(stated in a Japanese-language source)	elecom.co.jp	2026-10-08
ELECOM TK-TCP018BK key count	20 keys	elecom.co.jp	2026-10-08
ELECOM TK-TCP018BK extra keys the maker lists	(stated in a Japanese-language source)	elecom.co.jp	2026-10-08
ELECOM TK-TCP018BK key switch	(stated in a Japanese-language source)	elecom.co.jp	2026-10-08
ELECOM TK-TCP018BK width	89.6 mm	elecom.co.jp	2026-10-08
ELECOM TK-TCP018BK depth	127.5 mm	elecom.co.jp	2026-10-08
ELECOM TK-TCP018BK height	21.5 mm	elecom.co.jp	2026-10-08
ELECOM TK-TCP018BK weight	110 g	elecom.co.jp	2026-10-08
ELECOM TK-TCP018BK supported OS	(stated in a Japanese-language source)	elecom.co.jp	2026-10-08
ELECOM TK-TCP018BK connector	USB-A	elecom.co.jp	2026-10-08
ELECOM TK-TDM017BK connection	(stated in a Japanese-language source)	elecom.co.jp	2026-10-08
ELECOM TK-TDM017BK key count	20 keys	elecom.co.jp	2026-10-08
ELECOM TK-TDM017BK extra keys the maker lists	(stated in a Japanese-language source)	elecom.co.jp	2026-10-08
ELECOM TK-TDM017BK key switch	(stated in a Japanese-language source)	elecom.co.jp	2026-10-08
ELECOM TK-TDM017BK width	90 mm	elecom.co.jp	2026-10-08
ELECOM TK-TDM017BK depth	140 mm	elecom.co.jp	2026-10-08
ELECOM TK-TDM017BK height	26.5 mm	elecom.co.jp	2026-10-08
ELECOM TK-TDM017BK weight	107 g	elecom.co.jp	2026-10-08
ELECOM TK-TDM017BK battery life (maker figure)	1.3 years	elecom.co.jp	2026-10-08
ELECOM TK-TDM017BK battery type	(stated in a Japanese-language source)	elecom.co.jp	2026-10-08
ELECOM TK-TDM017BK supported OS	(stated in a Japanese-language source)	elecom.co.jp	2026-10-08
ELECOM TK-TDM017BK connector (receiver)	USB-A	elecom.co.jp	2026-10-08
ELECOM TK-TDM022SKBK connection	(stated in a Japanese-language source)	elecom.co.jp	2026-10-08
ELECOM TK-TDM022SKBK key count	20 keys	elecom.co.jp	2026-10-08
ELECOM TK-TDM022SKBK extra keys the maker lists	(stated in a Japanese-language source)	elecom.co.jp	2026-10-08
ELECOM TK-TDM022SKBK key switch	(stated in a Japanese-language source)	elecom.co.jp	2026-10-08
ELECOM TK-TDM022SKBK width	89.3 mm	elecom.co.jp	2026-10-08
ELECOM TK-TDM022SKBK depth	135.3 mm	elecom.co.jp	2026-10-08
ELECOM TK-TDM022SKBK height	26.6 mm	elecom.co.jp	2026-10-08
ELECOM TK-TDM022SKBK weight	80 g	elecom.co.jp	2026-10-08
ELECOM TK-TDM022SKBK battery life (maker figure)	1.5 years	elecom.co.jp	2026-10-08
ELECOM TK-TDM022SKBK battery type	(stated in a Japanese-language source)	elecom.co.jp	2026-10-08
ELECOM TK-TDM022SKBK supported OS	(stated in a Japanese-language source)	elecom.co.jp	2026-10-08
ELECOM TK-TDM022SKBK connector (receiver)	USB-A	elecom.co.jp	2026-10-08
ELECOM TK-TDP019BK connection	(stated in a Japanese-language source)	elecom.co.jp	2026-10-08
ELECOM TK-TDP019BK key count	20 keys	elecom.co.jp	2026-10-08
ELECOM TK-TDP019BK extra keys the maker lists	(stated in a Japanese-language source)	elecom.co.jp	2026-10-08
ELECOM TK-TDP019BK key switch	(stated in a Japanese-language source)	elecom.co.jp	2026-10-08
ELECOM TK-TDP019BK width	89.6 mm	elecom.co.jp	2026-10-08
ELECOM TK-TDP019BK depth	127.5 mm	elecom.co.jp	2026-10-08
ELECOM TK-TDP019BK height	21.5 mm	elecom.co.jp	2026-10-08
ELECOM TK-TDP019BK weight	100 g	elecom.co.jp	2026-10-08
ELECOM TK-TDP019BK battery life (maker figure)	1 year	elecom.co.jp	2026-10-08
ELECOM TK-TDP019BK battery type	(stated in a Japanese-language source)	elecom.co.jp	2026-10-08
ELECOM TK-TDP019BK supported OS	(stated in a Japanese-language source)	elecom.co.jp	2026-10-08
ELECOM TK-TDP019BK connector (receiver)	USB-A	elecom.co.jp	2026-10-08
ELECOM TK-TBM023SKBK connection	Bluetooth	elecom.co.jp	2026-10-08
ELECOM TK-TBM023SKBK key count	20 keys	elecom.co.jp	2026-10-08
ELECOM TK-TBM023SKBK extra keys the maker lists	(stated in a Japanese-language source)	elecom.co.jp	2026-10-08
ELECOM TK-TBM023SKBK key switch	(stated in a Japanese-language source)	elecom.co.jp	2026-10-08
ELECOM TK-TBM023SKBK width	89.3 mm	elecom.co.jp	2026-10-08
ELECOM TK-TBM023SKBK depth	135.3 mm	elecom.co.jp	2026-10-08
ELECOM TK-TBM023SKBK height	26.6 mm	elecom.co.jp	2026-10-08
ELECOM TK-TBM023SKBK weight	80 g	elecom.co.jp	2026-10-08
ELECOM TK-TBM023SKBK battery life (maker figure)	10 months	elecom.co.jp	2026-10-08
ELECOM TK-TBM023SKBK battery type	(stated in a Japanese-language source)	elecom.co.jp	2026-10-08
ELECOM TK-TBM023SKBK supported OS	(stated in a Japanese-language source)	elecom.co.jp	2026-10-08
ELECOM TK-TBM023SKBK supported hosts	(stated in a Japanese-language source)	elecom.co.jp	2026-10-08
ELECOM TK-TBM023SKBK devices it can connect to	(stated in a Japanese-language source)	elecom.co.jp	2026-10-08
ELECOM TK-TBP020BK connection	Bluetooth	elecom.co.jp	2026-10-08
ELECOM TK-TBP020BK key count	20 keys	elecom.co.jp	2026-10-08
ELECOM TK-TBP020BK extra keys the maker lists	(stated in a Japanese-language source)	elecom.co.jp	2026-10-08
ELECOM TK-TBP020BK key switch	(stated in a Japanese-language source)	elecom.co.jp	2026-10-08
ELECOM TK-TBP020BK width	89.6 mm	elecom.co.jp	2026-10-08
ELECOM TK-TBP020BK depth	127.5 mm	elecom.co.jp	2026-10-08
ELECOM TK-TBP020BK height	21.5 mm	elecom.co.jp	2026-10-08
ELECOM TK-TBP020BK weight	100 g	elecom.co.jp	2026-10-08
ELECOM TK-TBP020BK battery life (maker figure)	1 year	elecom.co.jp	2026-10-08
ELECOM TK-TBP020BK battery type	(stated in a Japanese-language source)	elecom.co.jp	2026-10-08
ELECOM TK-TBP020BK supported OS	(stated in a Japanese-language source)	elecom.co.jp	2026-10-08
ELECOM TK-TBP020BK supported hosts	(stated in a Japanese-language source)	elecom.co.jp	2026-10-08
ELECOM TK-TBP020BK devices it can connect to	(stated in a Japanese-language source)	elecom.co.jp	2026-10-08
Sanwa Supply NT-BT21BK connection	Bluetooth	sanwa.co.jp	2026-10-08
Sanwa Supply NT-BT21BK key count	20 keys	sanwa.co.jp	2026-10-08
Sanwa Supply NT-BT21BK extra keys the maker lists	(stated in a Japanese-language source)	sanwa.co.jp	2026-10-08
Sanwa Supply NT-BT21BK key switch	(stated in a Japanese-language source)	sanwa.co.jp	2026-10-08
Sanwa Supply NT-BT21BK width	86 mm	sanwa.co.jp	2026-10-08
Sanwa Supply NT-BT21BK depth	130 mm	sanwa.co.jp	2026-10-08
Sanwa Supply NT-BT21BK height	17 mm	sanwa.co.jp	2026-10-08
Sanwa Supply NT-BT21BK weight	70 g	sanwa.co.jp	2026-10-08
Sanwa Supply NT-BT21BK battery life (maker figure)	150 hours	sanwa.co.jp	2026-10-08
Sanwa Supply NT-BT21BK battery type	(stated in a Japanese-language source)	sanwa.co.jp	2026-10-08
Sanwa Supply NT-BT21BK supported OS	(stated in a Japanese-language source)	sanwa.co.jp	2026-10-08
Sanwa Supply NT-BT21BK NumLock handling (maker statement)	(stated in a Japanese-language source)	sanwa.co.jp	2026-10-08
Sanwa Supply NT-BT23BK connection	Bluetooth	sanwa.co.jp	2026-10-08
Sanwa Supply NT-BT23BK key count	20 keys	sanwa.co.jp	2026-10-08
Sanwa Supply NT-BT23BK extra keys the maker lists	(stated in a Japanese-language source)	sanwa.co.jp	2026-10-08
Sanwa Supply NT-BT23BK key switch	(stated in a Japanese-language source)	sanwa.co.jp	2026-10-08
Sanwa Supply NT-BT23BK width	86 mm	sanwa.co.jp	2026-10-08
Sanwa Supply NT-BT23BK depth	130 mm	sanwa.co.jp	2026-10-08
Sanwa Supply NT-BT23BK height	17 mm	sanwa.co.jp	2026-10-08
Sanwa Supply NT-BT23BK weight	70 g	sanwa.co.jp	2026-10-08
Sanwa Supply NT-BT23BK battery life (maker figure)	150 hours	sanwa.co.jp	2026-10-08
Sanwa Supply NT-BT23BK battery type	(stated in a Japanese-language source)	sanwa.co.jp	2026-10-08
Sanwa Supply NT-BT23BK supported OS	(stated in a Japanese-language source)	sanwa.co.jp	2026-10-08
Sanwa Supply NT-BT23BK NumLock handling (maker statement)	(stated in a Japanese-language source)	sanwa.co.jp	2026-10-08
Sanwa Supply NT-WL21BK connection	(stated in a Japanese-language source)	sanwa.co.jp	2026-10-08
Sanwa Supply NT-WL21BK key count	20 keys	sanwa.co.jp	2026-10-08
Sanwa Supply NT-WL21BK extra keys the maker lists	(stated in a Japanese-language source)	sanwa.co.jp	2026-10-08
Sanwa Supply NT-WL21BK key switch	(stated in a Japanese-language source)	sanwa.co.jp	2026-10-08
Sanwa Supply NT-WL21BK width	86 mm	sanwa.co.jp	2026-10-08
Sanwa Supply NT-WL21BK depth	130 mm	sanwa.co.jp	2026-10-08
Sanwa Supply NT-WL21BK height	17 mm	sanwa.co.jp	2026-10-08
Sanwa Supply NT-WL21BK weight	70 g	sanwa.co.jp	2026-10-08
Sanwa Supply NT-WL21BK battery life (maker figure)	70 hours	sanwa.co.jp	2026-10-08
Sanwa Supply NT-WL21BK battery type	(stated in a Japanese-language source)	sanwa.co.jp	2026-10-08
Sanwa Supply NT-WL21BK supported OS	(stated in a Japanese-language source)	sanwa.co.jp	2026-10-08
Sanwa Supply NT-WL21BK NumLock handling (maker statement)	(stated in a Japanese-language source)	sanwa.co.jp	2026-10-08
Sanwa Supply NT-WL23BK connection	(stated in a Japanese-language source)	sanwa.co.jp	2026-10-08
Sanwa Supply NT-WL23BK key count	20 keys	sanwa.co.jp	2026-10-08
Sanwa Supply NT-WL23BK extra keys the maker lists	(stated in a Japanese-language source)	sanwa.co.jp	2026-10-08
Sanwa Supply NT-WL23BK key switch	(stated in a Japanese-language source)	sanwa.co.jp	2026-10-08
Sanwa Supply NT-WL23BK width	86 mm	sanwa.co.jp	2026-10-08
Sanwa Supply NT-WL23BK depth	130 mm	sanwa.co.jp	2026-10-08
Sanwa Supply NT-WL23BK height	17 mm	sanwa.co.jp	2026-10-08
Sanwa Supply NT-WL23BK weight	70 g	sanwa.co.jp	2026-10-08
Sanwa Supply NT-WL23BK battery life (maker figure)	70 hours	sanwa.co.jp	2026-10-08
Sanwa Supply NT-WL23BK battery type	(stated in a Japanese-language source)	sanwa.co.jp	2026-10-08
Sanwa Supply NT-WL23BK supported OS	(stated in a Japanese-language source)	sanwa.co.jp	2026-10-08
Sanwa Supply NT-WL23BK NumLock handling (maker statement)	(stated in a Japanese-language source)	sanwa.co.jp	2026-10-08
Sanwa Supply NT-BT27BK connection	Bluetooth	sanwa.co.jp	2026-10-08
Sanwa Supply NT-BT27BK key count	39 keys	sanwa.co.jp	2026-10-08
Sanwa Supply NT-BT27BK extra keys the maker lists	(stated in a Japanese-language source)	sanwa.co.jp	2026-10-08
Sanwa Supply NT-BT27BK key switch	(stated in a Japanese-language source)	sanwa.co.jp	2026-10-08
Sanwa Supply NT-BT27BK width	147.7 mm	sanwa.co.jp	2026-10-08
Sanwa Supply NT-BT27BK depth	117.8 mm	sanwa.co.jp	2026-10-08
Sanwa Supply NT-BT27BK height	18 mm	sanwa.co.jp	2026-10-08
Sanwa Supply NT-BT27BK weight	95 g	sanwa.co.jp	2026-10-08
Sanwa Supply NT-BT27BK battery life (maker figure)	240 hours	sanwa.co.jp	2026-10-08
Sanwa Supply NT-BT27BK battery type	(stated in a Japanese-language source)	sanwa.co.jp	2026-10-08
Sanwa Supply NT-BT27BK supported OS	(stated in a Japanese-language source)	sanwa.co.jp	2026-10-08
Sanwa Supply NT-BT27BK NumLock handling (maker statement)	(stated in a Japanese-language source)	sanwa.co.jp	2026-10-08
Sanwa Supply NT-WL27BK connection	(stated in a Japanese-language source)	sanwa.co.jp	2026-10-08
Sanwa Supply NT-WL27BK key count	39 keys	sanwa.co.jp	2026-10-08
Sanwa Supply NT-WL27BK extra keys the maker lists	(stated in a Japanese-language source)	sanwa.co.jp	2026-10-08
Sanwa Supply NT-WL27BK key switch	(stated in a Japanese-language source)	sanwa.co.jp	2026-10-08
Sanwa Supply NT-WL27BK width	147.7 mm	sanwa.co.jp	2026-10-08
Sanwa Supply NT-WL27BK depth	117.8 mm	sanwa.co.jp	2026-10-08
Sanwa Supply NT-WL27BK height	18 mm	sanwa.co.jp	2026-10-08
Sanwa Supply NT-WL27BK weight	110 g	sanwa.co.jp	2026-10-08
Sanwa Supply NT-WL27BK battery life (maker figure)	800 hours	sanwa.co.jp	2026-10-08
Sanwa Supply NT-WL27BK battery type	(stated in a Japanese-language source)	sanwa.co.jp	2026-10-08
Sanwa Supply NT-WL27BK supported OS	(stated in a Japanese-language source)	sanwa.co.jp	2026-10-08
Sanwa Supply NT-WL27BK NumLock handling (maker statement)	(stated in a Japanese-language source)	sanwa.co.jp	2026-10-08
Sanwa Supply NT-27UBK connection	(stated in a Japanese-language source)	sanwa.co.jp	2026-10-08
Sanwa Supply NT-27UBK key count	39 keys	sanwa.co.jp	2026-10-08
Sanwa Supply NT-27UBK extra keys the maker lists	(stated in a Japanese-language source)	sanwa.co.jp	2026-10-08
Sanwa Supply NT-27UBK key switch	(stated in a Japanese-language source)	sanwa.co.jp	2026-10-08
Sanwa Supply NT-27UBK width	147.7 mm	sanwa.co.jp	2026-10-08
Sanwa Supply NT-27UBK depth	117.8 mm	sanwa.co.jp	2026-10-08
Sanwa Supply NT-27UBK height	18 mm	sanwa.co.jp	2026-10-08
Sanwa Supply NT-27UBK weight	110 g	sanwa.co.jp	2026-10-08
Sanwa Supply NT-27UBK supported OS	(stated in a Japanese-language source)	sanwa.co.jp	2026-10-08
Sanwa Supply NT-27UBK NumLock handling (maker statement)	(stated in a Japanese-language source)	sanwa.co.jp	2026-10-08
BUFFALO BSTK130BK connection	(stated in a Japanese-language source)	buffalo.jp	2026-10-08
BUFFALO BSTK130BK key count	20 keys	buffalo.jp	2026-10-08
BUFFALO BSTK130BK extra keys the maker lists	(stated in a Japanese-language source)	buffalo.jp	2026-10-08
BUFFALO BSTK130BK key switch	(stated in a Japanese-language source)	buffalo.jp	2026-10-08
BUFFALO BSTK130BK width	90 mm	buffalo.jp	2026-10-08
BUFFALO BSTK130BK depth	130 mm	buffalo.jp	2026-10-08
BUFFALO BSTK130BK height	20 mm	buffalo.jp	2026-10-08
BUFFALO BSTK130BK weight	115 g	buffalo.jp	2026-10-08
BUFFALO BSTK130BK supported OS	(stated in a Japanese-language source)	buffalo.jp	2026-10-08
BUFFALO BSTK130BK NumLock handling (maker statement)	(stated in a Japanese-language source)	buffalo.jp	2026-10-08
BUFFALO’s description of the NumLock problem	(stated in a Japanese-language source)	buffalo.jp	2026-10-08
BUFFALO BSTK135BK connection	(stated in a Japanese-language source)	buffalo.jp	2026-10-08
BUFFALO BSTK135BK key count	20 keys	buffalo.jp	2026-10-08
BUFFALO BSTK135BK extra keys the maker lists	(stated in a Japanese-language source)	buffalo.jp	2026-10-08
BUFFALO BSTK135BK key switch	(stated in a Japanese-language source)	buffalo.jp	2026-10-08
BUFFALO BSTK135BK width	90 mm	buffalo.jp	2026-10-08
BUFFALO BSTK135BK depth	130 mm	buffalo.jp	2026-10-08
BUFFALO BSTK135BK height	20 mm	buffalo.jp	2026-10-08
BUFFALO BSTK135BK weight	115 g	buffalo.jp	2026-10-08
BUFFALO BSTK135BK supported OS	(stated in a Japanese-language source)	buffalo.jp	2026-10-08
BUFFALO BSTK135BK NumLock handling (maker statement)	(stated in a Japanese-language source)	buffalo.jp	2026-10-08
BUFFALO BSTK120BKZ connection	(stated in a Japanese-language source)	buffalo.jp	2026-10-08
BUFFALO BSTK120BKZ key count	19 keys	buffalo.jp	2026-10-08
BUFFALO BSTK120BKZ extra keys the maker lists	(stated in a Japanese-language source)	buffalo.jp	2026-10-08
BUFFALO BSTK120BKZ key switch	(stated in a Japanese-language source)	buffalo.jp	2026-10-08
BUFFALO BSTK120BKZ width	90 mm	buffalo.jp	2026-10-08
BUFFALO BSTK120BKZ depth	145 mm	buffalo.jp	2026-10-08
BUFFALO BSTK120BKZ height	24 mm	buffalo.jp	2026-10-08
BUFFALO BSTK120BKZ weight	110 g	buffalo.jp	2026-10-08
BUFFALO BSTK120BKZ supported OS	(stated in a Japanese-language source)	buffalo.jp	2026-10-08
BUFFALO BSTK120BKZ NumLock handling (maker statement)	(stated in a Japanese-language source)	buffalo.jp	2026-10-08
Microsoft Number Pad connection	(stated in a Japanese-language source)	microsoft.com Bluetooth 5.0 connection	2026-10-08
Microsoft Number Pad devices it can pair with	3 devices	support.microsoft.com You can pair your number pad with up to three devices	2026-10-08
Microsoft Number Pad battery life (maker figure)	up to 24 months	microsoft.com long battery life up to 24 months (battery life may vary)	2026-10-08
Microsoft Number Pad battery type	(stated in a Japanese-language source)	support.microsoft.com Your number pad uses one coin-cell (CR2032) battery	2026-10-08
Microsoft Number Pad supported OS	(stated in a Japanese-language source)	support.microsoft.com a PC that’s running Windows 11 or Windows 10 with support for Bluetooth 4.0 or l	2026-10-08
Microsoft Number Pad calculator key	(stated in a Japanese-language source)	support.microsoft.com Press the Calculator key to open the Calculator app on your Windows PC	2026-10-08
Satechi Bluetooth Extended Keypad connection	Bluetooth	satechi.com Tech Specs: Connector Type Bluetooth	2026-10-08
Satechi Bluetooth Extended Keypad size and weight (as the maker lists them)	(stated in a Japanese-language source)	satechi.com Tech Specs: Dimensions & Weight 5.8 x 4.5 x 0.4 inches, 4.4 oz	2026-10-08
Satechi Bluetooth Extended Keypad supported OS	(stated in a Japanese-language source)	satechi.com Compatibility: Windows list; Important: Does not support Windows 7 or 8, please	2026-10-08
Perixx PERIPAD-202 connection	(stated in a Japanese-language source)	perixx.com Specs & Detail: Connection: USB-A cable, USB-C cable	2026-10-08
Perixx PERIPAD-202 key switch	(stated in a Japanese-language source)	perixx.com Switch type: Scissor; X-type scissor keys with non-synchronized function	2026-10-08
Perixx PERIPAD-202 supported OS	(stated in a Japanese-language source)	perixx.com Compatibility: Fully compatible: Windows 7 or higher, MacOS Monterey 12.2 or hig	2026-10-08
Kensington Slim Type Wired Numeric Keypad key count	21 keys	kensington.com 21-Key Number Pad with Four Shortcut Keys	2026-10-08
Kensington Slim Type Wired Numeric Keypad connection	(stated in a Japanese-language source)	kensington.com Plug-and-Play Wired USB-A Connection	2026-10-08
Kensington Slim Type Wired Numeric Keypad supported OS	(stated in a Japanese-language source)	kensington.com Compatible with Windows, macOS, and ChromeOS (some shortcut keys are only Window	2026-10-08
Windows startup NumLock official description (legacy reference)	(stated in a Japanese-language source)	learn.microsoft.com InitialKeyboardIndicators: Determines whether the NUMLOCK key is on or off when	2026-10-08
BUFFALO’s description of the Tab and 00 keys	(stated in a Japanese-language source)	buffalo.jp	2026-10-08
ELECOM TK-TCM015BK key descriptions	(stated in a Japanese-language source)	elecom.co.jp	2026-10-08
Satechi Slim Aluminum Bluetooth Keypad (ST-SALKPM) connection	Bluetooth	satechi.com Tech Specs: Connector Type Bluetooth	2026-10-11
Satechi Slim Aluminum Bluetooth Keypad size and weight (as the maker lists them)	(stated in a Japanese-language source)	satechi.com Tech Specs: Dimensions & Weight 4.5 x 3.125 x 0.31 inches, 3.5 ounces	2026-10-11
Satechi Slim Aluminum Bluetooth Keypad supported OS	(stated in a Japanese-language source)	satechi.com Compatibility: Windows list; Important: Does not support Windows 7 or 8, please	2026-10-11
Satechi Slim Aluminum Bluetooth Keypad extra keys the maker lists	(stated in a Japanese-language source)	satechi.com	2026-10-11
Satechi Slim Aluminum Bluetooth Keypad thinness	(stated in a Japanese-language source)	satechi.com	2026-10-11
Perixx PERIPAD-202C (USB-C, two-port hub) connection	(stated in a Japanese-language source)	perixx.com Connection type: USB Type-C / USB hub: 2 × USB Type-A ports（SKU PPD-202C-12060）	2026-10-11
Perixx PERIPAD-202C size and weight (as the maker lists them)	(stated in a Japanese-language source)	perixx.com Specs: Dimensions 131 × 87 × 20 mm, Weight 100 g	2026-10-11
Perixx PERIPAD-202C key switch	(stated in a Japanese-language source)	perixx.com Specs: Switch type: Scissor	2026-10-11
Perixx PERIPAD-202C supported OS	(stated in a Japanese-language source)	perixx.com Compatibility: Windows 7 or higher / MacOS Monterey 12.2 or higher	2026-10-11
Perixx PERIPAD-202C extra keys the maker lists	(stated in a Japanese-language source)	perixx.com	2026-10-11
Perixx PERIPAD-202C thinness	(stated in a Japanese-language source)	perixx.com Keypad Design: 19 keys full-size slim keypad; 131 × 87 × 20 mm	2026-10-11

Find a numeric keypad by connection and keys: quick reference

The calculator in this article shows the result for your own settings. Settings can be set in the URL; each link below opens the calculator with those settings.

Settings	Result	Calculator
Wired USB, Tab, 00, Any thickness, NumLock doesn’t matter	✓ 9 models match. ELECOM TK-TCM011BK, ELECOM TK-TCM012BK, ELECOM TK-TCM013BK, ELECOM TK-TCM015BK, ELECOM TK-TCM021SKBK, ELECOM TK-TCP018BK, BUFFALO BSTK130BK, BUFFALO BSTK135BK, BUFFALO BSTK120BKZ. ELECOM TK-TCM011BK: wired USB, listed keys Tab, 00, membrane, 23.5 mm tall, 110 g. NumLock (ELECOM TK-TCM011BK): Solves the NumLock problem in hardware; no software needed. ELECOM TK-TCM012BK: wired USB, listed keys Tab, 00, membrane, 23.5 mm tall, 114 g. NumLock (ELECOM TK-TCM012BK): Solves the NumLock problem in hardware; no software needed. ELECOM TK-TCM013BK: wired USB, listed keys Tab, 00, membrane, 24 mm tall, 97 g. Other matching models: TK-TCM015BK, TK-TCM021SKBK, TK-TCP018BK, BSTK130BK, BSTK135BK, BSTK120BKZ. No Windows 11 listing (not used for picks): BSTK130BK, BSTK135BK, BSTK120BKZ. Keys are judged by whether the maker lists them. A key that isn’t listed may still be there, so check the layout diagram	Open with these settings
Wired USB, Tab, 00, Any thickness, NumLock handling stated	✓ 5 models match. ELECOM TK-TCM011BK, ELECOM TK-TCM012BK, BUFFALO BSTK130BK, BUFFALO BSTK135BK, BUFFALO BSTK120BKZ. ELECOM TK-TCM011BK: wired USB, listed keys Tab, 00, membrane, 23.5 mm tall, 110 g. NumLock (ELECOM TK-TCM011BK): Solves the NumLock problem in hardware; no software needed. ELECOM TK-TCM012BK: wired USB, listed keys Tab, 00, membrane, 23.5 mm tall, 114 g. NumLock (ELECOM TK-TCM012BK): Solves the NumLock problem in hardware; no software needed. BUFFALO BSTK130BK: wired USB, listed keys Tab, 00, membrane, 20 mm tall, 115 g. NumLock (BUFFALO BSTK130BK): Designed so the keypad’s NumLock does not sync with the PC’s NumLock (maker statement). Other matching models: BSTK135BK, BSTK120BKZ. No Windows 11 listing (not used for picks): BSTK130BK, BSTK135BK, BSTK120BKZ. Keys are judged by whether the maker lists them. A key that isn’t listed may still be there, so check the layout diagram. NumLock handling is quoted from each maker. Models that don’t state it aren’t judged	Open with these settings
Wired USB, Tab, 00, Slim (pantograph keys), NumLock doesn’t matter	✓ 1 model matches. ELECOM TK-TCP018BK. ELECOM TK-TCP018BK: wired USB, listed keys Tab, 00, pantograph, 21.5 mm tall, 110 g. Keys are judged by whether the maker lists them. A key that isn’t listed may still be there, so check the layout diagram. Slim means the maker lists a pantograph key switch (this article’s grouping)	Open with these settings
Wired USB, Tab, 00, BS, =, Any thickness, NumLock doesn’t matter	✓ 1 model matches. ELECOM TK-TCM015BK. ELECOM TK-TCM015BK: wired USB, listed keys Tab, 00, BS, =, membrane, 24.5 mm tall, 143 g. Keys are judged by whether the maker lists them. A key that isn’t listed may still be there, so check the layout diagram	Open with these settings
Wired USB, Tab, =, Any thickness, NumLock handling stated	✓ 1 model matches. Sanwa Supply NT-27UBK. Sanwa Supply NT-27UBK: wired USB, listed keys Tab, =, pantograph, 18 mm tall, 110 g. NumLock (Sanwa Supply NT-27UBK): NumLock-independent (maker statement; has a NumLock LED). Keys are judged by whether the maker lists them. A key that isn’t listed may still be there, so check the layout diagram. NumLock handling is quoted from each maker. Models that don’t state it aren’t judged	Open with these settings
2.4 GHz (receiver), Tab, 00, Any thickness, NumLock doesn’t matter	✓ 5 models match. ELECOM TK-TDM017BK, ELECOM TK-TDM022SKBK, ELECOM TK-TDP019BK, Sanwa Supply NT-WL21BK, Sanwa Supply NT-WL23BK. ELECOM TK-TDM017BK: 2.4 GHz wireless (USB receiver), listed keys Tab, 00, membrane, 26.5 mm tall, 107 g. Battery (TK-TDM017BK): one AA battery, 1.3 years (estimated battery life, with alkaline batteries, maker’s own test). ELECOM TK-TDM022SKBK: 2.4 GHz wireless (USB receiver), listed keys Tab, 00, membrane, 26.6 mm tall, 80 g. Battery (TK-TDM022SKBK): one AAA battery, 1.5 years (estimated battery life, with alkaline batteries, assuming 8 hours of PC use a day). ELECOM TK-TDP019BK: 2.4 GHz wireless (USB receiver), listed keys Tab, 00, pantograph, 21.5 mm tall, 100 g. Battery (TK-TDP019BK): one AAA battery, 1 year (estimated battery life, with alkaline batteries, maker’s own test). Other matching models: NT-WL21BK, NT-WL23BK. Keys are judged by whether the maker lists them. A key that isn’t listed may still be there, so check the layout diagram. Battery figures are the makers’ own, stated in different ways (estimated battery life vs. continuous operating time), so they aren’t compared	Open with these settings
2.4 GHz (receiver), Tab, 00, Slim (pantograph keys), NumLock doesn’t matter	✓ 3 models match. ELECOM TK-TDP019BK, Sanwa Supply NT-WL21BK, Sanwa Supply NT-WL23BK. ELECOM TK-TDP019BK: 2.4 GHz wireless (USB receiver), listed keys Tab, 00, pantograph, 21.5 mm tall, 100 g. Battery (TK-TDP019BK): one AAA battery, 1 year (estimated battery life, with alkaline batteries, maker’s own test). Sanwa Supply NT-WL21BK: 2.4 GHz wireless (USB receiver), listed keys Tab, 00, pantograph, 17 mm tall, 70 g. NumLock (Sanwa Supply NT-WL21BK): NumLock-independent (maker statement). Battery (NT-WL21BK): one AAA battery, 70 hours (continuous operating time, with alkaline batteries). Sanwa Supply NT-WL23BK: 2.4 GHz wireless (USB receiver), listed keys Tab, 00, pantograph, 17 mm tall, 70 g. NumLock (Sanwa Supply NT-WL23BK): NumLock-independent (maker statement). Battery (NT-WL23BK): one AAA battery, 70 hours (continuous operating time, with alkaline batteries). Keys are judged by whether the maker lists them. A key that isn’t listed may still be there, so check the layout diagram. Slim means the maker lists a pantograph key switch (this article’s grouping). Battery figures are the makers’ own, stated in different ways (estimated battery life vs. continuous operating time), so they aren’t compared	Open with these settings
2.4 GHz (receiver), Tab, 00, Any thickness, NumLock handling stated	✓ 2 models match. Sanwa Supply NT-WL21BK, Sanwa Supply NT-WL23BK. Sanwa Supply NT-WL21BK: 2.4 GHz wireless (USB receiver), listed keys Tab, 00, pantograph, 17 mm tall, 70 g. NumLock (Sanwa Supply NT-WL21BK): NumLock-independent (maker statement). Battery (NT-WL21BK): one AAA battery, 70 hours (continuous operating time, with alkaline batteries). Sanwa Supply NT-WL23BK: 2.4 GHz wireless (USB receiver), listed keys Tab, 00, pantograph, 17 mm tall, 70 g. NumLock (Sanwa Supply NT-WL23BK): NumLock-independent (maker statement). Battery (NT-WL23BK): one AAA battery, 70 hours (continuous operating time, with alkaline batteries). Keys are judged by whether the maker lists them. A key that isn’t listed may still be there, so check the layout diagram. NumLock handling is quoted from each maker. Models that don’t state it aren’t judged. Battery figures are the makers’ own, stated in different ways (estimated battery life vs. continuous operating time), so they aren’t compared	Open with these settings
2.4 GHz (receiver), Tab, 00, BS, =, Any thickness, NumLock doesn’t matter	✕ No model in this research matches. Keys are judged by whether the maker lists them. A key that isn’t listed may still be there, so check the layout diagram	Open with these settings
Bluetooth, Tab, 00, Any thickness, NumLock doesn’t matter	✓ 4 models match. ELECOM TK-TBM023SKBK, ELECOM TK-TBP020BK, Sanwa Supply NT-BT21BK, Sanwa Supply NT-BT23BK. ELECOM TK-TBM023SKBK: Bluetooth, listed keys Tab, 00, membrane, 26.6 mm tall, 80 g. Battery (TK-TBM023SKBK): one AAA battery, 10 months (estimated battery life, with alkaline batteries, assuming 8 hours of PC use a day). Devices (TK-TBM023SKBK): One device at a time (paired). ELECOM TK-TBP020BK: Bluetooth, listed keys Tab, 00, pantograph, 21.5 mm tall, 100 g. Battery (TK-TBP020BK): one AAA battery, 1 year (estimated battery life, with alkaline batteries, maker’s own test). Devices (TK-TBP020BK): One device at a time (paired). Sanwa Supply NT-BT21BK: Bluetooth, listed keys Tab, 00, pantograph, 17 mm tall, 70 g. NumLock (Sanwa Supply NT-BT21BK): NumLock-independent (maker statement). Battery (NT-BT21BK): one AAA battery, 150 hours (continuous operating time). Other matching models: NT-BT23BK. Keys are judged by whether the maker lists them. A key that isn’t listed may still be there, so check the layout diagram. Battery figures are the makers’ own, stated in different ways (estimated battery life vs. continuous operating time), so they aren’t compared	Open with these settings
Bluetooth, Tab, 00, Slim (pantograph keys), NumLock doesn’t matter	✓ 3 models match. ELECOM TK-TBP020BK, Sanwa Supply NT-BT21BK, Sanwa Supply NT-BT23BK. ELECOM TK-TBP020BK: Bluetooth, listed keys Tab, 00, pantograph, 21.5 mm tall, 100 g. Battery (TK-TBP020BK): one AAA battery, 1 year (estimated battery life, with alkaline batteries, maker’s own test). Devices (TK-TBP020BK): One device at a time (paired). Sanwa Supply NT-BT21BK: Bluetooth, listed keys Tab, 00, pantograph, 17 mm tall, 70 g. NumLock (Sanwa Supply NT-BT21BK): NumLock-independent (maker statement). Battery (NT-BT21BK): one AAA battery, 150 hours (continuous operating time). Sanwa Supply NT-BT23BK: Bluetooth, listed keys Tab, 00, pantograph, 17 mm tall, 70 g. NumLock (Sanwa Supply NT-BT23BK): NumLock-independent (maker statement). Battery (NT-BT23BK): one AAA battery, 150 hours (continuous operating time). Keys are judged by whether the maker lists them. A key that isn’t listed may still be there, so check the layout diagram. Slim means the maker lists a pantograph key switch (this article’s grouping). Battery figures are the makers’ own, stated in different ways (estimated battery life vs. continuous operating time), so they aren’t compared	Open with these settings
Bluetooth, Tab, 00, Any thickness, NumLock handling stated	✓ 2 models match. Sanwa Supply NT-BT21BK, Sanwa Supply NT-BT23BK. Sanwa Supply NT-BT21BK: Bluetooth, listed keys Tab, 00, pantograph, 17 mm tall, 70 g. NumLock (Sanwa Supply NT-BT21BK): NumLock-independent (maker statement). Battery (NT-BT21BK): one AAA battery, 150 hours (continuous operating time). Sanwa Supply NT-BT23BK: Bluetooth, listed keys Tab, 00, pantograph, 17 mm tall, 70 g. NumLock (Sanwa Supply NT-BT23BK): NumLock-independent (maker statement). Battery (NT-BT23BK): one AAA battery, 150 hours (continuous operating time). Keys are judged by whether the maker lists them. A key that isn’t listed may still be there, so check the layout diagram. NumLock handling is quoted from each maker. Models that don’t state it aren’t judged. Battery figures are the makers’ own, stated in different ways (estimated battery life vs. continuous operating time), so they aren’t compared	Open with these settings
Bluetooth, Any keys, Any thickness, NumLock doesn’t matter	✓ 6 models match. ELECOM TK-TBM023SKBK, ELECOM TK-TBP020BK, Sanwa Supply NT-BT21BK, Sanwa Supply NT-BT23BK, Sanwa Supply NT-BT27BK, Satechi ST-XLABKM. ELECOM TK-TBM023SKBK: Bluetooth, listed keys Tab, 00, membrane, 26.6 mm tall, 80 g. Battery (TK-TBM023SKBK): one AAA battery, 10 months (estimated battery life, with alkaline batteries, assuming 8 hours of PC use a day). Devices (TK-TBM023SKBK): One device at a time (paired). ELECOM TK-TBP020BK: Bluetooth, listed keys Tab, 00, pantograph, 21.5 mm tall, 100 g. Battery (TK-TBP020BK): one AAA battery, 1 year (estimated battery life, with alkaline batteries, maker’s own test). Devices (TK-TBP020BK): One device at a time (paired). Sanwa Supply NT-BT21BK: Bluetooth, listed keys Tab, 00, pantograph, 17 mm tall, 70 g. NumLock (Sanwa Supply NT-BT21BK): NumLock-independent (maker statement). Battery (NT-BT21BK): one AAA battery, 150 hours (continuous operating time). Other matching models: NT-BT23BK, NT-BT27BK, ST-XLABKM. No Windows 11 listing (not used for picks): ST-XLABKM. Battery figures are the makers’ own, stated in different ways (estimated battery life vs. continuous operating time), so they aren’t compared	Open with these settings
Any, Tab, 00, BS, =, Any thickness, NumLock doesn’t matter	✓ 1 model matches. ELECOM TK-TCM015BK. ELECOM TK-TCM015BK: wired USB, listed keys Tab, 00, BS, =, membrane, 24.5 mm tall, 143 g. Keys are judged by whether the maker lists them. A key that isn’t listed may still be there, so check the layout diagram	Open with these settings

Vue.js for beginners — components, props, and two-way binding, explained with real code

Fri, 09 Apr 2021 00:00:00 GMT

2026 update: a post ported over from the old VuePress blog. The shape of the explanation still holds, but the tooling and Vue version are frozen at the time of writing and differ from the current release.

For anyone picking up Vue.js — or a JavaScript framework — for the first time, the part to understand first is components.

This post starts from what a component is, so it should read cleanly even without prior framework experience.

The short version

A component is a reusable UI part — it raises maintainability and lets you reuse work
Data flows between a parent and a child through binding
Use one-way binding (v-bind / v-on) and two-way binding (v-bind + v-on, or .sync) depending on whether the data needs transforming

What is a component?

Vue.js and Angular both build on the idea of a component.

Vue uses components to consolidate code. Consolidation cuts down the amount of code you maintain and makes it easier to change later.

Every Vue project uses components in some form, so understanding them is non-optional.

A reusable part

A component, in one line, is a reusable part.

A .vue file can import and use another .vue file inside its own template.

Build a button as a reusable component, for example, and any caller can render the same look and the same behaviour just by referencing it.

Example: a button

Goal: reuse a 20 px square button labelled “click here”.

How:

In file A, define a 20 px square button labelled “click here”.

In file B, render the button component.

Result: the button shows up in file B.

The side being called (file A) and the side calling it (file B) are usually referred to as the child and the parent.

From here on:

The caller = parent
The called = child

What is binding?

In the button example above, two limitations stand out:

The button text is hard-coded
The click handler has to live in the child

Binding is what removes those limits.

Component-to-component data flow

Binding, in one line, is the data flow between components.

With binding in place you can:

Let the parent decide the child’s text
Let the child propagate an event up to the parent

Both of the issues above go away once binding is added.

The relationship between the parent and the child matters here:

A one-way trip — parent to child, or child to parent — is one-way binding
A round trip between the two is two-way binding

Components do not click into place until binding does, so this is the section worth re-reading.

Working examples

From here, the explanation is anchored to real Vue.js code.

If .vue file basics are unfamiliar, the earlier post covers them:

Vue.js for beginners — the basics

Using a component

Start with the plainest case: rendering a component with no binding at all.

The child shows a string. The parent renders the child.

Child: ChildComponent.vue

<template>
  string
</template>

The child has no logic — it just prints text.

Parent: ParentComponent.vue

<template>
  <ChildComponent />
</template>

<script>
import ChildComponent from "xxxx/ChildComponent.vue";

export default {
  components: { ChildComponent },
};
</script>

The parent imports ChildComponent.vue and registers it under components. On the template side, dropping <ChildComponent /> where you want it to render is enough.

v-bind — one-way binding, parent to child

With the call mechanics out of the way, the next step is parent-to-child binding.

The idea is: let the parent decide the child’s text.

The snippet below passes a string set in the parent down to the child.

Child: ChildComponent.vue

<template>
  {{bindName}}
</template>

<script>
export default {
  props: ["bindName"],
};
</script>

To receive data from the parent, the child declares props.

The prop name has to match on both sides. Here it is bindName.

props is an array, so a child can declare multiple inbound props at once.

Parent: ParentComponent.vue

<template>
  <ChildComponent :bindName="dataName" />
</template>

<script>
import ChildComponent from "xxxx/ChildComponent.vue";

export default {
  components: { ChildComponent },

  data() {
    return {
      dataName: 'string'
    };
  },
};
</script>

The parent renders the child with :bindName="dataName". That attribute is the bind itself.

bindName is the prop name on the child
dataName is the data field on the parent

That is the full parent-to-child flow.

v-on — one-way binding, child to parent

Next, child-to-parent binding.

The idea is: let the parent observe an event that happened inside the child.

In the snippet below, clicking the child’s text fires an event in the parent and flips the parent’s text from string → モジ.

Child: ChildComponent.vue

<template>
  <div>
    <div @click="clickText()">
      クリックして変更
    </div>
  </div>
</template>

<script>
export default {

  methods: {
    clickText() {
      this.$emit('onName', 'モジ');
    }
  }
};
</script>

The child wires up a click handler — @click is the directive.

Inside the method, the line that lifts the event up to the parent is emit.

The first argument, onName, is the event name
The second argument, モジ, is the payload (an object works too)

Clicking “クリックして変更” propagates up to the parent’s onName listener.

Parent: ParentComponent.vue

<template>
  {{dataName}}
  <ChildComponent @onName="changeName"/>
</template>

<script>
import ChildComponent from "xxxx/ChildComponent.vue";

export default {
  components: { ChildComponent },

  data() {
    return {
      dataName: 'string'
    };
  },

  methods: {
    changeName(item) {
      this.dataName = item
    }
  }
};
</script>

The parent looks much like the v-bind case.

The new piece is changeName — its argument is the payload the child emitted.

That is the full child-to-parent flow.

Going further

The one-way patterns above cannot move data in both directions on their own.

Two-way binding is what closes that loop. It is the trickiest part of components, so this section is worth slowing down on.

v-bind + v-on — two-way binding

Vue.js has two common ways to do two-way binding. The most fundamental one comes first.

It is the combination of:

v-bind — one-way, parent to child
v-on — one-way, child to parent

The flow:

The parent sets the string the child will show (string)
The child sends a click event up to the parent (payload: モジ)
The parent’s string flips from string → モジ
The child re-renders with モジ

Previous sections led with the child. From here, the parent comes first — it reads more naturally.

Parent: ParentComponent.vue

<template>
  <ChildComponent :bindName="dataName" @onName="changeName"/>
</template>

<script>
import ChildComponent from "xxxx/ChildComponent.vue";

export default {
  components: { ChildComponent },

  data() {
    return {
      dataName: 'string'
    };
  },

  methods: {
    changeName(item) {
      this.dataName = item
    }
  }
};
</script>

It is just v-bind and v-on running side by side. The parent stays straightforward.

Child: ChildComponent.vue

<template>
  <div>
    {{bindName}}
    <div @click="clickText()">
      クリックして変更
    </div>
  </div>
</template>

<script>
export default {
  props: ["bindName"],

  methods: {
    clickText() {
      this.$emit('onName', 'モジ');
    }
  }
};
</script>

The child is the same shape — v-bind plus v-on.

If this stops making sense, walking through the four-step flow above in order tends to put it back together.

.sync — two-way binding

v-bind + v-on works, but it gets verbose.

.sync is the shortcut for the cases that do not need the wiring.

In one line, .sync turns v-bind into a synced binding.

The code below produces the same end behaviour as the previous snippet:

The parent sets the string the child will show (string)
The child sends a click event up to the parent (payload: モジ)
The parent’s string flips from string → モジ
The child re-renders with モジ

Parent: ParentComponent.vue

<template>
  <ChildComponent :bindName.sync="dataName" />
</template>

<script>
import ChildComponent from "xxxx/ChildComponent.vue";

export default {
  components: { ChildComponent },

  data() {
    return {
      dataName: 'string'
    };
  },
};
</script>

The parent gets noticeably lighter.

The key is bindName.sync — adding .sync removes the explicit assignment that v-on was doing.

Child: ChildComponent.vue

<template>
  <div>
    {{bindName}}
    <div @click="clickText()">
      クリックして変更
    </div>
  </div>
</template>

<script>
export default {
  props: ["bindName"],

  methods: {
    clickText() {
      this.$emit('update:bindName', 'モジ');
    }
  }
};
</script>

The child changes slightly — the first argument to emit becomes update:bindName.

The update: prefix is what wires the emit back into the parent’s .sync.

v-on vs. .sync — when to use which

.sync is shorter, but it has a caveat.

Because it is a synced bind, there is no place to hook in transformation logic. If the parent needs to filter or reshape the child’s payload before assigning it, .sync will not fit.

A rough rule for two-way binding:

If the data is transformed between files → v-bind + v-on
If the data is passed through unchanged → .sync

Wrap-up

Components and binding are two of the load-bearing pieces of Vue.js, and the rest of the framework leans on them. Worth re-reading and writing out by hand until they feel automatic.

Best MacBook for Blogging: Smallest Config by Use

Thu, 08 Apr 2021 00:00:00 GMT

For the October 2026 lineup, the smallest configuration is 8 GB of memory and 256 GB of storage. That holds with no Lightroom Classic AI features and external displays capped at 1 display. It’s borderline against the 8 GB minimum for Lightroom Classic and Photoshop.

With two external displays, or Lightroom Classic’s AI features with up to two, the smallest option is an M5 configuration with 16 GB of memory and 512 GB of storage. It’s borderline against both the recommended 16 GB or more and Lightroom Classic’s 16 GB for AI features. With 3 displays you need the M5 Pro chip, and with 4 displays the M5 Max.

This covers the MacBooks Apple sells now, checked as of October 2026. Lightroom Classic’s AI features are the ones that process photos with AI. Nothing was tested on real hardware, so speed and comfort aren’t judged.

The smallest config depends on your use and your external displays

The table ignores weight; weight only comes in through the carry option in the tool below. For plain writing, only the number of external displays matters. For photo editing, the pick is the first config, going from Neo to Air to Pro, that meets Adobe’s minimum RAM.

External displays	Writing only (RAM not judged)	Lightroom Classic (minimum 8 GB)	Photoshop (minimum 8 GB)	Lightroom Classic AI features (16 GB)
Up to 1 display	MacBook Neo, 8 GB / 256 GB, memory not judged	MacBook Neo, 8 GB / 256 GB, borderline	MacBook Neo, 8 GB / 256 GB, borderline	13-inch MacBook Air (M5), 16 GB / 512 GB, borderline
2 displays	13-inch MacBook Air (M5), 16 GB / 512 GB, memory not judged	13-inch MacBook Air (M5), 16 GB / 512 GB	13-inch MacBook Air (M5), 16 GB / 512 GB	13-inch MacBook Air (M5), 16 GB / 512 GB, borderline
3 displays	14-inch MacBook Pro (M5 Pro), 24 GB / 1 TB, memory not judged	14-inch MacBook Pro (M5 Pro), 24 GB / 1 TB	14-inch MacBook Pro (M5 Pro), 24 GB / 1 TB	14-inch MacBook Pro (M5 Pro), 24 GB / 1 TB
4 displays	14-inch MacBook Pro (M5 Max), 36 GB / 2 TB, memory not judged	14-inch MacBook Pro (M5 Max), 36 GB / 2 TB	14-inch MacBook Pro (M5 Max), 36 GB / 2 TB	14-inch MacBook Pro (M5 Max), 36 GB / 2 TB

“Borderline” means the spare RAM (the configuration’s RAM minus the minimum) is under 10% of the minimum. Per Adobe’s official requirements, each official figure is for that app on its own. A figure for running Lightroom Classic and Photoshop together wasn’t found in Adobe’s requirements, so this article doesn’t judge it.

Only the maximum display count was checked, not resolution or refresh rate combinations.

Minimum MacBook configuration for blogging

Pick what you do (writing only, Lightroom Classic, Photoshop, or AI features), how much you carry it and how many external displays you use. It rounds Adobe's official memory figures up to Apple's memory tiers, shows the lightest configuration that fits, and suggests one MacBook you can buy.

Is the MacBook Neo enough?

The MacBook Neo (8 GB) is exactly at the 8 GB minimum for Lightroom Classic and Photoshop, so it’s borderline. It doesn’t meet the recommendation (16 GB or more) or the 16 GB listed for the AI features, and it drives only 1 display.

Per Apple’s spec pages, the 256 GB Neo has no Touch ID, while the 512 GB model has a Magic Keyboard with Touch ID.

How much RAM and storage to pick

Lightroom Classic and Photoshop for Mac need at least 8 GB of RAM, with 16 GB or more recommended. For Lightroom Classic’s AI features, Adobe lists 16 GB of unified memory. On these Macs the RAM and the GPU share that one unified memory, so the figure is the same RAM, not extra.

Lightroom Classic needs 8 GB of free space and Photoshop 10 GB, with 100 GB recommended for Photoshop. Even the smallest current drive, the Neo’s 256 GB, is above those figures. Space taken by macOS and your photos isn’t calculated.

Keyboard and SD card

Per Apple’s US spec pages, the keyboard is ANSI with 78 keys, or ISO with 79, and both have 12 full-height function keys. Retail listings can differ, so check the layout on the one you buy.

Per Apple’s spec pages, an SDXC card slot appears on the MacBook Pro spec page, but not on the Air or Neo spec pages.

Weight and screen size

Taking 1.3 kg as the limit for daily carry, only the Neo and the 13-inch Air fit, both at 1.23 kg. If you need 16 GB, that’s the 13-inch Air, and nothing this light drives three or more external displays.

Model	Screen	Resolution	Weight	External displays	Web browsing	Ports
Neo	13.0 in	2,408 × 1,506	1.23 kg	1 display	11 hours	1 × USB 3, 1 × USB 2
Air 13	13.6 in	2,560 × 1,664	1.23 kg	2 displays	15 hours	2 × Thunderbolt 4
Air 15	15.3 in	2,880 × 1,864	1.51 kg	2 displays	15 hours	2 × Thunderbolt 4
Pro 14 M5 Pro	14.2 in	3,024 × 1,964	1.60 kg	3 displays	14 hours	Thunderbolt, HDMI

Web browsing is Apple’s maximum time from its own tests. The Neo’s USB 3 port can output video.

Picks by condition

One pick per condition, from Amazon.com:

Sold new by Amazon.com, with a configuration that matches Apple’s spec
At least 100 ratings and 4 stars or more

Prices vary, so check them on the retailer’s page.

Writing only, or photo editing with the 8 GB minimum, no Lightroom Classic AI features, up to 1 display

It’s exactly at the 8 GB minimum, so it’s borderline. In this calculation, it falls short of Adobe’s recommended 16 GB or more.

Other colors are separate listings. Whether ratings are pooled was not checked.

Recommended 16 GB or more, Lightroom Classic AI features, up to 2 displays, 1.3 kg or less

It’s exactly at the 16 GB for the AI features, so it’s borderline.

15.3 in screen, Lightroom Classic AI features, up to 2 displays

Also exactly at the 16 GB for the AI features, so borderline.

3 displays external displays and Lightroom Classic’s AI features

Its standard 24 GB leaves 8 GB of room over the 16 GB for the AI features.

FAQ

Is the MacBook Neo enough for just writing a blog?

For the October 2026 lineup, writing alone has no official memory figure, so this article doesn’t judge memory for it. Any current model is a candidate, and the choice comes down to external displays and weight.

Models that fit are in the picks section.

Can the MacBook Neo handle blog photo editing?

For the October 2026 lineup, the smallest-memory configuration, 8 GB, is exactly at the 8 GB minimum for Lightroom Classic and Photoshop, so it’s borderline. It doesn’t meet the recommendation (16 GB or more) or the 16 GB listed for the AI features, and it drives only 1 display.

Photoshop’s AI features aren’t judged. Matching models are listed under Picks by condition.

How much RAM does a MacBook need for blogging and photo editing?

Lightroom Classic and Photoshop for Mac need at least 8 GB of RAM, with 16 GB or more recommended. For Lightroom Classic’s AI features, Adobe lists 16 GB of unified memory.

On these Macs the RAM and the GPU share that one unified memory, so the figure is the same RAM, not extra. Among Apple’s current MacBooks as of October 2026, the lightest one that meets the recommended 16 GB or more weighs 1.23 kg.

It has the M5 chip and the standard 16 GB. That’s exactly the low end of the recommendation, so it’s borderline.

This article’s minimum configurations use the minimum, not the recommendation.

What keyboard layout and SD card slot do MacBooks have?

Per Apple’s US spec pages, the keyboard is ANSI with 78 keys, or ISO with 79, and both have 12 full-height function keys. Retail listings can differ, so check the layout on the one you buy.

On current MacBooks, an SDXC card slot appears on Apple’s spec pages for the 14.2-inch and 16.2-inch models, but not on the 13-inch or 15-inch models.

How many external displays can a MacBook run?

Per Apple’s spec pages, on current MacBooks the display count depends on the chip. The M5 drives up to 2 displays, the M5 Pro 3 displays and the M5 Max 4 displays.

The 8 GB configuration drives 1 display.

Data and formulas used (for verification)

How this was checked:

Sources: Apple specs for Neo, Air and Pro, and Compare Mac ／ Adobe Lightroom Classic system requirements and Adobe Photoshop system requirements

Method

Memory needed = Adobe’s official figure for the chosen use (Lightroom Classic for Mac minimum 8 GB, Photoshop minimum 8 GB, and the 16 GB of unified memory Adobe lists in the recommended column for Lightroom Classic’s AI features such as Denoise). Writing alone has no official figure, so memory isn’t judged for it. Photoshop’s macOS table gives no memory figure for AI features (No memory figure for AI features in the macOS table (RAM: 8 GB minimum, 16 GB or more recommended)). The official recommendations (Lightroom Classic 16 GB, Photoshop 16 GB) aren’t used for the minimum configuration; the lightest configuration that meets them is shown separately.
Minimum configuration = the smallest memory tier on Apple’s spec pages at or above the memory needed (Neo 8 GB; Air and 14-inch Pro with M5: 16, 24, 32 GB; M5 Pro: 24, 48, 64 GB; M5 Max from 36 GB). Models that meet the display count and weight limit are checked in the order Neo → 13-inch Air → 15-inch Air → Pro, and the first one is shown (the calculation confirms it’s also the lightest that fits). If the headroom (configured memory − memory needed) is under 10% of the memory needed, it’s marked borderline (a threshold set for this article).
External display counts are Apple’s spec-page maximums (Neo 1, Air 2, Pro with M5 2, M5 Pro 3, M5 Max 4). Weights are from Apple’s spec pages (Neo 1.23 kg, 13-inch Air 1.23 kg, 15-inch Air 1.51 kg, 14-inch Pro with M5 1.55 kg). The carrying limits (1.3 kg daily, 1.7 kg sometimes) are this article’s own choices.
Storage is only compared with Adobe’s official free-space figures (Lightroom Classic 8 GB, Photoshop 10 GB minimum and 100 GB recommended); space used by macOS, photos and catalogs isn’t calculated. SD card slots are compared from Apple’s spec pages in Japan and the US (Pro: SDXC card slot; Air: No SD card slot listed (MagSafe 3, two Thunderbolt 4, headphone jack); Neo: No SD card slot listed (one USB 3, one USB 2, headphone jack)).
Speed and comfort weren’t tested on real hardware and aren’t judged. Picks are new listings sold by Amazon with 100+ ratings, 4+ stars, and a configuration matching Apple’s specs, cheapest including shipping.

Official values used

Item	Value	Source	Checked
MacBook Air 13-inch display size (diagonal)	13.6 in	apple.com	2026-10-08
MacBook Air 13-inch native width	2,560 px	apple.com	2026-10-08
MacBook Air 13-inch native height	1,664 px	apple.com	2026-10-08
MacBook Air 13-inch weight	1.23 kg	apple.com	2026-10-08
MacBook Air 13-inch video streaming battery life (up to)	18 hours	apple.com	2026-10-08
MacBook Air 13-inch wireless web battery life (up to)	15 hours	apple.com	2026-10-08
MacBook Air (M5) maximum external displays	2	apple.com	2026-10-08
MacBook Air 13-inch base memory	16 GB	apple.com	2026-10-08
MacBook Air 13-inch base storage	512 GB	apple.com	2026-10-08
MacBook Air (M5) maximum memory option	32 GB	apple.com	2026-10-08
MacBook Air (M5) maximum storage option	4 TB	apple.com	2026-10-08
MacBook Air 15-inch display size (diagonal)	15.3 in	apple.com	2026-10-08
MacBook Air 15-inch native width	2,880 px	apple.com	2026-10-08
MacBook Air 15-inch native height	1,864 px	apple.com	2026-10-08
MacBook Air 15-inch weight	1.51 kg	apple.com	2026-10-08
MacBook Air 15-inch video streaming battery life (up to)	18 hours	apple.com	2026-10-08
MacBook Air 15-inch wireless web battery life (up to)	15 hours	apple.com	2026-10-08
MacBook Pro 14-inch display size (diagonal)	14.2 in	apple.com	2026-10-08
MacBook Pro 14-inch native width	3,024 px	apple.com	2026-10-08
MacBook Pro 14-inch native height	1,964 px	apple.com	2026-10-08
MacBook Pro 14-inch (M5) weight	1.55 kg	apple.com	2026-10-08
MacBook Pro 14-inch (M5 Pro) weight	1.6 kg	apple.com	2026-10-08
MacBook Pro 14-inch (M5) video streaming battery life (up to)	24 hours	apple.com	2026-10-08
MacBook Pro 14-inch (M5) wireless web battery life (up to)	16 hours	apple.com	2026-10-08
MacBook Pro 14-inch (M5 Pro) video streaming battery life (up to)	22 hours	apple.com	2026-10-08
MacBook Pro 14-inch (M5 Pro) wireless web battery life (up to)	14 hours	apple.com	2026-10-08
MacBook Pro (M5) maximum external displays	2	apple.com	2026-10-08
MacBook Pro (M5 Pro) maximum external displays	3	apple.com	2026-10-08
MacBook Pro (M5 Max) maximum external displays	4	apple.com	2026-10-08
MacBook Pro 14-inch (M5) base memory	16 GB	apple.com	2026-10-08
MacBook Pro 14-inch (M5) base storage	1 TB	apple.com	2026-10-08
MacBook Pro 14-inch (M5 Pro) base memory	24 GB	apple.com	2026-10-08
MacBook Pro (M5 Pro) maximum memory option	64 GB	apple.com	2026-10-08
MacBook Pro 16-inch display size (diagonal)	16.2 in	apple.com	2026-10-08
MacBook Pro 16-inch (M5 Pro) weight	2.14 kg	apple.com	2026-10-08
MacBook Pro 16-inch (M5 Pro) wireless web battery life (up to)	17 hours	apple.com	2026-10-08
MacBook Neo display size (diagonal)	13.0 in	apple.com	2026-10-08
MacBook Neo native width	2,408 px	apple.com	2026-10-08
MacBook Neo native height	1,506 px	apple.com	2026-10-08
MacBook Neo weight	1.23 kg	apple.com	2026-10-08
MacBook Neo video streaming battery life (up to)	16 hours	apple.com	2026-10-08
MacBook Neo wireless web battery life (up to)	11 hours	apple.com	2026-10-08
MacBook Neo memory (the only configuration listed on the spec page)	8 GB	apple.com	2026-10-08
MacBook Neo storage (model 1)	256 GB	apple.com	2026-10-08
MacBook Neo storage (model 2)	512 GB	apple.com	2026-10-08
MacBook Neo maximum external displays	1	apple.com	2026-10-08
MacBook Neo USB 3 (USB-C) ports	1	apple.com	2026-10-08
MacBook Neo USB 2 (USB-C) ports	1	apple.com	2026-10-08
MacBook Pro 16-inch native width	3,456 px	apple.com	2026-10-08
MacBook Pro 16-inch native height	2,234 px	apple.com	2026-10-08
MacBook Pro 16-inch (M5 Max) weight	2.15 kg	apple.com	2026-10-08
MacBook Pro 16-inch (M5 Pro) maximum external displays	3	apple.com	2026-10-08
MacBook Pro 16-inch (M5 Max) maximum external displays	4	apple.com	2026-10-08
MacBook Pro 16-inch base memory (M5 Pro)	24 GB	apple.com	2026-10-08
MacBook Pro 16-inch base storage (M5 Pro)	1 TB	apple.com	2026-10-08
MacBook Pro 16-inch (M5 Pro) video streaming battery life (up to)	24 hours	apple.com	2026-10-08
MacBook Air 15-inch base memory	16 GB	apple.com	2026-10-08
MacBook Air 15-inch base storage	512 GB	apple.com	2026-10-08
MacBook Pro 14-inch (M5 Max) weight	1.62 kg	apple.com	2026-10-08
MacBook Pro (M5 Max) smallest memory option	36 GB	apple.com	2026-10-08
MacBook Pro (M5 Max) base storage	2 TB	apple.com	2026-10-08
MacBook Pro 14-inch (M5) maximum memory option	32 GB	apple.com	2026-10-08
MacBook Pro (M5 Max) maximum memory option	128 GB	apple.com	2026-10-08
Lightroom Classic (Mac) minimum RAM	8 GB	helpx.adobe.com	2026-10-08
Lightroom Classic (Mac) recommended RAM	16 GB	helpx.adobe.com	2026-10-08
Lightroom Classic (Mac) unified memory for full GPU acceleration and AI features	16 GB	helpx.adobe.com	2026-10-08
Lightroom Classic (Mac) free disk space for installation	8 GB	helpx.adobe.com	2026-10-08
Photoshop (Mac) minimum RAM	8 GB	helpx.adobe.com	2026-10-08
Photoshop (Mac) recommended RAM	16 GB	helpx.adobe.com	2026-10-08
Photoshop (Mac) free disk space for installation	10 GB	helpx.adobe.com	2026-10-08
Photoshop (Mac) recommended free disk space	100 GB	helpx.adobe.com	2026-10-08
Keyboard layout of MacBooks on Apple Japan (default)	(stated in a Japanese-language source)	apple.com	2026-10-08
US layout option on Apple Japan spec pages	(stated in a Japanese-language source)	apple.com	2026-10-08
Keyboard of MacBooks on Apple US (ANSI / ISO)	(stated in a Japanese-language source)	apple.com Apple US MacBook Air Tech Specs: 78 (ANSI) or 79 (ISO) keys including 12 full-he	2026-10-08
MacBook Neo 256GB model Touch ID	(stated in a Japanese-language source)	apple.com	2026-10-08
MacBook Pro SD card slot	(stated in a Japanese-language source)	apple.com Apple US MacBook Pro Tech Specs, Charging and Expansion: SDXC card slot, HDMI po	2026-10-08
MacBook Air Thunderbolt 4 (USB-C) ports	2	apple.com	2026-10-08
MacBook Air ports (SD card slot)	(stated in a Japanese-language source)	apple.com	2026-10-08
MacBook Neo ports (SD card slot)	(stated in a Japanese-language source)	apple.com	2026-10-08
Photoshop (Mac) memory figure for AI features	(stated in a Japanese-language source)	helpx.adobe.com	2026-10-08

Minimum MacBook configuration for blogging: quick reference

The calculator in this article shows the result and a matching product for your own settings. Settings can be set in the URL; each link below opens the calculator with those settings.

Settings	Result	Calculator
Writing only, Daily (1.3 kg or less), None	△ MacBook Neo (memory not judged). Writing alone has no official memory figure, so memory isn’t judged. The configuration below is chosen only by display count and weight. The weight limits (1.3 kg and 1.7 kg) are this article’s own choices. Lightest configuration: MacBook Neo, 8 GB memory, 256 GB storage. Weight: 1.23 kg (Apple’s spec page). The MacBook Neo drives up to 1 external display. Other models that fit: 13-inch MacBook Air (M5)	Open with these settings
Writing only, Daily (1.3 kg or less), 1	△ MacBook Neo (memory not judged). Writing alone has no official memory figure, so memory isn’t judged. The configuration below is chosen only by display count and weight. The weight limits (1.3 kg and 1.7 kg) are this article’s own choices. Lightest configuration: MacBook Neo, 8 GB memory, 256 GB storage. Weight: 1.23 kg (Apple’s spec page). The MacBook Neo drives up to 1 external display. The Neo outputs video only from its single USB 3 port, so with a display connected you’re left with one USB 2 port. Other models that fit: 13-inch MacBook Air (M5)	Open with these settings
Writing only, Doesn’t matter, 3	△ 14-inch MacBook Pro (M5 Pro) (memory not judged). Writing alone has no official memory figure, so memory isn’t judged. The configuration below is chosen only by display count and weight. Lightest configuration: 14-inch MacBook Pro (M5 Pro), 24 GB memory, 1 TB storage. Weight: 1.60 kg (Apple’s spec page). The 14-inch MacBook Pro (M5 Pro) drives up to 3 external displays. Other models that fit: 16-inch MacBook Pro (M5 Pro), 14-inch MacBook Pro (M5 Max), 16-inch MacBook Pro (M5 Max)	Open with these settings
+ Lightroom Classic, Daily (1.3 kg or less), None	△ The MacBook Neo just meets it. Memory needed: 8 GB (Lightroom Classic’s official minimum). Free space needed: 8 GB (Adobe’s official figure; photos and catalogs not included). The weight limits (1.3 kg and 1.7 kg) are this article’s own choices. Lightest configuration: MacBook Neo, 8 GB memory, 256 GB storage. Weight: 1.23 kg (Apple’s spec page). Memory is exactly the official figure with nothing to spare (borderline). The 256 GB drive is larger than that. Space for macOS and your photos isn’t calculated. Adobe recommends 16 GB or more (this check uses the minimum). Lightest configuration meeting the recommendation: 13-inch MacBook Air (M5), 16 GB (exactly the low end, borderline). The MacBook Neo drives up to 1 external display. The MacBook Neo (8 GB) is exactly at the 8 GB minimum for Lightroom Classic and Photoshop, and doesn’t meet the recommended 16 GB or more or the 16 GB for AI features. Pick one use. Each official figure is for that app on its own; running Lightroom Classic and Photoshop together has no official figure and isn’t judged. Other models that fit: 13-inch MacBook Air (M5)	Open with these settings
+ Lightroom Classic, Daily (1.3 kg or less), 2	✓ The 13-inch MacBook Air (M5) meets the official minimum. Memory needed: 8 GB (Lightroom Classic’s official minimum). Free space needed: 8 GB (Adobe’s official figure; photos and catalogs not included). The weight limits (1.3 kg and 1.7 kg) are this article’s own choices. Lightest configuration: 13-inch MacBook Air (M5), 16 GB memory, 512 GB storage. Weight: 1.23 kg (Apple’s spec page). Memory headroom: 8 GB. The 512 GB drive is larger than that. Space for macOS and your photos isn’t calculated. Adobe recommends 16 GB or more (this check uses the minimum). The 13-inch MacBook Air (M5) drives up to 2 external displays. Pick one use. Each official figure is for that app on its own; running Lightroom Classic and Photoshop together has no official figure and isn’t judged	Open with these settings
+ Lightroom Classic, Doesn’t matter, 3	✓ The 14-inch MacBook Pro (M5 Pro) meets the official minimum. Memory needed: 8 GB (Lightroom Classic’s official minimum). Free space needed: 8 GB (Adobe’s official figure; photos and catalogs not included). Lightest configuration: 14-inch MacBook Pro (M5 Pro), 24 GB memory, 1 TB storage. Weight: 1.60 kg (Apple’s spec page). Memory headroom: 16 GB. The 1 TB drive is larger than that. Space for macOS and your photos isn’t calculated. Adobe recommends 16 GB or more (this check uses the minimum). The 14-inch MacBook Pro (M5 Pro) drives up to 3 external displays. Pick one use. Each official figure is for that app on its own; running Lightroom Classic and Photoshop together has no official figure and isn’t judged. Other models that fit: 16-inch MacBook Pro (M5 Pro), 14-inch MacBook Pro (M5 Max), 16-inch MacBook Pro (M5 Max)	Open with these settings
+ Photoshop, Daily (1.3 kg or less), 1	△ The MacBook Neo just meets it. Memory needed: 8 GB (Photoshop’s official minimum). Free space needed: 10 GB, 100 GB recommended (Adobe’s official figure; photos and catalogs not included). The weight limits (1.3 kg and 1.7 kg) are this article’s own choices. Lightest configuration: MacBook Neo, 8 GB memory, 256 GB storage. Weight: 1.23 kg (Apple’s spec page). Memory is exactly the official figure with nothing to spare (borderline). The 256 GB drive is larger than that. Space for macOS and your photos isn’t calculated. Adobe recommends 16 GB or more (this check uses the minimum). Lightest configuration meeting the recommendation: 13-inch MacBook Air (M5), 16 GB (exactly the low end, borderline). The MacBook Neo drives up to 1 external display. The MacBook Neo (8 GB) is exactly at the 8 GB minimum for Lightroom Classic and Photoshop, and doesn’t meet the recommended 16 GB or more or the 16 GB for AI features. The Neo outputs video only from its single USB 3 port, so with a display connected you’re left with one USB 2 port. There’s no official memory figure for Photoshop’s AI features, so they aren’t judged. Pick one use. Each official figure is for that app on its own; running Lightroom Classic and Photoshop together has no official figure and isn’t judged. Other models that fit: 13-inch MacBook Air (M5)	Open with these settings
+ Photoshop, Sometimes (1.7 kg or less), 2	✓ The 13-inch MacBook Air (M5) meets the official minimum. Memory needed: 8 GB (Photoshop’s official minimum). Free space needed: 10 GB, 100 GB recommended (Adobe’s official figure; photos and catalogs not included). The weight limits (1.3 kg and 1.7 kg) are this article’s own choices. Lightest configuration: 13-inch MacBook Air (M5), 16 GB memory, 512 GB storage. Weight: 1.23 kg (Apple’s spec page). Memory headroom: 8 GB. The 512 GB drive is larger than that. Space for macOS and your photos isn’t calculated. Adobe recommends 16 GB or more (this check uses the minimum). The 13-inch MacBook Air (M5) drives up to 2 external displays. There’s no official memory figure for Photoshop’s AI features, so they aren’t judged. Pick one use. Each official figure is for that app on its own; running Lightroom Classic and Photoshop together has no official figure and isn’t judged. Other models that fit: 15-inch MacBook Air (M5), 14-inch MacBook Pro (M5), 14-inch MacBook Pro (M5 Pro), 14-inch MacBook Pro (M5 Max)	Open with these settings
+ AI features (Lightroom Classic), Daily (1.3 kg or less), None	△ The 13-inch MacBook Air (M5) just meets it. Memory needed: 16 GB (unified memory Adobe lists for Lightroom Classic’s AI features). Free space needed: 8 GB (Adobe’s official figure; photos and catalogs not included). The weight limits (1.3 kg and 1.7 kg) are this article’s own choices. Lightest configuration: 13-inch MacBook Air (M5), 16 GB memory, 512 GB storage. Weight: 1.23 kg (Apple’s spec page). Memory is exactly the official figure with nothing to spare (borderline). The 512 GB drive is larger than that. Space for macOS and your photos isn’t calculated. Adobe recommends 16 GB or more (this check uses the minimum). The 13-inch MacBook Air (M5) drives up to 2 external displays. Pick one use. Each official figure is for that app on its own; running Lightroom Classic and Photoshop together has no official figure and isn’t judged	Open with these settings
+ AI features (Lightroom Classic), Daily (1.3 kg or less), 2	△ The 13-inch MacBook Air (M5) just meets it. Memory needed: 16 GB (unified memory Adobe lists for Lightroom Classic’s AI features). Free space needed: 8 GB (Adobe’s official figure; photos and catalogs not included). The weight limits (1.3 kg and 1.7 kg) are this article’s own choices. Lightest configuration: 13-inch MacBook Air (M5), 16 GB memory, 512 GB storage. Weight: 1.23 kg (Apple’s spec page). Memory is exactly the official figure with nothing to spare (borderline). The 512 GB drive is larger than that. Space for macOS and your photos isn’t calculated. Adobe recommends 16 GB or more (this check uses the minimum). The 13-inch MacBook Air (M5) drives up to 2 external displays. Pick one use. Each official figure is for that app on its own; running Lightroom Classic and Photoshop together has no official figure and isn’t judged	Open with these settings
+ AI features (Lightroom Classic), Daily (1.3 kg or less), 3	✕ No model fits. Memory needed: 16 GB (unified memory Adobe lists for Lightroom Classic’s AI features). Free space needed: 8 GB (Adobe’s official figure; photos and catalogs not included). The weight limits (1.3 kg and 1.7 kg) are this article’s own choices. No model in this weight range supports 3 external displays. Set carrying to “Doesn’t matter” to see one	Open with these settings
+ AI features (Lightroom Classic), Sometimes (1.7 kg or less), 3	✓ The 14-inch MacBook Pro (M5 Pro) meets the official minimum. Memory needed: 16 GB (unified memory Adobe lists for Lightroom Classic’s AI features). Free space needed: 8 GB (Adobe’s official figure; photos and catalogs not included). The weight limits (1.3 kg and 1.7 kg) are this article’s own choices. Lightest configuration: 14-inch MacBook Pro (M5 Pro), 24 GB memory, 1 TB storage. Weight: 1.60 kg (Apple’s spec page). Memory headroom: 8 GB. The 1 TB drive is larger than that. Space for macOS and your photos isn’t calculated. Adobe recommends 16 GB or more (this check uses the minimum). The 14-inch MacBook Pro (M5 Pro) drives up to 3 external displays. Pick one use. Each official figure is for that app on its own; running Lightroom Classic and Photoshop together has no official figure and isn’t judged. Other models that fit: 14-inch MacBook Pro (M5 Max)	Open with these settings
+ AI features (Lightroom Classic), Doesn’t matter, 4	✓ The 14-inch MacBook Pro (M5 Max) meets the official minimum. Memory needed: 16 GB (unified memory Adobe lists for Lightroom Classic’s AI features). Free space needed: 8 GB (Adobe’s official figure; photos and catalogs not included). Lightest configuration: 14-inch MacBook Pro (M5 Max), 36 GB memory, 2 TB storage. Weight: 1.62 kg (Apple’s spec page). Memory headroom: 20 GB. The 2 TB drive is larger than that. Space for macOS and your photos isn’t calculated. Adobe recommends 16 GB or more (this check uses the minimum). The 14-inch MacBook Pro (M5 Max) drives up to 4 external displays. Pick one use. Each official figure is for that app on its own; running Lightroom Classic and Photoshop together has no official figure and isn’t judged. Other models that fit: 16-inch MacBook Pro (M5 Max)	Open with these settings
+ Lightroom Classic, Sometimes (1.7 kg or less), 4	✓ The 14-inch MacBook Pro (M5 Max) meets the official minimum. Memory needed: 8 GB (Lightroom Classic’s official minimum). Free space needed: 8 GB (Adobe’s official figure; photos and catalogs not included). The weight limits (1.3 kg and 1.7 kg) are this article’s own choices. Lightest configuration: 14-inch MacBook Pro (M5 Max), 36 GB memory, 2 TB storage. Weight: 1.62 kg (Apple’s spec page). Memory headroom: 28 GB. The 2 TB drive is larger than that. Space for macOS and your photos isn’t calculated. Adobe recommends 16 GB or more (this check uses the minimum). The 14-inch MacBook Pro (M5 Max) drives up to 4 external displays. Pick one use. Each official figure is for that app on its own; running Lightroom Classic and Photoshop together has no official figure and isn’t judged	Open with these settings

Which Logitech MX Mouse Should You Buy? Picks by Use

Wed, 07 Apr 2021 00:00:00 GMT

By this article’s rule of thumb, office work and programming call for the MX Master 4 with a thumb wheel (a side wheel for horizontal scrolling). Carrying calls for the 99 g MX Anywhere 3S, which clears the 120 g line.

The MX Master 4 beats the 3S Bluetooth Edition on two points. It comes with a Logi Bolt receiver, a USB receiver that plugs into the PC for a wireless link. It also has haptic feedback, which vibrates the mouse for actions you choose.

Nothing was tested on a real mouse, so comfort and speed aren’t judged.

Choosing by use and conditions

Carrying means 120 g or less, and the only current MX mouse that fits is the MX Anywhere 3S at 99 g. The Mac version weighs the same.

Condition	Matching models
Office work and programming (thumb wheel)	MX Master 4, MX Master 4 for Mac, MX Master 3S Bluetooth Edition
Carrying (this article’s rule: 120 g or less)	MX Anywhere 3S
Shape Logitech says reduces wrist pressure (vertical)	MX Vertical
Tilt Logitech says reduces forearm tension (trackball)	MX Ergo S
Switching between 2 devices	MX Master 4, MX Anywhere 3S, MX Anywhere 3S for Mac, MX Vertical, MX Ergo S (borderline)
Switching between 3 devices (borderline)	MX Master 4, MX Anywhere 3S, MX Anywhere 3S for Mac, MX Vertical

“Borderline” means the maximum device count equals the number you use (spare room under 10%).

Find the MX mouse that fits your setup

Pick your use (office and programming, carrying, or easing wrist strain), how many devices you switch between, your OS, how you connect (Bluetooth, a USB receiver or Logi Bolt), and whether you want quiet clicks or haptics. It lists the current MX mice whose official specs fit, and suggests one you can buy.

When wrist strain is the concern

To move the whole mouse by hand, pick the MX Vertical; to roll a ball with your thumb, pick the MX Ergo S trackball. Logitech says the Vertical’s 57° angle reduces wrist pressure and the Ergo S’s 20° tilt reduces forearm muscle tension. Percentages aren’t quoted because test conditions weren’t confirmed.

Per Logitech’s official pages as of October 2026, only the MX Vertical and MX Ergo S give a hand-size guide, and both say medium to large hands. Other models give none, so fit isn’t judged for them.

The MX Ergo S is right-handed only. For the other models, no statement on left-handed use turned up.

Using several computers and a Mac

The MX Master 4, MX Anywhere 3S and MX Vertical each pair with up to 3 devices. For three devices that’s exactly their maximum, so it’s borderline. The MX Ergo S pairs with up to 2 devices.

Per Logitech’s official pages, the MX Master 3S Bluetooth Edition and MX Master 4 for Mac listed no pairing count on the pages checked. They’re left out when you pick two or more devices.

The Mac versions list only macOS and iPadOS. If you also use Windows, pick the standard version, which supports macOS too.

The MX Master 4 for Mac and MX Anywhere 3S for Mac picks are US listings. As of October 2026, Logitech’s Japanese site has no MX Master 4 for Mac page.

MX Master 4 vs 3S, and the receiver

This compares the MX Master 4 with the MX Master 3S Bluetooth Edition. The standard 3S edition comes with a Logi Bolt receiver, but its availability wasn’t confirmed.

Per Logitech’s official pages, the MX Master 4 has haptic feedback and a Logi Bolt receiver in the box, which the 3S Bluetooth Edition lacks. It has 8 buttons to the 3S’s 7 and weighs 150 g to the 3S’s 141 g.

Haptic feedback is only on the MX Master 4 and its Mac version. Per the official page, it vibrates the mouse for specific actions, shortcuts and notifications, and it needs the Logi Options+ app.

If you don’t need the vibration cues or the receiver, the 3S Bluetooth Edition is lighter. Comfort isn’t judged.

A Logi Bolt receiver comes with the MX Master 4 and MX Ergo S. The MX Vertical includes a Unifying receiver, Logitech’s older receiver type. The MX Anywhere 3S and Mac versions take a Logi Bolt receiver sold separately, and the MX Master 3S Bluetooth Edition has none.

Logi Bolt is built for reliable connections where Wi-Fi and other wireless devices crowd the air. The MX Master 4 uses Bluetooth only on Windows 11 or later, so on Windows 10 connect it with the Logi Bolt receiver. If you’d rather skip the receiver, use Bluetooth.

One Logi Bolt receiver pairs with up to 6 Logi Bolt products. The MX Vertical uses Unifying, not Logi Bolt, so a Bolt receiver can’t connect it.

Main specs compared

DPI is the cursor speed setting, in dots per inch.

Per Logitech’s official pages, battery life is up to 70 days for the MX Master models and MX Anywhere 3S. The MX Ergo S is rated up to 120 days and the MX Vertical up to 4 months. These maximums vary with use, so they aren’t compared.

The MX Ergo S weight below is without its metal plate. With the plate it’s 259 g.

Model	Weight	Buttons (total)	Max devices	Battery (maker’s maximum)	Quick charge (maker’s figure)	DPI
MX Master 4	150 g	8	3 devices	up to 70 days	3 hours from a 1-minute charge	200–8,000 DPI
MX Master 3S Bluetooth Edition	141 g	7	Not found on the official page	up to 70 days	3 hours from a 1-minute charge	200–8,000 DPI
MX Anywhere 3S	99 g	6	3 devices	up to 70 days	3 hours from a 1-minute charge	200–8,000 DPI
MX Vertical	125 g	Total not found (4 customizable)	3 devices	up to 4 months	3 hours from a 1-minute charge	400–4,000 DPI
MX Ergo S	164 g	8	2 devices	up to 120 days	24 hours from a 1-minute charge	512–2,048 DPI

Every current MX mouse lists quiet clicks except the MX Vertical and MX Master 4 for Mac, which don’t say.

Picks by condition

One pick per condition, from Amazon.com:

Sold new by Amazon.com
At least 100 ratings and 4 stars or more
A model that matches Logitech’s official product

Prices vary, so check them on the retailer’s page.

With a receiver in the box, for office work and programming

※ The Pale Gray version is a separate listing (B0FC5V3YVY, sold by a third party). Both pages show the same 2,118 ratings, so the count is pooled across colors.

Without haptic feedback, for office work and programming

It’s the previous generation, with no receiver in the box, and Logi Bolt support wasn’t found on the official page.

Whether the rating count is shared with other models on the listing wasn't checked.

Carrying (this article’s rule

At 99 g, it clears 120 g by 21 g.

Rose and Pale Grey versions are separate listings (for example B0BPY51TZ3) showing the same 2,312 ratings, so the count is pooled across colors.

For Mac users, office work and programming

※ A White Silver listing (B0FC5C7VCJ, sold by Amazon) shows the same 578 ratings, so the count is pooled across colors. Logitech US lists only Space Black for this model.

For Mac users, carrying

Vertical mouse, whose 57° angle Logitech says reduces wrist pressure

Thumb-operated trackball, whose 20° tilt Logitech says reduces forearm tension

FAQ

Which MX mouse is best for office work and programming?

Office work and programming need a thumb wheel. The MX Master 4, MX Master 4 for Mac and MX Master 3S Bluetooth Edition have one.

If you want a USB receiver in the box and haptic feedback (vibration for chosen actions), pick the MX Master 4.

MX Anywhere 3S or an MX Master for carrying?

Carrying means 120 g or less, and the only current MX mouse that fits is the MX Anywhere 3S at 99 g. The Mac version weighs the same.

Vertical mouse or trackball for wrist strain?

Percentages aren’t quoted because test conditions weren’t confirmed.

Which MX mouse works best across several computers?

The MX Master 4, MX Anywhere 3S and MX Vertical each pair with up to 3 devices. For three devices that’s exactly their maximum, so it’s borderline.

The MX Ergo S pairs with up to 2 devices. Per Logitech’s official pages as of October 2026, the MX Master 3S Bluetooth Edition and MX Master 4 for Mac listed no pairing count on the pages checked.

They’re left out when you pick two or more devices.

How does the MX Master 4 differ from the MX Master 3S, and is the 3S still sold?

Yes, Amazon.com still lists the MX Master 3S Bluetooth Edition as new stock (checked October 2026). Per Logitech’s official pages as of October 2026, the MX Master 4 has haptic feedback and a Logi Bolt receiver in the box, which the 3S Bluetooth Edition lacks.

It has 8 buttons to the 3S’s 7 and weighs 150 g to the 3S’s 141 g. The standard 3S edition comes with a Logi Bolt receiver, but its availability wasn’t confirmed, and the Bluetooth Edition is the 3S pick here.

If you want a receiver in the box, pick the MX Master 4.

Do MX mice work on a glass desk, and is the DPI (cursor speed setting) the same on all of them?

Logitech says the MX Master 4, MX Master 3S Bluetooth Edition and MX Anywhere 3S track on glass. The minimum glass thickness is 4 mm for the MX Master 4 and, on the US page, for the MX Master 3S Bluetooth Edition.

No thickness figure turned up for the MX Anywhere 3S. Most MX mice run 200–8,000 DPI, while the MX Vertical and MX Ergo S top out lower.

The spec table’s DPI column lists each model. Whether that’s enough depends on your monitor and taste, so no cutoff is set.

Data and formulas used (for verification)

How this was checked:

Sources: MX Master 4 and MX Master 3S ／ MX Anywhere 3S, MX Vertical and MX Ergo S ／ MX Master 4 for Mac (Logitech US)

Method

Covers the MX mice listed on Logitech’s official US and Japanese stores on 2026-10-08: MX Master 4, MX Master 4 for Mac (US only), MX Master 3S Bluetooth Edition, MX Anywhere 3S and its for Mac version, MX Ergo S and MX Vertical. Specs are from each official product page; none were tested in hand.
Office and programming = models with a thumb wheel for horizontal scrolling (this article’s condition). Carrying = 120 g or less (this article’s cutoff; the MX Anywhere 3S weighs 99 g and the next lightest, the MX Vertical, 125 g). Wrist strain = the shapes Logitech describes as easing strain: vertical (MX Vertical, 57 degrees) and trackball (MX Ergo S, 0–20° tilt). The maker’s strain-reduction percentages aren’t used because their test conditions weren’t confirmed.
Devices = the official maximum pairing count (MX Master 4 3, MX Anywhere 3S 3, MX Vertical 3, MX Ergo S 2) is at least the number you use. The MX Master 3S Bluetooth Edition and MX Master 4 for Mac didn’t state a count and are left out for two or more. A count equal to the maximum (headroom under 10% of your count) is marked borderline.
Receiver = one in the box (Logi Bolt for the MX Master 4 and MX Ergo S, Unifying for the MX Vertical). Use with Logi Bolt = models stated to support Logi Bolt, including sold-separately receivers (the MX Vertical isn’t Logi Bolt; the MX Master 3S Bluetooth Edition doesn’t say and is left out). One Logi Bolt receiver pairs with up to 6 Logi Bolt products.
Quiet clicks, haptics and OS support follow the official pages. Specs that couldn’t be confirmed (quiet clicks on the MX Vertical and MX Master 4 for Mac) aren’t treated as missing; those models drop out only when you require the feature. The for Mac models list only macOS and iPadOS, so they drop out for Windows or both. Hand-size guidance exists only for the MX Vertical and MX Ergo S (medium to large); fit isn’t judged for the others.
Battery figures are Logitech’s maximums (MX Master 4 70 days, MX Anywhere 3S 70 days, MX Ergo S 120 days, MX Vertical up to 4 months) and vary with use, so they aren’t compared. Picks are new listings sold by Amazon with 100+ ratings and 4+ stars whose model matches the official product, cheapest including shipping.

Official values used

Item	Value	Source	Checked
MX Master 4 height	128.15 mm	logicool.co.jp	2026-10-08
MX Master 4 width	88.35 mm	logicool.co.jp	2026-10-08
MX Master 4 depth	50.8 mm	logicool.co.jp	2026-10-08
MX Master 4 weight	150 g	logicool.co.jp	2026-10-08
MX Master 3S Bluetooth Edition height	124.9 mm	logicool.co.jp	2026-10-08
MX Master 3S Bluetooth Edition width	84.3 mm	logicool.co.jp	2026-10-08
MX Master 3S Bluetooth Edition depth	51 mm	logicool.co.jp	2026-10-08
MX Master 3S Bluetooth Edition weight	141 g	logicool.co.jp	2026-10-08
MX Anywhere 3S height	100.5 mm	logicool.co.jp	2026-10-08
MX Anywhere 3S width	65 mm	logicool.co.jp	2026-10-08
MX Anywhere 3S depth	34.4 mm	logicool.co.jp	2026-10-08
MX Anywhere 3S weight	99 g	logicool.co.jp	2026-10-08
MX Anywhere 3S for Mac height	100.5 mm	logicool.co.jp	2026-10-08
MX Anywhere 3S for Mac width	65 mm	logicool.co.jp	2026-10-08
MX Anywhere 3S for Mac depth	34.4 mm	logicool.co.jp	2026-10-08
MX Anywhere 3S for Mac weight	99 g	logicool.co.jp	2026-10-08
MX Vertical height	120 mm	logicool.co.jp	2026-10-08
MX Vertical width	79 mm	logicool.co.jp	2026-10-08
MX Vertical depth	78.5 mm	logicool.co.jp	2026-10-08
MX Vertical weight	125 g	logicool.co.jp	2026-10-08
MX Master 4 for Mac height	128.15 mm	logitech.com Logitech US official MX Master 4 for Mac page (specs / features): Height 5.05 in	2026-10-08
MX Master 4 for Mac width	88.35 mm	logitech.com Logitech US official MX Master 4 for Mac page (specs / features): Width 3.48 in	2026-10-08
MX Master 4 for Mac depth	50.8 mm	logitech.com Logitech US official MX Master 4 for Mac page (specs / features): Depth 2 in (50	2026-10-08
MX Master 4 for Mac weight	150 g	logitech.com Logitech US official MX Master 4 for Mac page (specs / features): Weight 5.29 oz	2026-10-08
MX Ergo S height	132.5 mm	logicool.co.jp	2026-10-08
MX Ergo S width	99.8 mm	logicool.co.jp	2026-10-08
MX Ergo S depth	51.4 mm	logicool.co.jp	2026-10-08
MX Ergo S weight (without metal plate or receiver)	164 g	logicool.co.jp	2026-10-08
MX Ergo S weight (with metal plate, without receiver)	259 g	logicool.co.jp	2026-10-08
MX Master 4 sensor minimum DPI	200 DPI	logicool.co.jp	2026-10-08
MX Master 4 sensor maximum DPI	8000 DPI	logicool.co.jp	2026-10-08
MX Master 4 for Mac sensor minimum DPI	200 DPI	logitech.com Logitech US official MX Master 4 for Mac page (specs / features): 200-8000 (incr	2026-10-08
MX Master 4 for Mac sensor maximum DPI	8000 DPI	logitech.com Logitech US official MX Master 4 for Mac page (specs / features): 200-8000 (incr	2026-10-08
MX Master 3S Bluetooth Edition sensor minimum DPI	200 DPI	logicool.co.jp	2026-10-08
MX Master 3S Bluetooth Edition sensor maximum DPI	8000 DPI	logicool.co.jp	2026-10-08
MX Anywhere 3S sensor minimum DPI	200 DPI	logicool.co.jp	2026-10-08
MX Anywhere 3S sensor maximum DPI	8000 DPI	logicool.co.jp	2026-10-08
MX Anywhere 3S for Mac sensor minimum DPI	200 DPI	logicool.co.jp	2026-10-08
MX Anywhere 3S for Mac sensor maximum DPI	8000 DPI	logicool.co.jp	2026-10-08
MX Ergo S sensor minimum DPI	512 DPI	logicool.co.jp	2026-10-08
MX Ergo S sensor maximum DPI	2048 DPI	logicool.co.jp	2026-10-08
MX Vertical sensor minimum DPI	400 DPI	logicool.co.jp	2026-10-08
MX Vertical sensor maximum DPI	4000 DPI	logicool.co.jp	2026-10-08
MX Master 4 number of buttons	8	logicool.co.jp	2026-10-08
MX Master 4 for Mac number of buttons	8	logitech.com Logitech US official MX Master 4 for Mac page (specs / features): Number of Butt	2026-10-08
MX Master 3S Bluetooth Edition number of buttons	7	logicool.co.jp	2026-10-08
MX Anywhere 3S number of buttons	6	logicool.co.jp	2026-10-08
MX Anywhere 3S for Mac number of buttons	6	logicool.co.jp	2026-10-08
MX Ergo S number of buttons	8	logicool.co.jp	2026-10-08
MX Ergo S number of customizable buttons	6	logicool.co.jp	2026-10-08
MX Vertical number of customizable buttons	4	logicool.co.jp	2026-10-08
MX Master 4 scroll wheels	MagSpeed wheel with SmartShift; thumb wheel (horizontal scroll)	logicool.co.jp	2026-10-08
MX Master 4 for Mac scroll wheels	MagSpeed wheel with SmartShift; thumb wheel	logitech.com Logitech US official MX Master 4 for Mac page (specs / features): Scroll Wheel:	2026-10-08
MX Master 3S Bluetooth Edition scroll wheels	Scroll wheel with SmartShift (MagSpeed electromagnetic); thumb wheel	logicool.co.jp	2026-10-08
MX Anywhere 3S scroll wheel	MagSpeed scroll wheel with auto-shift; no thumb wheel (horizontal scroll: hold the side button and turn the wheel)	logicool.co.jp	2026-10-08
MX Anywhere 3S for Mac scroll wheel	MagSpeed scroll wheel with auto-shift; no thumb wheel (horizontal scroll: hold the side button and turn the wheel)	logicool.co.jp	2026-10-08
MX Ergo S wheel and trackball	Scroll wheel with left/right tilt; thumb-operated trackball	logicool.co.jp	2026-10-08
MX Vertical scroll wheel	Precision wheel	logicool.co.jp	2026-10-08
MX Master 4 number of devices that can be paired	3	logicool.co.jp	2026-10-08
MX Anywhere 3S number of devices that can be paired	3	logicool.co.jp	2026-10-08
MX Anywhere 3S for Mac number of devices that can be paired	3	logicool.co.jp	2026-10-08
MX Ergo S number of devices that can be paired	2	logicool.co.jp	2026-10-08
MX Vertical number of devices that can be paired	3	logicool.co.jp	2026-10-08
MX Master 4 battery life on a full charge (up to)	70 days	logicool.co.jp	2026-10-08
MX Master 4 for Mac battery life on a full charge (up to)	70 days	logitech.com Logitech US official MX Master 4 for Mac page (specs / features): Battery life:	2026-10-08
MX Master 3S Bluetooth Edition battery life on a full charge	70 days	logicool.co.jp	2026-10-08
MX Anywhere 3S battery life on a full charge (up to)	70 days	logicool.co.jp	2026-10-08
MX Anywhere 3S for Mac battery life on a full charge (up to)	70 days	logicool.co.jp	2026-10-08
MX Ergo S battery life on a full charge (up to)	120 days	logicool.co.jp	2026-10-08
MX Vertical battery life	up to 4 months on a full charge with daily use; 3 hours from a 1-minute charge	logicool.co.jp	2026-10-08
MX Master 4 quick charge	3 hours of use from a 1-minute charge (USB-C; cable not included)	logicool.co.jp	2026-10-08
MX Master 4 connection	Bluetooth Low Energy 5.1, or the USB-C Logi Bolt receiver in the box (also works in USB-A ports)	logicool.co.jp	2026-10-08
MX Master 4 for Mac connection	Bluetooth Low Energy 5.1. Logi Bolt receiver supported but not included	logitech.com Logitech US official MX Master 4 for Mac page (specs / features): Package: Mouse	2026-10-08
MX Master 3S Bluetooth Edition connection	Bluetooth Low Energy (Bluetooth Edition, no receiver). The standard edition includes a Logi Bolt receiver	logicool.co.jp	2026-10-08
MX Anywhere 3S connection	Bluetooth Low Energy. Works with the Logi Bolt receiver (sold separately, not in the box)	logicool.co.jp	2026-10-08
MX Anywhere 3S for Mac connection	Bluetooth Low Energy. Works with the Logi Bolt receiver (sold separately)	logicool.co.jp	2026-10-08
MX Ergo S connection	Bluetooth Low Energy, or the Logi Bolt USB receiver in the box	logicool.co.jp	2026-10-08
MX Vertical connection	Bluetooth Low Energy 4.2, or the Unifying USB receiver in the box (not Logi Bolt)	logicool.co.jp	2026-10-08
MX Master 4 supported OS	Bluetooth: Windows 11 or later, macOS 13 or later, Linux, ChromeOS, iPadOS 15 or later, Android 12 or later. Logi Bolt: Windows 10/11 or later, macOS 13 or later. Logi Options+ app is for Windows and macOS	logicool.co.jp	2026-10-08
MX Master 4 for Mac supported OS	Bluetooth: macOS 13 or later, iPadOS 15 or later	logitech.com Logitech US official MX Master 4 for Mac page (specs / features): Bluetooth: Req	2026-10-08
MX Master 3S Bluetooth Edition supported OS	Windows 10/11 or later, macOS 10.15 or later, Linux, ChromeOS, iPadOS 14 or later, Android 8.0 or later (Bluetooth). Logi Options+ is for Windows and macOS	logicool.co.jp	2026-10-08
MX Anywhere 3S supported OS	Windows 10/11 or later, macOS 10.15 or later, Linux, ChromeOS, iPadOS 14 or later, Android 8.0 or later	logicool.co.jp	2026-10-08
MX Anywhere 3S for Mac supported OS	macOS 10.15 or later, iPadOS 14 or later	logicool.co.jp	2026-10-08
MX Ergo S supported OS	Windows 10/11 or later, macOS 12 or later, Linux, ChromeOS. Over Bluetooth also iPadOS 15 or later and Android 12 or later	logicool.co.jp	2026-10-08
MX Vertical supported OS	Windows 10/11 or later, macOS 10.15 or later, ChromeOS, Linux. Over Bluetooth also iPadOS 13.4 or later (AssistiveTouch required)	logicool.co.jp	2026-10-08
MX Master 4 haptic feedback	Yes. Customizable haptic feedback on specific actions, shortcuts and notifications (needs Logi Options+; recent Windows and macOS only)	logicool.co.jp	2026-10-08
MX Master 4 for Mac haptic feedback	Yes (same wording on the product page)	logitech.com Logitech US official MX Master 4 for Mac page (specs / features): Haptic Sense P	2026-10-08
MX Master 4 quiet clicks	Yes. 90% lower sound output on left/right clicks compared with MX Master 3 (maker figure)	logicool.co.jp	2026-10-08
MX Master 3S Bluetooth Edition quiet clicks	Yes (Quiet Clicks). 90% less sound power level on left/right click than MX Master 3, measured at 1 m (maker figure)	logicool.co.jp	2026-10-08
MX Anywhere 3S quiet clicks	Yes (quieter clicks than MX Anywhere 3)	logicool.co.jp	2026-10-08
MX Anywhere 3S for Mac quiet clicks	Yes (quieter clicks than MX Anywhere 3)	logicool.co.jp	2026-10-08
MX Ergo S quiet clicks	Yes (quiet clicks)	logicool.co.jp	2026-10-08
MX Master 4 tracking on glass	Tracks on glass (4 mm minimum glass thickness)	logicool.co.jp	2026-10-08
MX Master 3S Bluetooth Edition tracking on glass	Tracks virtually anywhere, even on glass (4 mm minimum glass thickness)	logicool.co.jp	2026-10-08
MX Anywhere 3S tracking on glass	Tracks on glass (minimum glass thickness not captured)	logicool.co.jp	2026-10-08
MX Anywhere 3S for Mac tracking on glass	Tracks on glass (minimum glass thickness not captured)	logicool.co.jp	2026-10-08
MX Ergo S left-handed version	No. Currently made for right-handed use only	logicool.co.jp	2026-10-08
MX Ergo S hand size	Fits medium to large hands comfortably	logicool.co.jp	2026-10-08
MX Vertical hand size	Best for medium to large hands	logicool.co.jp	2026-10-08
MX Vertical vertical angle	57 degrees	logicool.co.jp	2026-10-08
MX Ergo S tilt and the maker’s wrist/forearm statement	Adjustable hinge tilts the trackball between 0 and 20 degrees; the maker says the 20-degree tilt reduces forearm muscle strain	logicool.co.jp	2026-10-08
MX Master 4 shape statement	The maker says the tilt angle is optimized for a natural posture (no angle figure captured)	logicool.co.jp	2026-10-08
MX Vertical left-handed version	not found on the product page	logicool.co.jp	2026-10-08
number of Logi Bolt mice/keyboards one Logi Bolt receiver can pair	6	hub.sync.logitech.com Users can pair up to six Logi Bolt wireless mice and keyboards to a single Logi	2026-10-08
Logi Bolt connection security (maker statement)	Bluetooth Security Mode 1, Level 4 (Secure Connections Only), described by Logitech as FIPS compliant	hub.sync.logitech.com Bluetooth security mode 1, level 4 (also known as Secure Connections Only mode),	2026-10-08
who a Logi Bolt receiver connects to	A Logi Bolt receiver emits an encrypted signal and connects only with Logi Bolt products; it cannot be paired with non-Logi Bolt devices	hub.sync.logitech.com A Logi Bolt receiver is emitting an encrypted signal that only connects with Log	2026-10-08
Logi Bolt wireless range (maker figure)	10 m	logitech.com up to 10 meters wireless range	2026-10-08
Logi Bolt and radio interference (maker statement)	Logitech says it is engineered for reliable connections even in wireless environments congested with interference from Wi-Fi access points or surrounding wireless devices	logitech.com engineered for reliable connections, even in wireless environments congested wit	2026-10-08
MX Master 4 for Mac on Logitech’s Japanese site	Not found on Logitech’s Japanese site (the product page URL returns 404)	logicool.co.jp	2026-10-08
MX Master 3S Bluetooth Edition availability on Logitech’s Japanese site	The product page is on Logitech’s Japanese site and shows it in stock	logicool.co.jp	2026-10-08

Find the MX mouse that fits your setup: quick reference

The calculator in this article shows the result and a matching product for your own settings. Settings can be set in the URL; each link below opens the calculator with those settings.

Settings	Result	Calculator
Office & programming, 1 device, Windows and Mac, Bluetooth is fine, Quiet clicks optional, Haptics optional	✓ Matches: MX Master 4, MX Master 3S Bluetooth Edition. Office and programming here means a thumb wheel for horizontal scrolling (this article’s condition). MX Master 4: 150 g, 8 buttons, Bluetooth, Logi Bolt receiver included. MX Master 3S Bluetooth Edition: 141 g, 7 buttons, Bluetooth (no receiver). The for Mac models list only macOS and iPadOS, so if you also use Windows, pick the standard version	Open with these settings
Office & programming, 3 devices, Windows and Mac, Want a USB receiver, Quiet clicks optional, Haptics optional	△ MX Master 4 (device count is borderline). Office and programming here means a thumb wheel for horizontal scrolling (this article’s condition). MX Master 4: 150 g, 8 buttons, Bluetooth, Logi Bolt receiver included, up to 3 devices. Your device count is exactly the MX Master 4’s maximum, so it’s borderline	Open with these settings
Office & programming, 1 device, Mac, Bluetooth is fine, Quiet clicks optional, Want haptic feedback	✓ Matches: MX Master 4, MX Master 4 for Mac. Office and programming here means a thumb wheel for horizontal scrolling (this article’s condition). MX Master 4: 150 g, 8 buttons, Bluetooth, Logi Bolt receiver included. MX Master 4 for Mac: 150 g, 8 buttons, Bluetooth (Logi Bolt receiver sold separately). Haptic feedback needs Logi Options+ (recent Windows and macOS only)	Open with these settings
Office & programming, 2 devices, Windows, Bluetooth is fine, Want quiet clicks, Haptics optional	✓ Matches: MX Master 4. Office and programming here means a thumb wheel for horizontal scrolling (this article’s condition). MX Master 4: 150 g, 8 buttons, Bluetooth, Logi Bolt receiver included, up to 3 devices. Left out because the official page didn’t state it: MX Master 3S Bluetooth Edition (device count). Using the MX Master 4 over Bluetooth needs Windows 11 or later (Windows 10 works with the Logi Bolt receiver)	Open with these settings
Office & programming, 1 device, Windows, Use with Logi Bolt, Quiet clicks optional, Haptics optional	✓ Matches: MX Master 4. Office and programming here means a thumb wheel for horizontal scrolling (this article’s condition). MX Master 4: 150 g, 8 buttons, Bluetooth, Logi Bolt receiver included. Left out because the official page didn’t state it: MX Master 3S Bluetooth Edition (Logi Bolt support). The for Mac models list only macOS and iPadOS, so if you also use Windows, pick the standard version. One Logi Bolt receiver pairs with up to 6 Logi Bolt products	Open with these settings
Office & programming, 3 devices, Mac, Bluetooth is fine, Want quiet clicks, Want haptic feedback	△ MX Master 4 (device count is borderline). Office and programming here means a thumb wheel for horizontal scrolling (this article’s condition). MX Master 4: 150 g, 8 buttons, Bluetooth, Logi Bolt receiver included, up to 3 devices. Your device count is exactly the MX Master 4’s maximum, so it’s borderline. Left out because the official page didn’t state it: MX Master 4 for Mac (device count, quiet clicks). Haptic feedback needs Logi Options+ (recent Windows and macOS only)	Open with these settings
Carrying, 1 device, Windows and Mac, Bluetooth is fine, Quiet clicks optional, Haptics optional	✓ Matches: MX Anywhere 3S. Carrying here means 120 g or less (this article’s cutoff). MX Anywhere 3S: 99 g, 6 buttons, Bluetooth (Logi Bolt receiver sold separately). The for Mac models list only macOS and iPadOS, so if you also use Windows, pick the standard version	Open with these settings
Carrying, 3 devices, Mac, Use with Logi Bolt, Want quiet clicks, Haptics optional	△ MX Anywhere 3S, MX Anywhere 3S for Mac (device count is borderline). Carrying here means 120 g or less (this article’s cutoff). MX Anywhere 3S: 99 g, 6 buttons, Bluetooth (Logi Bolt receiver sold separately), up to 3 devices. Your device count is exactly the MX Anywhere 3S’s maximum, so it’s borderline. MX Anywhere 3S for Mac: 99 g, 6 buttons, Bluetooth (Logi Bolt receiver sold separately), up to 3 devices. Your device count is exactly the MX Anywhere 3S for Mac’s maximum, so it’s borderline. One Logi Bolt receiver pairs with up to 6 Logi Bolt products	Open with these settings
Carrying, 1 device, Windows, Want a USB receiver, Quiet clicks optional, Haptics optional	✕ No model fits. Carrying here means 120 g or less (this article’s cutoff)	Open with these settings
Ease wrist strain, Vertical, 1 device, Windows, Want a USB receiver, Quiet clicks optional, Haptics optional	✓ Matches: MX Vertical. Logitech says the MX Vertical’s 57° vertical angle reduces wrist pressure. The maker’s percentages aren’t quoted because their test conditions weren’t confirmed. MX Vertical: 125 g, 4 customizable buttons, Bluetooth, Unifying receiver included (not Logi Bolt). Logitech rates the MX Vertical for medium to large hands	Open with these settings
Ease wrist strain, Trackball, 2 devices, Windows and Mac, Want a USB receiver, Want quiet clicks, Haptics optional	△ MX Ergo S (device count is borderline). Logitech says the MX Ergo S’s 20° tilt reduces forearm muscle strain. The maker’s percentages aren’t quoted because their test conditions weren’t confirmed. MX Ergo S: 164 g, 8 buttons, Bluetooth, Logi Bolt receiver included, up to 2 devices. Your device count is exactly the MX Ergo S’s maximum, so it’s borderline. Logitech rates the MX Ergo S for medium to large hands	Open with these settings
Ease wrist strain, Either, 3 devices, Windows and Mac, Bluetooth is fine, Quiet clicks optional, Haptics optional	△ MX Vertical (device count is borderline). Logitech says the MX Vertical’s 57° vertical angle reduces wrist pressure. Logitech says the MX Ergo S’s 20° tilt reduces forearm muscle strain. The maker’s percentages aren’t quoted because their test conditions weren’t confirmed. MX Vertical: 125 g, 4 customizable buttons, Bluetooth, Unifying receiver included (not Logi Bolt), up to 3 devices. Your device count is exactly the MX Vertical’s maximum, so it’s borderline. Logitech rates the MX Vertical for medium to large hands. Pairs with fewer devices than you need: MX Ergo S (up to 2)	Open with these settings
Ease wrist strain, Trackball, 3 devices, Windows, Bluetooth is fine, Quiet clicks optional, Haptics optional	✕ No model fits. Logitech says the MX Ergo S’s 20° tilt reduces forearm muscle strain. The maker’s percentages aren’t quoted because their test conditions weren’t confirmed. Pairs with fewer devices than you need: MX Ergo S (up to 2)	Open with these settings

Is the iPad mini Enough? Screen Size vs Paper Size

Mon, 05 Apr 2021 00:00:00 GMT

For reading, notes and pressure-sensitive handwriting with pages up to A6 shown at actual size, the smallest and lightest current 8.3 in model is enough. Among current iPads, a whole A5 page at actual size needs a 10.86 in or larger screen, and 10.86 in and 11.1 in are borderline. That compares the whole screen with the page, not the app’s drawing area after toolbars.

This compares the iPads Apple sells now, checked as of October 2026 against Apple’s spec pages and CLIP STUDIO PAINT’s official requirements. Nothing was tested on a real iPad, so speed and comfort aren’t judged. “Fits” means a page shown whole at actual size, so an A5 PDF can still be read zoomed or scrolled.

How to compare screen size and paper size

The iPad mini’s screen is about 176 × 116 mm. Compared as the whole screen before app toolbars, A6 fits at actual size, but A5 doesn’t. The real display area has rounded corners and is slightly smaller.

Apple doesn’t publish screen dimensions, so these are calculated from the diagonal and resolution.

Paper size	iPad mini	Base iPad (A16 chip)	11-inch iPad Air
A6 (105 × 148 mm)	Fits whole at actual size	Fits whole at actual size	Fits whole at actual size
A5 (148 × 210 mm)	Doesn’t fit whole at actual size	Fits (borderline)	Fits (borderline)
A4 (210 × 297 mm)	Doesn’t fit whole at actual size	Doesn’t fit whole at actual size	Doesn’t fit whole at actual size

“Borderline” means the spare room is under 10% of the paper’s side.

The fit compares the whole screen with the page, not the app’s drawing area after toolbars. Calculated from Apple’s diagonal and resolution, no current iPad shows a whole A4 page at actual size. Even the 13-inch screens have a short side of about 198 mm, short of A4’s 210 mm.

Pick a use and the weight you’ll carry, and the tool shows the first model that fits.

Is the iPad mini enough for your use?

Pick your use, how much weight you'll carry and a paper size. It checks whether the iPad mini meets Apple's specs and CLIP STUDIO PAINT's official requirements, and shows the first model that fits (checking iPad mini, iPad, iPad Air, iPad Pro in that order), plus one iPad you can buy that fits.

Apple Pencil pressure and hover

In Apple’s comparison table (Pro vs. USB-C), pressure sensing is on Apple Pencil Pro, not on the USB-C model. The iPad mini supports the Pro, but the base iPad (A16 chip) doesn’t.

Per Apple’s Apple Pencil pages, the iPad mini supports Apple Pencil hover. No hover support was found for the base iPad (A16 chip) on Apple’s pages.

Only the base iPad (A16 chip) can use the Apple Pencil 1st generation, with an adapter. The iPad mini can’t use the 1st or 2nd generation. Apple’s table doesn’t cover the 1st generation, so its pressure support isn’t judged there, but CLIP STUDIO PAINT’s list includes it as a pressure pen.

When the base iPad or the 11-inch iPad Air makes sense

Among current iPads, a whole A5 page at actual size needs 10.86 in diagonal or more, judged on the whole screen, not the app’s drawing area. At 10.86 in and 11.1 in the margin is under 10%, so it’s borderline. From 12.9 in up there’s room.

For pressure-sensitive handwriting with A5 at actual size, the 11-inch iPad Air just fits, so it’s borderline. Without pressure sensing, the base iPad (A16 chip) also just fits.

CLIP STUDIO PAINT notes

Per CLIP STUDIO PAINT’s official requirements, the iPad mini is on the supported-device list for standard-size work, not the high-resolution list. The listed pressure pens include Apple Pencil Pro. The screen is below the recommended 10.5 in, and the required 3 GB of memory can’t be judged from Apple’s specs.

The iPad Air (M4) isn’t on the supported-device page, so this article doesn’t judge it.

Apple Pencil (USB-C) isn’t in the list of pressure-sensitive pens.

The listed OS versions are iPadOS 18 and iPadOS 26; the current iPadOS 27 wasn’t found there.

Main specs compared (weight, storage, Apple Pencil, memory)

Comparing Apple’s current iPads on its spec pages (October 2026), the iPad mini is the smallest and lightest model at 293 g, 171 g lighter than the 11-inch iPad Air.

Model	Screen diagonal	Weight	Storage	Apple Pencil	Memory
iPad mini	8.3 in	293 g	128 GB, 256 GB, 512 GB	Pro, USB-C	Not stated by Apple
Base iPad (A16 chip)	10.86 in	477 g	128 GB, 256 GB, 512 GB	USB-C, 1st gen (adapter)	Not stated by Apple
11-inch iPad Air	10.86 in	464 g	128 GB, 256 GB, 512 GB, 1 TB	Pro, USB-C	12 GB

Weights are for the Wi-Fi models. The iPad mini with Wi-Fi + Cellular weighs 297 g.
Apple’s tests put battery life at up to 10 hours. That’s Wi-Fi web browsing and video playback, not time spent drawing.
Comparing Apple’s current iPads on its spec pages (October 2026), ProMotion is listed only for the iPad Pro; it wasn’t found for the iPad mini.
The iPad mini can connect an external display over USB-C. Apple’s spec page lists native DisplayPort output over USB-C, and VGA, HDMI and DVI output with adapters (sold separately).
Per Apple’s spec pages, memory for the iPad mini and the base iPad (A16 chip) wasn’t found. It is listed for the iPad Air and the iPad Pro.
Per Apple’s spec pages, this article judges only official specs and requirements; speed and feel weren’t tested on real devices and aren’t judged.

Picks by condition

One pick per condition, from Amazon.com:

Sold new by Amazon.com, with a configuration that matches Apple’s spec
At least 100 ratings and 4 stars or more
If several listings qualify, the cheapest of those checked

Prices vary, so check them on the retailer’s page.

Reading and notes without pressure sensing, or pressure-sensitive handwriting with Apple Pencil Pro, 350 g or less, pages up to A6 at actual size

Calculated from diagonal and resolution, the screen is about 176 × 116 mm, so A5 doesn’t fit.

Color variants are listed; price may vary by color. The rating count is shared with other models on the same listing.

A5 paper without pressure sensing, or a screen at or above CLIP STUDIO PAINT’s recommended 10.5 in

The A5 margin is under 10% of the paper’s side, so it’s borderline. Per CLIP STUDIO PAINT’s official requirements, this model is on the standard-size list, and its screen meets the recommended 10.5 in by a small margin, so it’s borderline. The required 3 GB of memory can’t be judged from Apple’s specs. The pressure pen is the 1st generation with an adapter, and Apple Pencil (USB-C) isn’t on the pressure-pen list.

The rating count is shared with other models on the same listing.

Pressure-sensitive handwriting with A5 at actual size

The A5 margin is small here too, so it’s borderline.

The rating count is shared with other models on the same listing.

FAQ

Should you buy the iPad mini?

That compares the whole screen with the page, not the app’s drawing area after toolbars.

What’s the difference between the iPad Air and the iPad mini?

The current model with an 8.3 in screen is the smaller, lighter one at 293 g. The current 11-inch model with the M4 chip is 10.86 in diagonal and 464 g, with 12 GB of memory per Apple’s specs.

No memory figure was found for the 8.3 in model. Which is which is in the main specs table.

Base iPad or iPad mini?

Among current iPads, for pressure-sensitive handwriting with pages up to A6 at actual size, pick the 8.3 in model. For A5 at actual size with pressure, pick the 11-inch model with the M4 chip, which just fits; without pressure sensing, the A16-chip model also just fits.

Per Apple’s Apple Pencil pages, the current model with an 8.3 in screen supports Apple Pencil hover. No hover support was found for the current A16-chip model on Apple’s pages.

The models are in the main specs table.

How much storage should you get on the iPad mini?

Storage on the current model with an 8.3 in screen comes in three sizes: 128 GB, 256 GB, 512 GB. Every option exceeds CLIP STUDIO PAINT’s 6 GB free-space requirement, but Apple gives no guide to how much each use needs, so that isn’t calculated.

Which Apple Pencil works with the iPad mini?

Apple Pencil Pro and Apple Pencil (USB-C) both work with the current model with an 8.3 in screen, but the 1st and 2nd generations don’t. Only the Pro senses pressure, per Apple’s comparison table (Pro vs. USB-C), so the USB-C model is enough without it.

Is the iPad mini’s screen too small? Can you draw on it?

Calculated from Apple’s diagonal and resolution, an 8.3 in diagonal screen works out to about 176 × 116 mm. Compared as the whole screen before app toolbars, A6 fits at actual size, but A5 doesn’t.

The real display area has rounded corners and is slightly smaller. Per CLIP STUDIO PAINT’s official requirements, the current model with an 8.3 in screen is on the supported-device list for standard-size work, not the high-resolution list.

The listed pressure pens include Apple Pencil Pro. The screen is below the recommended 10.5 in, and the required 3 GB of memory can’t be judged from Apple’s specs.

Nothing was tested on a real iPad, so comfort isn’t judged.

Data and formulas used (for verification)

How this was checked:

Sources: Apple specs for iPad mini, iPad (A16) and iPad Air, and Compare iPad ／ Apple Pencil and Apple Pencil compatibility ／ CLIP STUDIO PAINT system requirements, supported-device FAQ and the ISO 216 table (University of Cambridge)

Method

Current lineup: the six models in ‘Current iPad models’ on Apple’s compare page: 13- and 11-inch iPad Pro (M5), 13- and 11-inch iPad Air (M4), iPad (A16) and iPad mini (A17 Pro). Older models are not covered.
Specs: Apple’s spec page for each model (Japan and US) and the compare page. Memory is quoted only where Apple states it (iPad Pro and iPad Air); otherwise ‘not stated’ and no figure is guessed.
Battery life is Apple’s test figure on pre-production units, not drawing time.
Apple Pencil support: Apple’s ‘Apple Pencil compatibility’ page (published 2026-03-31) and the Apple Pencil page. The page states pressure sensing is on Apple Pencil Pro but not Apple Pencil (USB-C).
App requirements: Procreate 5.4.14 (current per its Help Center) needs iPadOS 16.3 or later; CLIP STUDIO PAINT for iPad per its official requirements (iPadOS 18 / 26, 3GB required, 4GB recommended, 6GB for high-resolution work) and its supported-device FAQ.
Screen dimensions: Apple’s diagonal (measured as a rectangle, in inches) × 25.4, split into long and short sides by the resolution’s aspect ratio. Long = diagonal mm × horizontal px ÷ √(h px² + v px²); short = diagonal mm × vertical px ÷ √(h px² + v px²). Example: the iPad mini’s 8.3 × 25.4 = 210.8 mm split by 2266:1488 gives about 176 × 116 mm. The long side matches resolution ÷ ppi within 1.5 mm. As Apple notes, the real display area has rounded corners and is smaller.
Whole page at actual size: the paper’s long and short sides (ISO 216: A6 105 × 148 mm, A5 148 × 210 mm, A4 210 × 297 mm) compared with the screen’s. If the margin (screen − paper) is under 10% of that paper side, it’s borderline.
Use conditions: reading/notes without Apple Pencil has none. Pressure-sensitive handwriting needs Apple Pencil Pro support (Apple’s Apple Pencil comparison covers Pro and USB-C, and only Pro has pressure sensing). Pressure on the Apple Pencil (1st generation) isn’t in that table, so it isn’t judged.
CLIP STUDIO PAINT: being on its supported-device page and able to use a pen from its pressure-pen list (Apple Pencil 1st gen, 2nd gen, Pro) are checked as required; a 10.5 in or larger screen is checked separately as a recommendation (borderline if Apple’s diagonal exceeds it by under 10%). Memory (3 GB required) is judged only where Apple states it (iPad Air, iPad Pro); for the iPad mini and iPad (A16) it can’t be judged. The iPad Air (M4), absent from the supported-device page, is treated as not listed, not as unsupported.
Calculator order: it looks for the first model that fits in the order iPad mini → iPad (A16) → 11-inch iPad Air → 13-inch iPad Air → 11-inch iPad Pro → 13-inch iPad Pro (smaller lines first). This is not a price comparison. Carry-weight limits (350 g, 500 g) use Wi-Fi weights; the Cellular models fall in the same bands.
Products: new Wi-Fi models with standard glass on Amazon.co.jp and Amazon.com. A candidate needs 100+ ratings, 4.0+ stars, Amazon as seller, a configuration matching Apple’s spec, and (Japan) no strong sakura warning.

Official values used

Item	Value	Source	Checked
13-inch iPad Pro (M5): screen diagonal (measured as a rectangle)	13 inch	apple.com	2026-10-08
13-inch iPad Pro (M5): resolution (long side)	2752 px	apple.com	2026-10-08
13-inch iPad Pro (M5): resolution (short side)	2064 px	apple.com	2026-10-08
13-inch iPad Pro (M5): pixel density	264 ppi	apple.com	2026-10-08
13-inch iPad Pro (M5): weight (Wi-Fi model)	579 g	apple.com	2026-10-08
13-inch iPad Pro (M5): weight (Wi-Fi + Cellular model)	582 g	apple.com	2026-10-08
13-inch iPad Pro (M5): body height	281.6 mm	apple.com	2026-10-08
13-inch iPad Pro (M5): body width	215.5 mm	apple.com	2026-10-08
13-inch iPad Pro (M5): body thickness	5.1 mm	apple.com	2026-10-08
13-inch iPad Pro (M5): chip	M5 text	apple.com	2026-10-08
13-inch iPad Pro (M5): storage options	256GB, 512GB, 1TB, 2TB	apple.com	2026-10-08
13-inch iPad Pro (M5): memory	12GB on 256GB/512GB models, 16GB on 1TB/2TB models (unified memory)	apple.com	2026-10-08
13-inch iPad Pro (M5): port	Thunderbolt / USB 4 port (USB-C connector). USB 4 (up to 40Gb/s)	apple.com	2026-10-08
13-inch iPad Pro (M5): battery capacity	38.99 Wh	apple.com	2026-10-08
13-inch iPad Pro (M5): battery life (Apple test figure, web browsing and video over Wi-Fi)	up to 10 hours	apple.com	2026-10-08
13-inch iPad Pro (M5): refresh rate	ProMotion technology with an adaptive refresh rate of 10Hz to 120Hz	apple.com	2026-10-08
13-inch iPad Pro (M5): display type, lamination and color gamut	Ultra Retina XDR (tandem OLED). Fully laminated. Wide color (P3)	apple.com	2026-10-08
13-inch iPad Pro (M5): Nano-texture glass	Nano-texture display glass is an option on 1TB and 2TB models	apple.com	2026-10-08
13-inch iPad Pro (M5): supported Apple Pencil	Apple Pencil Pro, Apple Pencil (USB-C)	support.apple.com	2026-10-08
13-inch iPad Pro (M5): Apple Pencil hover	Supported	apple.com	2026-10-08
11-inch iPad Pro (M5): screen diagonal (measured as a rectangle)	11.1 inch	apple.com	2026-10-08
11-inch iPad Pro (M5): resolution (long side)	2420 px	apple.com	2026-10-08
11-inch iPad Pro (M5): resolution (short side)	1668 px	apple.com	2026-10-08
11-inch iPad Pro (M5): pixel density	264 ppi	apple.com	2026-10-08
11-inch iPad Pro (M5): weight (Wi-Fi model)	444 g	apple.com	2026-10-08
11-inch iPad Pro (M5): weight (Wi-Fi + Cellular model)	446 g	apple.com	2026-10-08
11-inch iPad Pro (M5): body height	249.7 mm	apple.com	2026-10-08
11-inch iPad Pro (M5): body width	177.5 mm	apple.com	2026-10-08
11-inch iPad Pro (M5): body thickness	5.3 mm	apple.com	2026-10-08
11-inch iPad Pro (M5): chip	M5 text	apple.com	2026-10-08
11-inch iPad Pro (M5): storage options	256GB, 512GB, 1TB, 2TB	apple.com	2026-10-08
11-inch iPad Pro (M5): memory	12GB on 256GB/512GB models, 16GB on 1TB/2TB models (unified memory)	apple.com	2026-10-08
11-inch iPad Pro (M5): port	Thunderbolt / USB 4 port (USB-C connector). USB 4 (up to 40Gb/s)	apple.com	2026-10-08
11-inch iPad Pro (M5): battery capacity	31.29 Wh	apple.com	2026-10-08
11-inch iPad Pro (M5): battery life (Apple test figure, web browsing and video over Wi-Fi)	up to 10 hours	apple.com	2026-10-08
11-inch iPad Pro (M5): refresh rate	ProMotion technology with an adaptive refresh rate of 10Hz to 120Hz	apple.com	2026-10-08
11-inch iPad Pro (M5): display type, lamination and color gamut	Ultra Retina XDR (tandem OLED). Fully laminated. Wide color (P3)	apple.com	2026-10-08
11-inch iPad Pro (M5): Nano-texture glass	Nano-texture display glass is an option on 1TB and 2TB models	apple.com	2026-10-08
11-inch iPad Pro (M5): supported Apple Pencil	Apple Pencil Pro, Apple Pencil (USB-C)	support.apple.com	2026-10-08
11-inch iPad Pro (M5): Apple Pencil hover	Supported	apple.com	2026-10-08
13-inch iPad Air (M4): screen diagonal (measured as a rectangle)	12.9 inch	apple.com	2026-10-08
13-inch iPad Air (M4): resolution (long side)	2732 px	apple.com	2026-10-08
13-inch iPad Air (M4): resolution (short side)	2048 px	apple.com	2026-10-08
13-inch iPad Air (M4): pixel density	264 ppi	apple.com	2026-10-08
13-inch iPad Air (M4): weight (Wi-Fi model)	616 g	apple.com	2026-10-08
13-inch iPad Air (M4): weight (Wi-Fi + Cellular model)	617 g	apple.com	2026-10-08
13-inch iPad Air (M4): body height	280.6 mm	apple.com	2026-10-08
13-inch iPad Air (M4): body width	214.9 mm	apple.com	2026-10-08
13-inch iPad Air (M4): body thickness	6.1 mm	apple.com	2026-10-08
13-inch iPad Air (M4): chip	M4 text	apple.com	2026-10-08
13-inch iPad Air (M4): storage options	128GB, 256GB, 512GB, 1TB	apple.com	2026-10-08
13-inch iPad Air (M4): memory	12GB (unified memory)	apple.com	2026-10-08
13-inch iPad Air (M4): port	USB-C port. USB 3 (up to 10Gb/s)	apple.com	2026-10-08
13-inch iPad Air (M4): battery capacity	36.59 Wh	apple.com	2026-10-08
13-inch iPad Air (M4): battery life (Apple test figure, web browsing and video over Wi-Fi)	up to 10 hours	apple.com	2026-10-08
13-inch iPad Air (M4): refresh rate (not stated by Apple)	No ProMotion or refresh-rate figure stated by Apple	apple.com	2026-10-08
13-inch iPad Air (M4): display type, lamination and color gamut	Liquid Retina (IPS). Fully laminated. Wide color (P3)	apple.com	2026-10-08
13-inch iPad Air (M4): supported Apple Pencil	Apple Pencil Pro, Apple Pencil (USB-C)	support.apple.com	2026-10-08
13-inch iPad Air (M4): Apple Pencil hover	Supported	apple.com	2026-10-08
11-inch iPad Air (M4): screen diagonal (measured as a rectangle)	10.86 inch	apple.com	2026-10-08
11-inch iPad Air (M4): resolution (long side)	2360 px	apple.com	2026-10-08
11-inch iPad Air (M4): resolution (short side)	1640 px	apple.com	2026-10-08
11-inch iPad Air (M4): pixel density	264 ppi	apple.com	2026-10-08
11-inch iPad Air (M4): weight (Wi-Fi model)	464 g	apple.com	2026-10-08
11-inch iPad Air (M4): weight (Wi-Fi + Cellular model)	465 g	apple.com	2026-10-08
11-inch iPad Air (M4): body height	247.6 mm	apple.com	2026-10-08
11-inch iPad Air (M4): body width	178.5 mm	apple.com	2026-10-08
11-inch iPad Air (M4): body thickness	6.1 mm	apple.com	2026-10-08
11-inch iPad Air (M4): chip	M4 text	apple.com	2026-10-08
11-inch iPad Air (M4): storage options	128GB, 256GB, 512GB, 1TB	apple.com	2026-10-08
11-inch iPad Air (M4): memory	12GB (unified memory)	apple.com	2026-10-08
11-inch iPad Air (M4): port	USB-C port. USB 3 (up to 10Gb/s)	apple.com	2026-10-08
11-inch iPad Air (M4): battery capacity	28.93 Wh	apple.com	2026-10-08
11-inch iPad Air (M4): battery life (Apple test figure, web browsing and video over Wi-Fi)	up to 10 hours	apple.com	2026-10-08
11-inch iPad Air (M4): refresh rate (not stated by Apple)	No ProMotion or refresh-rate figure stated by Apple	apple.com	2026-10-08
11-inch iPad Air (M4): display type, lamination and color gamut	Liquid Retina (IPS). Fully laminated. Wide color (P3)	apple.com	2026-10-08
11-inch iPad Air (M4): supported Apple Pencil	Apple Pencil Pro, Apple Pencil (USB-C)	support.apple.com	2026-10-08
11-inch iPad Air (M4): Apple Pencil hover	Supported	apple.com	2026-10-08
iPad (A16): screen diagonal (measured as a rectangle)	10.86 inch	apple.com	2026-10-08
iPad (A16): resolution (long side)	2360 px	apple.com	2026-10-08
iPad (A16): resolution (short side)	1640 px	apple.com	2026-10-08
iPad (A16): pixel density	264 ppi	apple.com	2026-10-08
iPad (A16): weight (Wi-Fi model)	477 g	apple.com	2026-10-08
iPad (A16): weight (Wi-Fi + Cellular model)	481 g	apple.com	2026-10-08
iPad (A16): body height	248.6 mm	apple.com	2026-10-08
iPad (A16): body width	179.5 mm	apple.com	2026-10-08
iPad (A16): body thickness	7 mm	apple.com	2026-10-08
iPad (A16): chip	A16 text	apple.com	2026-10-08
iPad (A16): storage options	128GB, 256GB, 512GB	apple.com	2026-10-08
iPad (A16): memory (not stated by Apple)	not stated by Apple	apple.com	2026-10-08
iPad (A16): port	USB-C port. USB 2.0 (up to 480Mb/s)	apple.com	2026-10-08
iPad (A16): battery capacity	28.93 Wh	apple.com	2026-10-08
iPad (A16): battery life (Apple test figure, web browsing and video over Wi-Fi)	up to 10 hours	apple.com	2026-10-08
iPad (A16): refresh rate (not stated by Apple)	No ProMotion or refresh-rate figure stated by Apple	apple.com	2026-10-08
iPad (A16): display type, lamination and color gamut	Liquid Retina (IPS). No full lamination stated. sRGB color (no P3 stated)	apple.com	2026-10-08
iPad (A16): supported Apple Pencil	Apple Pencil (USB-C), Apple Pencil (1st generation) (needs the USB-C to Apple Pencil Adapter)	support.apple.com	2026-10-08
iPad (A16): Apple Pencil hover	Not listed as supported	apple.com	2026-10-08
iPad mini (A17 Pro): screen diagonal (measured as a rectangle)	8.3 inch	apple.com	2026-10-08
iPad mini (A17 Pro): resolution (long side)	2266 px	apple.com	2026-10-08
iPad mini (A17 Pro): resolution (short side)	1488 px	apple.com	2026-10-08
iPad mini (A17 Pro): pixel density	326 ppi	apple.com	2026-10-08
iPad mini (A17 Pro): weight (Wi-Fi model)	293 g	apple.com	2026-10-08
iPad mini (A17 Pro): weight (Wi-Fi + Cellular model)	297 g	apple.com	2026-10-08
iPad mini (A17 Pro): body height	195.4 mm	apple.com	2026-10-08
iPad mini (A17 Pro): body width	134.8 mm	apple.com	2026-10-08
iPad mini (A17 Pro): body thickness	6.3 mm	apple.com	2026-10-08
iPad mini (A17 Pro): chip	A17 Pro text	apple.com	2026-10-08
iPad mini (A17 Pro): storage options	128GB, 256GB, 512GB	apple.com	2026-10-08
iPad mini (A17 Pro): memory (not stated by Apple)	not stated by Apple	apple.com	2026-10-08
iPad mini (A17 Pro): port	USB-C port. USB 3 (up to 10Gb/s)	apple.com	2026-10-08
iPad mini (A17 Pro): battery capacity	19.3 Wh	apple.com	2026-10-08
iPad mini (A17 Pro): battery life (Apple test figure, web browsing and video over Wi-Fi)	up to 10 hours	apple.com	2026-10-08
iPad mini (A17 Pro): refresh rate (not stated by Apple)	No ProMotion or refresh-rate figure stated by Apple	apple.com	2026-10-08
iPad mini (A17 Pro): display type, lamination and color gamut	Liquid Retina (IPS). Fully laminated. Wide color (P3)	apple.com	2026-10-08
iPad mini (A17 Pro): supported Apple Pencil	Apple Pencil Pro, Apple Pencil (USB-C)	support.apple.com	2026-10-08
iPad mini (A17 Pro): Apple Pencil hover	Supported	apple.com	2026-10-08
Apple Pencil Pro: pressure sensing	Supported	apple.com	2026-10-08
Apple Pencil (USB-C): pressure sensing	Not supported	apple.com	2026-10-08
Apple Pencil Pro: tilt sensing	Supported	apple.com	2026-10-08
Apple Pencil (USB-C): tilt sensing	Supported	apple.com	2026-10-08
Features only Apple Pencil Pro has	Squeeze, barrel roll, haptic feedback, double tap and Find My (not on Apple Pencil (USB-C))	apple.com	2026-10-08
iPad models with Apple Pencil hover	Supported on iPad Pro (M4, M5), iPad Air (M2, M3, M4) and iPad mini (A17 Pro), among others; iPad (A16) is not listed	apple.com	2026-10-08
iPadOS needed for Apple Pencil (USB-C)	iPadOS 17.1.1 or later	support.apple.com	2026-10-08
iPad (A16) and Apple Pencil (1st generation)	iPad (A16) also works with Apple Pencil (1st generation) when paired and charged through the USB-C to Apple Pencil Adapter	support.apple.com	2026-10-08
Apple Pencil (2nd generation) support	None of the six current iPads supports Apple Pencil (2nd generation)	support.apple.com	2026-10-08
iPad models Apple currently sells	13-inch iPad Pro (M5), 11-inch iPad Pro (M5), 13-inch iPad Air (M4), 11-inch iPad Air (M4), iPad (A16), iPad mini (A17 Pro)	apple.com	2026-10-08
Latest iPadOS version on Apple’s iPadOS page	iPadOS 27 text	apple.com	2026-10-08
Procreate for iPad: current version	5.4.14 text	help.procreate.com Procreate FAQ: The current version of Procreate for iPad is 5.4.14	2026-10-08
iPadOS required by Procreate 5.4.14 (or later)	16.3 text	help.procreate.com Procreate FAQ: Procreate 5.4.14 requires an iPad running iPadOS 16.3 or newer	2026-10-08
Procreate: current iPads that qualify	Procreate lists no model names, only iPadOS 16.3 or later. All six current iPads run iPadOS 27, so all meet it	help.procreate.com	2026-10-08
Procreate: maximum canvas by device	134 megapixels (11,585px x 11,585px) on iPad Pro (M4, M5) and iPad Air (M2, M3); 67 megapixels (8,192px x 8,192px) on iPad mini (6th and 7th generation). iPad (A16) and iPad Air (M4) are not listed	help.procreate.com	2026-10-08
CLIP STUDIO PAINT for iPad: App Store version	5.1.4 text	apps.apple.com	2026-10-08
CLIP STUDIO PAINT for iPad: OS in the official requirements	iPadOS 18 / iPadOS 26 (iPadOS 27 is not mentioned)	clipstudio.net	2026-10-08
CLIP STUDIO PAINT for iPad: memory requirement	3GB or more required, 4GB or more recommended; 6GB or more recommended for high-resolution illustration and manga	clipstudio.net	2026-10-08
CLIP STUDIO PAINT for iPad: recommended display size	A display of 10.5 inches or larger is recommended	clipstudio.net	2026-10-08
CLIP STUDIO PAINT for iPad: free storage	6GB or more of free storage	clipstudio.net	2026-10-08
CLIP STUDIO PAINT for iPad: pressure-sensitive pens in the official requirements	Apple Pencil (1st generation), Apple Pencil (2nd generation), Apple Pencil Pro (Apple Pencil (USB-C) is not listed)	clipstudio.net	2026-10-08
CLIP STUDIO PAINT supported devices: high-resolution list	High-resolution illustration and manga list: 13- and 11-inch iPad Pro (M5), (M4), 13- and 11-inch iPad Air (M3, M2) and others. iPad Air (M4) is not on the page	support.clip-studio.com	2026-10-08
CLIP STUDIO PAINT supported devices: standard-size list	Standard-size illustration and web manga list: iPad (A16), iPad mini (A17 Pro) and others	support.clip-studio.com	2026-10-08
iPad Air (M4) and CLIP STUDIO PAINT	iPad Air (M4) is not on CLIP STUDIO PAINT’s supported-device page (FAQ updated 2025-12-02)	support.clip-studio.com	2026-10-08
A6 paper short side	105 mm	cl.cam.ac.uk	2026-10-08
A6 paper long side	148 mm	cl.cam.ac.uk	2026-10-08
A5 paper short side	148 mm	cl.cam.ac.uk	2026-10-08
A5 paper long side	210 mm	cl.cam.ac.uk	2026-10-08
A4 paper short side	210 mm	cl.cam.ac.uk	2026-10-08
A4 paper long side	297 mm	cl.cam.ac.uk	2026-10-08
Models in the feature comparison on Apple’s Apple Pencil page	The feature comparison covers only Apple Pencil Pro and Apple Pencil (USB-C); pressure for the 1st and 2nd generation is not in that table	apple.com	2026-10-08
iPad mini (A17 Pro): video output	native DisplayPort output over USB-C, and VGA, HDMI and DVI output with adapters (sold separately)	apple.com	2026-10-08

Is the iPad mini enough for your use?: quick reference

The calculator in this article shows the result and a matching product for your own settings. Settings can be set in the URL; each link below opens the calculator with those settings.

Settings	Result	Calculator
Reading / notes (no pressure needed), Doesn’t matter, Not set	✓ The iPad mini fits. Condition: no pressure sensing needed, so only screen and weight are checked. The iPad mini meets these conditions. First model that fits: iPad mini (A17 Pro) (293 g, 8.3-inch diagonal). Storage options for the iPad mini (A17 Pro): 128 GB, 256 GB, 512 GB. Apple gives no guide to how much each use needs, so it isn’t calculated	Open with these settings
Reading / notes (no pressure needed), 350 g or less, Not set	✓ The iPad mini fits. Condition: no pressure sensing needed, so only screen and weight are checked. The iPad mini meets these conditions. First model that fits: iPad mini (A17 Pro) (293 g, 8.3-inch diagonal). Storage options for the iPad mini (A17 Pro): 128 GB, 256 GB, 512 GB. Apple gives no guide to how much each use needs, so it isn’t calculated	Open with these settings
Reading / notes (no pressure needed), Doesn’t matter, A6	✓ The iPad mini fits. Condition: no pressure sensing needed, so only screen and weight are checked. Paper: a whole A6 page (105 × 148 mm) at actual size. The iPad mini meets these conditions. First model that fits: iPad mini (A17 Pro) (293 g, 8.3-inch diagonal). About 11 mm to spare for A6 (calculated from the diagonal measured as a rectangle; the real display area has rounded corners and is smaller). Storage options for the iPad mini (A17 Pro): 128 GB, 256 GB, 512 GB. Apple gives no guide to how much each use needs, so it isn’t calculated	Open with these settings
Reading / notes (no pressure needed), Doesn’t matter, A5	△ The iPad (A16) just fits. Condition: no pressure sensing needed, so only screen and weight are checked. Paper: a whole A5 page (148 × 210 mm) at actual size. A whole A5 page doesn’t fit at actual size on the iPad mini’s screen (about 176 × 116 mm). First model that fits: iPad (A16) (477 g, 10.86-inch diagonal). About 9 mm to spare for A5 (under 10% of the paper side), so it’s borderline. The real display area has rounded corners and is smaller than calculated. Lighter model that also fits: 11-inch iPad Pro (M5) (444 g). Storage options for the iPad (A16): 128 GB, 256 GB, 512 GB. Apple gives no guide to how much each use needs, so it isn’t calculated	Open with these settings
Reading / notes (no pressure needed), Doesn’t matter, A4	✕ No model fits. Condition: no pressure sensing needed, so only screen and weight are checked. Paper: a whole A4 page (210 × 297 mm) at actual size. A whole A4 page doesn’t fit at actual size on the iPad mini’s screen (about 176 × 116 mm). No current iPad shows a whole A4 page at actual size (even the 13-inch screens have a short side of about 198 mm)	Open with these settings
Handwriting with pressure, Doesn’t matter, Not set	✓ The iPad mini fits. Condition: Apple Pencil Pro support (in Apple’s comparison table of the Pro and the USB-C version, pressure sensing is on the Pro; the USB-C version doesn’t have it). The iPad mini meets these conditions. First model that fits: iPad mini (A17 Pro) (293 g, 8.3-inch diagonal). It also supports Apple Pencil hover. Storage options for the iPad mini (A17 Pro): 128 GB, 256 GB, 512 GB. Apple gives no guide to how much each use needs, so it isn’t calculated	Open with these settings
Handwriting with pressure, 350 g or less, Not set	✓ The iPad mini fits. Condition: Apple Pencil Pro support (in Apple’s comparison table of the Pro and the USB-C version, pressure sensing is on the Pro; the USB-C version doesn’t have it). The iPad mini meets these conditions. First model that fits: iPad mini (A17 Pro) (293 g, 8.3-inch diagonal). It also supports Apple Pencil hover. Storage options for the iPad mini (A17 Pro): 128 GB, 256 GB, 512 GB. Apple gives no guide to how much each use needs, so it isn’t calculated	Open with these settings
Handwriting with pressure, Doesn’t matter, A5	△ The 11-inch iPad Air (M4) just fits. Condition: Apple Pencil Pro support (in Apple’s comparison table of the Pro and the USB-C version, pressure sensing is on the Pro; the USB-C version doesn’t have it). Paper: a whole A5 page (148 × 210 mm) at actual size. A whole A5 page doesn’t fit at actual size on the iPad mini’s screen (about 176 × 116 mm). First model that fits: 11-inch iPad Air (M4) (464 g, 10.86-inch diagonal). About 9 mm to spare for A5 (under 10% of the paper side), so it’s borderline. The real display area has rounded corners and is smaller than calculated. Lighter model that also fits: 11-inch iPad Pro (M5) (444 g). It also supports Apple Pencil hover. Storage options for the 11-inch iPad Air (M4): 128 GB, 256 GB, 512 GB, 1 TB. Apple gives no guide to how much each use needs, so it isn’t calculated	Open with these settings
Handwriting with pressure, 350 g or less, A5	✕ No model fits. Condition: Apple Pencil Pro support (in Apple’s comparison table of the Pro and the USB-C version, pressure sensing is on the Pro; the USB-C version doesn’t have it). Paper: a whole A5 page (148 × 210 mm) at actual size. A whole A5 page doesn’t fit at actual size on the iPad mini’s screen (about 176 × 116 mm). No model in this weight range fits. Set carry weight to “Doesn’t matter” to see one	Open with these settings
Handwriting with pressure, 500 g or less, A5	△ The 11-inch iPad Air (M4) just fits. Condition: Apple Pencil Pro support (in Apple’s comparison table of the Pro and the USB-C version, pressure sensing is on the Pro; the USB-C version doesn’t have it). Paper: a whole A5 page (148 × 210 mm) at actual size. A whole A5 page doesn’t fit at actual size on the iPad mini’s screen (about 176 × 116 mm). First model that fits: 11-inch iPad Air (M4) (464 g, 10.86-inch diagonal). About 9 mm to spare for A5 (under 10% of the paper side), so it’s borderline. The real display area has rounded corners and is smaller than calculated. Lighter model that also fits: 11-inch iPad Pro (M5) (444 g). It also supports Apple Pencil hover. Storage options for the 11-inch iPad Air (M4): 128 GB, 256 GB, 512 GB, 1 TB. Apple gives no guide to how much each use needs, so it isn’t calculated	Open with these settings
Drawing in CLIP STUDIO, Doesn’t matter, Not set	△ The iPad (A16) just fits. Condition: listed on CLIP STUDIO PAINT’s supported-device page and able to use a pen from its pressure-pen list (Apple Pencil 1st gen, 2nd gen, Pro). A 10.5-inch or larger screen is recommended. The iPad mini is on the supported-device list (standard-size work) and Apple Pencil Pro is a listed pressure pen, but its screen is below the recommended 10.5 inches. Apple doesn’t state memory for the iPad mini or the iPad (A16), so whether they meet CLIP STUDIO PAINT’s required 3 GB can’t be judged. First model that also meets the recommended screen size: iPad (A16) (477 g, 10.86 in). Under 10% to spare against the recommended 10.5 inches, so it’s borderline. The iPad (A16) can use the Apple Pencil (1st generation) with an adapter, and that pen is on CLIP STUDIO PAINT’s pressure-pen list. First model whose memory Apple also states: 11-inch iPad Pro (M5) (12 GB, above the required 3 GB and the 6 GB recommended for high-resolution work). Storage options for the iPad (A16): 128 GB, 256 GB, 512 GB. Apple gives no guide to how much each use needs, so it isn’t calculated. The iPad Air (M4) isn’t on CLIP STUDIO PAINT’s supported-device page, so it isn’t judged. CLIP STUDIO PAINT lists iPadOS 18 / 26; the current iPadOS 27 isn’t mentioned	Open with these settings
Drawing in CLIP STUDIO, 350 g or less, Not set	△ The iPad mini is listed, but its screen is below the recommendation. Condition: listed on CLIP STUDIO PAINT’s supported-device page and able to use a pen from its pressure-pen list (Apple Pencil 1st gen, 2nd gen, Pro). A 10.5-inch or larger screen is recommended. The iPad mini is on the supported-device list (standard-size work) and Apple Pencil Pro is a listed pressure pen, but its screen is below the recommended 10.5 inches. Apple doesn’t state memory for the iPad mini or the iPad (A16), so whether they meet CLIP STUDIO PAINT’s required 3 GB can’t be judged. First model that fits: iPad mini (A17 Pro) (293 g, 8.3-inch diagonal). It also supports Apple Pencil hover. Storage options for the iPad mini (A17 Pro): 128 GB, 256 GB, 512 GB. Apple gives no guide to how much each use needs, so it isn’t calculated. The iPad Air (M4) isn’t on CLIP STUDIO PAINT’s supported-device page, so it isn’t judged. CLIP STUDIO PAINT lists iPadOS 18 / 26; the current iPadOS 27 isn’t mentioned	Open with these settings
Drawing in CLIP STUDIO, 500 g or less, Not set	△ The iPad (A16) just fits. Condition: listed on CLIP STUDIO PAINT’s supported-device page and able to use a pen from its pressure-pen list (Apple Pencil 1st gen, 2nd gen, Pro). A 10.5-inch or larger screen is recommended. The iPad mini is on the supported-device list (standard-size work) and Apple Pencil Pro is a listed pressure pen, but its screen is below the recommended 10.5 inches. Apple doesn’t state memory for the iPad mini or the iPad (A16), so whether they meet CLIP STUDIO PAINT’s required 3 GB can’t be judged. First model that also meets the recommended screen size: iPad (A16) (477 g, 10.86 in). Under 10% to spare against the recommended 10.5 inches, so it’s borderline. The iPad (A16) can use the Apple Pencil (1st generation) with an adapter, and that pen is on CLIP STUDIO PAINT’s pressure-pen list. First model whose memory Apple also states: 11-inch iPad Pro (M5) (12 GB, above the required 3 GB and the 6 GB recommended for high-resolution work). Storage options for the iPad (A16): 128 GB, 256 GB, 512 GB. Apple gives no guide to how much each use needs, so it isn’t calculated. The iPad Air (M4) isn’t on CLIP STUDIO PAINT’s supported-device page, so it isn’t judged. CLIP STUDIO PAINT lists iPadOS 18 / 26; the current iPadOS 27 isn’t mentioned	Open with these settings
Drawing in CLIP STUDIO, Doesn’t matter, A5	△ The iPad (A16) just fits. Condition: listed on CLIP STUDIO PAINT’s supported-device page and able to use a pen from its pressure-pen list (Apple Pencil 1st gen, 2nd gen, Pro). A 10.5-inch or larger screen is recommended. Paper: a whole A5 page (148 × 210 mm) at actual size. A whole A5 page doesn’t fit at actual size on the iPad mini’s screen (about 176 × 116 mm). First model that also meets the recommended screen size: iPad (A16) (477 g, 10.86 in). Under 10% to spare against the recommended 10.5 inches, so it’s borderline. Apple doesn’t state memory for the iPad mini or the iPad (A16), so whether they meet CLIP STUDIO PAINT’s required 3 GB can’t be judged. The iPad (A16) can use the Apple Pencil (1st generation) with an adapter, and that pen is on CLIP STUDIO PAINT’s pressure-pen list. First model whose memory Apple also states: 11-inch iPad Pro (M5) (12 GB, above the required 3 GB and the 6 GB recommended for high-resolution work). About 9 mm to spare for A5 (under 10% of the paper side), so it’s borderline. The real display area has rounded corners and is smaller than calculated. Storage options for the iPad (A16): 128 GB, 256 GB, 512 GB. Apple gives no guide to how much each use needs, so it isn’t calculated. The iPad Air (M4) isn’t on CLIP STUDIO PAINT’s supported-device page, so it isn’t judged. CLIP STUDIO PAINT lists iPadOS 18 / 26; the current iPadOS 27 isn’t mentioned	Open with these settings

Logitech MX MASTER 3 review — the natural sequel to the 2S, with USB Type-C and a MagSpeed wheel

Sat, 03 Apr 2021 00:00:00 GMT

2026 update: Ported and revised from the old VuePress blog. The MX MASTER 3 itself is an older model in 2026 — the current line is the MX Master 3S (higher sensor resolution and quieter clicks). The review angles in this piece (charging port, MagSpeed wheel, thumb wheel) carry over to the 3S, so it should still be useful whether you are looking at the 3 or the 3S. The screenshots in this article are from Logicool Options at the time of writing; today it has been replaced by Logi Options+.

I used the Logitech MX MASTER 3 for over a year at work and at home. This is the write-up from that time.

I covered the previous model in a separate post: the MX MASTER 2S review. This is the sibling piece — focused on what changed and what did not between the 2S and the 3. If you are deciding between the 3, the 2S, and the 3S for a new purchase, the comparison should give you enough to choose.

The verdict — if you do not already own a 2S, get the 3 (or the 3S)

Short answer: the move to USB Type-C alone is reason enough to upgrade from a 2S. Add the MagSpeed wheel and the wider thumb wheel, and two-axis scrolling in Excel, code editors, and Figma gets visibly smoother.

The three main deltas from the 2S:

Charging port: micro-USB → USB Type-C
Main wheel: mechanical → MagSpeed (electromagnetic), which switches between ratcheted and free-spin automatically based on how hard you flick it
Thumb wheel (horizontal scroll): larger, easier to roll with the pad of your thumb

What did not change is the size lineup (full-size only) and the distinctive sculpted shape. For smaller hands it is still a large mouse, and that stays true on the 3S too.

Build and feel

Short answer: lined up next to a 2S, the 3 reads as a more angular sibling. USB Type-C is on the front, and the enlarged thumb wheel is on the left flank.

A Logicool logo sits on the left click. Directly above the main wheel is the scroll-mode toggle (ratcheted / free-spin). On the thumb side is the gesture button.

Connection is over Bluetooth or the USB receiver that succeeded Unifying. The receiver is included (the first run of the MX MASTER 3 shipped with the receiver; some 3S configurations sell it separately).

The underside carries the power switch and a device-pair toggle (up to three devices). The sensor is rated 1000 dpi (the 3S raises this to 8000 dpi).

USB Type-C lives on the front. This is the single biggest change from the 2S’s micro-USB Type-B, and the day-to-day payoff — one cable type across the whole desk — is bigger than it sounds.

From behind, the body has a clear waist through the middle. You grip it with the thumb on one side of the waist and the pinky on the other — the same hold as the 2S.

The left flank carries two side buttons (back/forward) and the horizontal thumb wheel. The thumb wheel is now nearly the size of the main wheel — easier to roll than the narrow strip on the 2S. A short green indicator blinks when you turn the mouse on.

The right side is plain — no buttons.

What worked

Short answer: the “it just switches” feel of the MagSpeed wheel, the wider thumb wheel, and USB Type-C charging.

MagSpeed — ratcheted and free-spin, switched automatically

This is the headline feature of the MX MASTER 3. Roll the wheel a little and you get the familiar ratcheted clicks; flick it hard and it shifts into free-spin mode, flying to the bottom of the page in one motion.

The switch happens automatically based on rotation speed. If you want to lock one mode, the button above the wheel toggles between ratcheted and free-spin explicitly.

The place this pays off most is moving up and down long files — code, long articles, deep spreadsheets. On the 2S I worked around it by holding Shift and tapping a button, or by keeping a second mouse on the desk for free-spinning. On the 3, one wheel covers both jobs.

A thumb wheel you can actually roll

Functionally it is the same “dedicated horizontal-scroll wheel” as on the 2S, but the larger shape lets the pad of the thumb sit on it comfortably.

Horizontal pans in Excel, canvas panning in Figma, scrubbing a video timeline — already comfortable on the 2S, now resolved with even less finger travel. Rolling the main wheel and the thumb wheel at the same time gives you simultaneous two-axis scrolling, and that is hard to go back from once it sinks in.

Charging is USB Type-C

The single biggest weak point of the 2S, gone. I almost wrote “charging only happens every two or three months, so it does not really matter” and then deleted it — being able to keep a single cable type on the desk turns out to matter more than the charge interval would suggest.

Battery life

Logitech lists 70 days. In actual use, charging every two or three months has been enough. After a full charge it feels the same as a freshly charged 2S.

What didn’t

Short answer: still full-size only, still on the expensive side, and the gesture button is “somewhat better” than the 2S rather than “fixed”.

No size options

The 3, like the 2S, comes in a single full size. For smaller hands it is still a large mouse.

If a smaller body matters more than the thumb wheel, the MX Anywhere 3 is the other Logicool option. The Anywhere 3 has the MagSpeed main wheel but no thumb wheel — so the “horizontal scrolling is great” part of this review does not apply.

Price stayed high

Around 13,000 yen at the time of writing. Whether the deltas from the 2S — Type-C, MagSpeed, the larger thumb wheel — justify that depends almost entirely on how much horizontal scrolling you do in a day.

For someone in Excel, Figma, or a video editor every day, the payback comes quickly. For browsing-heavy use, a used 2S or waiting for the 3S below is often the more satisfying buy.

The gesture button is still the soft spot

In the 2S review, I wrote about the gesture button sticking in the pressed state after a hard press mid-game. The same structure carries into the 3, and the same risk exists for users who hammer it during long gaming sessions (my own unit has not reproduced it).

For gaming, a dedicated gaming mouse is still the honest recommendation.

Comparison: MX MASTER 2S vs 3 vs 3S

The three siblings, side by side.

Angle	MX MASTER 2S	MX MASTER 3	MX Master 3S
Charging	micro-USB Type-B	USB Type-C	USB Type-C
Main wheel	Mechanical (ratcheted)	MagSpeed (electromagnetic, auto switch)	MagSpeed (electromagnetic, auto switch)
Thumb wheel	Yes (narrow)	Yes (larger)	Yes (larger, same as the 3)
Sensor resolution	4000 dpi (rated)	1000 dpi (rated)	8000 dpi (rated)
Click acoustics	Standard	Standard	Quiet click
Battery (rated)	70 days	70 days	70 days
Sizes	Full-size only	Full-size only	Full-size only
Software	Logicool Options	Logicool Options / Options+	Logi Options+
Released	2017	2019	2022

For a fresh purchase, the 3S is the easiest recommendation. If the price gap is large enough on a clearance 3, the 3 still holds up. The 2S is hard to find new in 2026; if you want USB Type-C, skip it.

The software (Logicool Options / Logi Options+)

The companion app remaps every button and tunes wheel sensitivity. At the time of writing it was Logicool Options; in 2026 it has been replaced by Logi Options+.

The mouse view changes each button’s action. The thumb wheel accepts horizontal scroll, next/previous, volume, tab switching, and more.

The gesture button can carry four directional actions on top of its press. I do not use it heavily, so mine is set to Enter.

Pointer speed, thumb-wheel sensitivity, and scroll direction are all tunable in fine steps. Per-app tuning works too — for example, dialling down the scroll distance just for Excel.

FAQ

Q. Does it work with a Mac? A. Yes. Logi Options+ has a Mac build. It pairs over Bluetooth or the USB receiver, and macOS-specific gestures (Mission Control and similar) can be assigned.

Q. Is it worth upgrading from the MX MASTER 2S? A. The two big wins are USB Type-C charging and the MagSpeed wheel’s free-spin. If you do a lot of horizontal scrolling, the upgrade is worth it. If you mostly browse and the micro-USB port on the 2S does not bother you, there is no urgency.

Q. Now that the MX Master 3S exists, is there a reason to buy the 3? A. If the street-price gap is clearly in the 3’s favor, yes. The 3S’s main improvements are sensor resolution (1000 → 8000 dpi) and quieter clicks; the horizontal-scrolling and thumb-wheel feel are the same as on the 3. Pick the 3S if quiet offices or late-night sessions matter to you. Pick the 3 if price matters more.

Q. How usable are the gestures? A. Holding the thumb gesture button while flicking the mouse up, down, left, or right triggers a configurable action in each direction. It is genuinely handy for Mission Control and virtual-desktop switching, but the press feel is unusual and it took me a few days to get used to. In the end I drifted away from gestures and settled on a single-press action (Enter).

Verdict — pays off if you do horizontal scrolling every day

Short answer: the natural sequel to the 2S. USB Type-C closes the most painful gap, and the main wheel jumps a tier with MagSpeed. For a new purchase, lean 3S; for a price-led decision, the 3 is still a fine pick.

If you live in Excel, Figma, or a code editor and move around the screen on both axes, the price comes back in saved time within a few months. If you mostly browse, have smaller hands, or care most about quiet clicks — pick whichever of those applies and look at the 3S or the Anywhere 3 instead.

For the sibling review, see the MX MASTER 2S.

Vue.js basics — a working engineer's guide to the file structure and lifecycle

Fri, 02 Apr 2021 00:00:00 GMT

2026-05-16 update: This post is a few years old. It’s written against Vue 2 + the Options API — the data() { return ... } + methods: {} style. The current mainstream in 2026 is Vue 3 + the Composition API (<script setup> with ref / reactive / computed), and lifecycle hooks are functions like onMounted / onUnmounted. The skeleton — the three-block structure and the created-vs-mounted distinction — still holds, but the code itself isn’t something you’d drop into a new project as-is.

I’ve been working with Vue.js for over a year now. This post is the bare-bones basics of Vue.js, with a bit of working experience folded in.

The takeaway

A Vue file puts HTML, JavaScript, and CSS in a single file
The three-block structure — template / script / style — is enough to get started
In the script block, the first thing to learn is data, methods, and the lifecycle
For the lifecycle, mainly learn the distinction between created and mounted

Understand the structure of a Vue file

The first thing to understand the structure of a Vue file. A Vue file holds all three languages:

HTML
JavaScript
CSS

It looks like this:

<template>
  <div>
    html記述
  </div>
</template>

<script>
  export default {
    name: "XXX",
    data() {
      return {
        xxx: yyy
      }
    }
  }
</script>

<style lang="css">

</style>

Each language can be split out into its own file, but I’ll skip that here.
The script and style blocks can be swapped for other languages.

I usually go with TypeScript for script and SCSS or Stylus for style.

Write HTML in the template block

This is where you write HTML in the template block.

It’s mostly the same as plain HTML, with a few Vue-specific additions.

I’ll cover the Vue-specific bits in a separate post.

Write data and methods in the script block

This is where the JavaScript side lives. The entry point is to write data and methods in the script block.

I’ll use plain JavaScript for the examples here.

The base shape:

export default {
  data() {
    return {
      xxx: "yyy"
    }
  },

  methods: {
    hogeMethod(str) {
      return str
    },
    hugaMethod() {
      let test = this.xxx;
      test = this.hogeMethod(test);
      return test;
    }
  }
}

data is where you declare and initialise variables that are available inside the Vue file.
It’s used a lot when you want to render processed data into the HTML.

Note: you’re returning an object literal here.

methods is just a place for regular functions.

To reference a variable or function defined in the same file, prefix it with this.
Local variables inside the same function don’t need this..

The relevant part:

hugaMethod() {
  let test = this.xxx;
  test = this.hogeMethod(test);
  return test;
}

Use lifecycle hooks to control timing

The lifecycle is the trickiest part of this post. Read it with this in mind: use lifecycle hooks to control timing.
The official reference is here:

Vue.js docs: lifecycle diagram

The lifetime of a Vue file — from load to teardown — is divided into 8 points, and you can hook JavaScript into each one.

The eight points:

beforeCreate — before instance creation
created — after instance creation
beforeMount — before mount
mounted — after mount
beforeUpdate — before update
updated — after update
beforeDestroy — before teardown
destroyed — after teardown

In code:

export default {
  data() {},
  beforeCreate() {},
  created() {},
  beforeMount() {},
  mounted() {},
  beforeUpdate() {},
  updated() {},
  beforeDestroy() {},
  destroyed() {},
  methods: {}
}

The two you’ll actually use most are created and mounted.
The difference between them is when the DOM is loaded.

created runs before the DOM is loaded
mounted runs after the DOM is loaded

The rough rule: use created for script-side initialisation and mounted for anything that touches the HTML.

Write CSS in the style block

The CSS lives here — write CSS in the style block.
The syntax is plain CSS, so there’s nothing new to learn.

.display-flex {
  display: flex;
}

Closing

That covers the bare basics of Vue.js.

The main acts — Components and Binding — are in separate posts.

Related posts:

Angular vs Vue.js — what a year of using both in production actually felt like

Wed, 31 Mar 2021 00:00:00 GMT

2026 update: Originally written in 2021, when Vue 2 and Angular 11 were the mainstream versions. Today the defaults are Vue 3 (Composition API, <script setup>) and Angular 17+ (Signals, standalone components, the new control-flow syntax), and Vue’s TypeScript story has improved a lot. The overall framing — Angular as batteries-included, Vue as minimal — still mostly holds, but the code samples and APIs here are from the original version and don’t match current behavior exactly.

There are three JavaScript frameworks people usually mention in the same breath, and I’ve shipped production work in two of them. React isn’t one of them, so this post is just about Angular and Vue.js, and how they actually differ when you use them day-to-day.

The short version

Angular
- Batteries included
- Strict syntax conventions
Vue.js
- Minimal core
- Loose, flexible conventions

Angular

Overview

Angular is the framework Google maintains. The v1 line was called AngularJS; from v2 onward it’s just Angular. Google has said major versions ship roughly every six months, and that major versions don’t guarantee backwards compatibility — worth keeping in mind.

In practice, upgrades on the projects I worked on were less painful than that warning suggested.

The two things that stood out to me:

It’s a full-stack framework
TypeScript by default

Full-stack framework

Angular ships with routing and other common pieces out of the box. Install it and you basically have a working project skeleton. Looking things up is also easier because the official docs cover most of what you need.

A component’s folder layout looks like this:

component
  ├── component.ts
  ├── component.html
  ├── component.css
  └── component.spec.ts

Keeping the languages in separate files inside one folder makes the code easier to manage at scale.

The downside of “batteries included” is that you carry code you don’t necessarily use, and that has a real performance cost.

TypeScript

Angular’s recommended language is TypeScript. TypeScript is a Microsoft project, so you end up with a Google framework written in a Microsoft language — which on paper is a pretty solid combination.

Static typing makes the code feel tight, in a good way. But for anyone new to TypeScript, you’re effectively learning two things at once, and the ramp-up is steep.

When I’d reach for Angular

The conventional wisdom is that Angular fits large projects, and my experience matches that.

The strict structure and conventions — for better or worse — meant handoffs and parallel work between team members went smoothly. The trade-off is the learning cost: plan extra time at the start of any handoff for people to get comfortable with the framework.

Vue.js

Overview

Vue.js is community-developed and has been picking up adoption steadily. The two things that stood out to me:

A minimal core
Flexibility

Minimal core

Unlike Angular, Vue doesn’t bundle routing and other extras by default. You install the libraries you need separately. Vue CLI makes that pretty painless.

A component is a single file:

component
  └── component.vue

Very simple. HTML, JS, and CSS all live inside the .vue file.

You can split JS and CSS into separate files if you want:

component
  ├── component.vue
  ├── component.js
  └── component.css

The split is manual, though — there’s no automatic separation.

The flip side of a minimal core is that you have to find and pick libraries yourself. The upside is the bundle stays light, and performance is good by default.

Flexibility

Vue defaults to plain JavaScript. TypeScript works too, but setting it up properly used to take some effort (current note: Vue 3 with <script setup> has improved TypeScript support significantly, and as of 2026 the setup is no longer the same hassle it was).

The key difference from Angular is how much room Vue gives you. You can shape a project to fit the team’s skill level and the use case — that flexibility is genuinely useful.

You’ll need to do more of your own research, but the barrier to entry is clearly lower.

When I’d reach for Vue

Vue fits anything that isn’t a strict programmer-and-designer split with formal handoffs.

The catch: document your library choices, your project structure, and the build setup. The flexibility that makes Vue pleasant in solo work becomes a maintenance burden in a team if nothing is written down. For company projects, I’d strongly recommend setting up a formatter and a style guide on day one.

For personal projects, Vue.js is hard to beat.

Wrap-up

That’s the comparison, with my own bias from working in both.

If you forced me to pick one, I’d go with Vue.js plus TypeScript. The setup takes some patience, but the readability and the resulting code are worth it.

YAMAHA TW-E3A review — a year of commuting with YAMAHA's honest, even-handed true wireless

Tue, 30 Mar 2021 00:00:00 GMT

2026 update: Ported from the old VuePress blog. The TW-E3A is an older model in 2026, with successors like the TW-E3B and TW-E3C now on sale. The review angles here (the honest tuning, the fit, the physical-button quirks) still hold up. Spec numbers are based on the manufacturer’s published figures at the time of writing.

I used the YAMAHA TW-E3A true wireless earbuds for about a year of daily commuting. This is the write-up of what showed up — strengths and weaknesses — once they were folded into the rhythm of regular use.

The review is framed around the fact that the TW-E3A was YAMAHA’s first true wireless earbud. So instead of grading it head-to-head against the latest competition, I focus on whether it fits its price band and intended use.

The verdict — for listeners who want tuning without a personality

Short answer: the sound follows YAMAHA’s tuning brief without unnecessary flavor, and the fit is stable enough for a commute. The deep physical-button press, and the way the controls are split across left and right, are the parts that will pick their audience.

The reason is simple. The TW-E3A does not chase a single headline trait — no thumping bass, no aggressive noise cancellation. Instead it settles into a balance that does not tire the ear over long sessions. The design leans toward background listening during a commute or while working, rather than dedicated music-appreciation sessions.

That said, you should probably look elsewhere if any of these apply:

You’re used to touch controls or multi-button setups: the TW-E3A funnels every function through a single physical button on each side
You need active noise cancellation: the TW-E3A does not have ANC (based on the manufacturer’s spec sheet)
You want to use them for hard workouts: the fit is secure, but there are no sport fins for high-impact movement

Build and feel

Short answer: the case is a small rectangle, roughly the size of a miniature bento box, that slips into a chest pocket. The earbuds themselves are slightly elongated and sit deep in the ear.

The lid carries a matte YAMAHA wordmark. The finish does not pick up fingerprints easily.

The case measures roughly 70 × 35 × 35 mm (manufacturer’s published figures). That is compact for its price band, and it fits cleanly into a jacket’s chest pocket.

The lid pops open with a light push and falls shut under its own weight, locking with a small click. In a year of bag-carrying it never opened on its own.

The nozzle is long. Once seated, it sits about as deep as the first knuckle of an index finger, which is what gives the fit its security. The R / L marks are embossed on the inside of each shell, so you can tell them apart by feel in the dark.

On its own, each bud reads as slightly elongated. Each side weighs 6.3 g (manufacturer’s published figures).

The back of the case carries a USB Type-C port and a charge-status LED. USB-C was not a given on early-2021 models, and in 2026 it is still useful for keeping one cable type across devices.

What worked

Short answer: honest tuning, a fit that holds up at commute intensity, and a stable Bluetooth 5.0 link.

YAMAHA’s flat, low-fatigue tuning

The tuning does not push the low end and does not sharpen the upper mids. Vocals and acoustic instruments come through naturally; genres that lean on a forward low end — EDM, for example — can feel underweight to some listeners.

Two- or three-hour sessions stay comfortable, which is the practical payoff of that tuning choice.

Stays in for commutes and walking

The deep nozzle keeps the bud seated through the shoves of a packed train or a short jog. Across a year of daily use, neither side ever fell out.

There are no sport fins in the box, though, so for the gym or longer runs it is worth looking at a model built for that.

Bluetooth 5.0 holds up in crowded RF

With Bluetooth 5.0 (based on the manufacturer’s spec sheet), the link stayed clean inside busy stations and trains where RF gets congested. Dropouts and single-ear lockouts were rare. Compared to the 4.x-era buds I used at the same price point, the stability is clearly a step up.

What didn’t

Short answer: the buttons need a firm press, the controls are split across the two sides, and there is no ANC.

Buttons want a deep press

These are physical buttons rather than touch sensors. That cuts down on accidental taps, but the TW-E3A’s travel is on the deep side. To register a press you end up pushing the bud further into the ear, and each time you do, the seal shifts.

A touch sensor — or even just a lighter actuation — would have been easier to live with.

Left and right do different things, and it’s hard to memorize

Because there is only one button per side, playback, volume, and track skipping are spread across the two sides by click count.

Action	Mapping
Play / pause	Either side, single click
Volume up	Right, double click
Volume down	Left, double click
Next track	Right, long press
Previous track	Left, long press

If “volume on the right, track skip on the right” had been the rule, the mapping would have stayed intuitive by direction. As it stands you have to memorize the chart. For someone who only touches the controls a few times a day, that is more friction than it should be.

No active noise cancellation

The canal-type fit gives you some passive isolation, but there is no active NC (based on the manufacturer’s spec sheet). If your goal is to flatten out rail noise or HVAC drone, this is not the right pick.

Comparison — where the TW-E3A sits in mid-range true wireless

Short answer: if you care about even-handed tuning and a secure fit, it competes well at this price. If ANC or rich touch controls are non-negotiable, look elsewhere.

Angle	YAMAHA TW-E3A	Typical mid-range peers	YAMAHA’s higher tier (TW-E5B / E7B family)
Tuning	Flat, low-fatigue	Often bass-forward	Listening-focused, bass with detail
Active noise cancellation	None	Increasingly common	Available on some models
Controls	One physical button per side	Touch is the norm	Touch plus companion app
Connection	Bluetooth 5.0	5.0 to 5.3	5.2 and up is common
Charging port	USB Type-C	Mostly USB-C now	USB Type-C
Bud-only playback	~6 hours (rated)	5 to 8 hours	6 to 10 hours (rated)

If you are buying new and you specifically want the YAMAHA house sound, the more natural move is to compare the TW-E5B / E7B siblings as well. Conversely, on the second-hand or clearance market, where the TW-E3A turns up cheap, its solid baseline of tuning and fit still earns it a place.

FAQ

Q. Will it pair without issues on both iPhone and Android? A. Any Bluetooth 5.0 host is fine in practice. There is no multipoint support (based on the manufacturer’s spec sheet), so it is not a good fit for someone who switches between a PC and a phone all day. Treat it as a single-device companion, or expect to re-pair each time.

Q. How long does the battery last? A. The published figures are around 6 hours on a charge and roughly 24 hours including the case (manufacturer’s published figures). With a one-hour commute each way, the case lands back on the dock about once every three or four weekdays.

Q. Is it usable for remote-work video calls? A. There is a built-in mic and calls work, but the noise reduction is not what you would expect from a dedicated headset (based on the manufacturer’s spec sheet). It is fine for a 1:1 from a quiet home; for calls in a cafe or an open office, a real headset is the safer bet.

Q. Does it hold up for running or light workouts? A. The deep nozzle keeps it secure through short jogs on a commute. Heavy vertical motion during running, or sweat-heavy gym sessions, are outside what it was built for, and long-term sweat resistance is not guaranteed (check the manufacturer’s IP rating). For serious training, use something built for it.

Verdict — for daily listening with YAMAHA’s tuning

Short answer: not a model that sells on flash. It earns its keep through the quiet quality of the tuning and the fit, which is the kind of thing you only notice after weeks of use. The cleanest way to read it: a first-generation product that did not try to do too much.

The two things to weigh are the button feel and the awkward control mapping. If a store has a demo unit, putting them in and running through the left-right controls before deciding will save regret later. If you are also weighing newer generations, look at the higher-tier models and the successor lineup (ANC, multipoint, touch controls) at the same time — the decision gets less wobbly with that context.

Anker PowerCore 10000 review — the palm-sized power bank I used for years (possible recall)

Mon, 29 Mar 2021 00:00:00 GMT

2026 update: Ported from the old VuePress blog. The PowerCore 10000 is an older model in 2026 — the current line has moved to USB Type-C I/O in the newer PowerCore series. The framing here (how to weigh capacity, size, and ports in the 10000mAh class) still applies.

Recall notice (added 2026-10-11): The PowerCore 10000 in this review is likely model A1263 (the article does not state a model number, and I could not confirm it from the product page). On 2025-06-12 the US Consumer Product Safety Commission announced a recall of the A1263 for fire and burn hazards (recall no. 25-338; units with qualifying serial numbers sold in the US from June 2016 through December 2022). The model number and “SN” serial are printed on the bottom of the unit. Check whether yours is covered on Anker’s recall page. If it is, stop using it and follow the official steps (CPSC notice).

I could not confirm a recall in Japan: the Anker Japan recall page does not list the A1263.

I used the Anker PowerCore 10000 as my main mobile battery for several years, carried for the usual reason — a backup against a dead phone outside the house. The 10000mAh class strikes a workable balance between capacity (two or three charges for a 2021-era phone) and “small enough to carry”, which makes it an easy first battery to land on.

The I/O ports, though, are this generation’s weak spot. From a 2026 vantage point, I want to write down which parts still hold up and which parts I would no longer pick.

The verdict — a solid default for size-vs-capacity, but the ports are dated

Short answer: fitting 10000mAh into a pocketable shell still earns the recommendation. The charging behaviour is steady in the way Anker batteries usually are, which makes it easy to trust as an outing backup. The catch is the USB Type-A input and micro-B output — that pairing does not match the cables most people already carry in 2026.

The reason is simple: phones, laptops, and wireless earbuds have been converging on USB Type-C. For anyone trying to standardise on a single Type-C cable, packing a dedicated micro-B cable just for this one battery is hard to justify.

A few cautions before buying:

You want to standardise on USB Type-C cables: a Type-C version of the same series is the cleaner choice
You want to charge a laptop with PD (Power Delivery): this unit tops out at USB-A / 2.4A and does not support laptop fast-charge
You are buying brand new: the current Type-C successor will keep working with the rest of your gear for longer

Look and size

Short answer: compact for a 10000mAh battery. Anker traded height for thickness, so it still slips into a jeans pocket.

The shell is matte black, the same finish Anker uses across the mobile battery range, and it hides fingerprints and minor scuffs well.

Stated specs are below (source: as printed on the packaging).

Spec	Value
Size	~92 x 60 x 22mm
Weight	~180g
Capacity	10000mAh
Output	Up to 2.4A (PowerIQ)
Input port	USB Type-A
Output port	micro-B

180g is about average for the 10000mAh class. The short height means it sits well in a closed hand.

The battery charges over USB Type-A and outputs over micro-B. At the time of writing (2021), USB Type-A was the default everywhere. In 2026 the layout reads as an older generation.

The power button is on the side. A single press lights four blue LEDs that show the remaining charge. Plug in a phone, press the button, and it starts pushing power — there is nothing to think about.

What worked

Short answer: Anker’s reliable charging behaviour and a genuinely pocketable 10000mAh form factor.

The charging behaviour is steady

I have owned several Anker batteries and never run into power cutting out mid-charge or the device failing to register. For a battery that gets plugged and unplugged constantly throughout the day, that kind of quiet reliability matters more than a spec sheet.

Anker’s own PowerIQ (current adjustment based on the connected device) seems to do its job within the 2.4A range — though this is based on day-to-day use rather than measured output.

Pocketable for a 10000mAh battery

10000mAh covered about two to three charges of a 2021-era phone, or just under one full tablet charge. A 2026 phone with a 4,000mAh-class battery gets only about 1.3 to 1.8 full charges (0% to 100%, assuming 60-70% efficiency). Fitting that into a pocketable shape was the headline at launch and still is.

On a day out without a bag, it goes in a jacket or back pocket and lasts the day.

Four-LED battery indicator, no fuss

One button press, four LEDs. There is no app, no Bluetooth pairing — and as a result, no ambiguity about how much is left. It is hard to walk out the door with a battery you thought was full but wasn’t.

What didn’t

Short answer: USB Type-A / micro-B ports, no PD support, and slow input charging.

USB Type-A / micro-B I/O

Normal in 2021, a clear weakness in 2026. If your cable kit is already standardised on USB Type-C, carrying a micro-B cable just for this one battery starts to feel like overhead.

For a new purchase I would point people at the Type-C version of the same line (PowerCore III 10000 or similar).

No USB PD

The output is a single USB-A port at up to 2.4A — no USB Power Delivery (PD).

That rules out fast-charging a recent phone, or driving a laptop or tablet at any useful speed. As a topping-up battery for a phone it is still fine.

Slow to refill itself

Charging the battery itself is over micro-B, and even on the bundled wall adapter, it feels slow to reach full.

Plug it in before bed and it is ready in the morning, which is the usual pattern. As a “top up just before leaving” battery, it is not the right tool.

Comparison: choosing in the 10000mAh class

10000mAh is the capacity most people land on for a first battery. A short way to compare options inside that class:

Angle	PowerCore 10000 (this unit)	Type-C PowerCore successor	20000mAh class
Ports	USB-A / micro-B	USB-C in/out	Varies by model
Weight	~180g	~190g (rated)	~350g
Capacity	10000mAh	10000mAh	20000mAh
Best for	2–3 charges of a 2021-era phone	Phone + light tablet	Phone + laptop
Launch price (at time of writing)	~2,500–3,500 yen	~3,500–5,000 yen	~5,000–8,000 yen

The decision is short: “do I want everything on Type-C?” and “do I need to charge a laptop?” Two no’s and this class is enough. One yes and a different model fits better.

FAQ

Q. Does it work with both iPhone and Android? A. Yes. The output is USB Type-A, so as long as you carry the right cable for your device (Lightning / USB-C / micro-B), either side works. There is one USB-A output port on the battery.

Q. Can I take it on a plane? A. Carry-on, yes. 10000mAh works out to roughly 37Wh, comfortably under the 100Wh limit major airlines apply (airlines and routes vary, check the current rules for your specific flight). Checked baggage is not allowed.

Q. How many times can it charge a phone? A. Depends on the phone’s battery and charging losses, but for a 3000–4000mAh phone of the 2021 era, two to three full charges was the practical figure. For a current 4,000mAh-class phone it is about 1.3 to 1.8 full charges (0% to 100%, assuming 60-70% efficiency). The four-LED indicator gives you a rough sense of what’s left as you go.

Q. Can it charge a laptop? A. Not really. The output is USB-A at 2.4A with no USB PD, so it does not meet the fast-charge requirements of most laptops. For a laptop, look at a PD-capable battery with at least 30W output.

Verdict — still usable as a first battery, but a successor is the smarter new purchase

As a default in the 10000mAh class, it is still easy to live with. Anker’s steady charging behaviour and the palm-sized shell make it a quiet, dependable backup when leaving the house.

For a fresh purchase in 2026, though, the Type-C versions of the current line will hurt less in the long run, simply because everything else you carry is already on Type-C.

If you already own a PowerCore 10000, there is no reason to retire it — keep a micro-B cable in the bag and carry on. When it does come time to replace it, the Type-C successor is the obvious next stop.

Windows Laptop Specs for Programming: Minimum RAM, SSD and CPU

Sun, 28 Mar 2021 00:00:00 GMT

Working from the official requirements as of October 2026, a Windows laptop for programming needs at least 8 GB of RAM if you skip the Android emulator. That’s each tool’s official minimum (8 GB for Docker and Android Studio) rounded up to a standard size, separate from Windows 11’s own 4 GB. Docker and Android Studio sit exactly at 8 GB, and the emulator at 16 GB, so all three are borderline.

The official minimums are per app, and running several at once isn’t judged. Only official minimum RAM, free-space figures and Arm support are used; speed and comfort weren’t tested on real hardware and aren’t judged.

Minimum RAM and SSD by type of development

The table rounds each official minimum up to a common laptop RAM and SSD size, and marks the borderline cells.

Development	RAM	SSD	CPU
Web and WSL2 (VS Code)	8 GB	256 GB	Either x64 or Arm
Docker	8 GB (exactly the minimum, borderline)	256 GB	Either (Docker labels the Arm version Early Access)
Visual Studio (.NET and more)	8 GB	256 GB (512 GB with every workload)	Either x64 or Arm
Visual Studio (Python and data science)	8 GB	256 GB (512 GB with every workload)	x64 (Intel or AMD) required
Android (no emulator)	8 GB (exactly the minimum, borderline)	256 GB	x64 (Intel or AMD) required
Android (with emulator)	16 GB (exactly the minimum, borderline)	256 GB	x64 (Intel or AMD) required

“Borderline” means the spare RAM (the rounded-up size minus the minimum) is under 10% of the minimum.

Minimum Windows laptop spec for your development work

Pick your kind of development and CPU type. It rounds the official minimum requirements for Docker, Visual Studio, Android Studio and more up to common memory and SSD sizes, and shows the minimum spec and whether you need x64 (Intel or AMD) or can use Arm (Snapdragon).

Start at 8 GB of RAM and a 256 GB SSD

Per each vendor’s official requirements, the minimum RAM is 8 GB for Docker Desktop, 8 GB for Android Studio, and 16 GB with the emulator. Each official RAM figure is given per app, and none of the pages gives a combined figure. So running several, such as Docker and Android Studio together, isn’t judged here.

With 16 GB of memory in a programming laptop, everything except the Android emulator clears the official minimum with room to spare. With the emulator it’s exactly the minimum, so it’s borderline, and Android Studio recommends 32 GB with the emulator.

Windows 11’s 64 GB minimum plus each dev tool’s official free space fits in 256 GB, counting Visual Studio at its typical 20–50 GB install. A full Visual Studio install, at 274 GB, needs 512 GB. Setups between typical and full, and your own files and Docker images, aren’t calculated.

CPU cores and type, and tools that don’t run on Arm

Per the Visual Studio 2026 system requirements, a quad-core or better CPU is recommended. Per Android Studio’s official requirements, Intel Core N-series and U-series processors aren’t recommended due to insufficient performance. With the emulator, the minimum GPU has 4 GB of VRAM.

Per the makers’ spec pages, two chips both called Ryzen AI 5 differ: the 330 has 4 cores and the 340 has 6 cores. Check the core count on the spec page, not the name.

Android Studio needs an x64 CPU, while VS Code and WSL 2 also run on Windows on Arm. Visual Studio supports Arm64 too, but Python development, data science and other workloads aren’t supported. Before choosing Arm, check the workloads you use against Microsoft’s list.

Per the vendors’ official documentation, Docker calls its Arm build Early Access, but Microsoft says it runs natively, so the two differ.

WSL2 and Docker on Windows 11 Home, and the RAM cap

Per Microsoft’s documentation, WSL 2 works on Windows 11 Home and supports both x64 and Arm64 CPUs.

WSL 2 can use up to 50% of Windows’ memory by default. On a 16 GB laptop that’s 8 GB, and you can change it in .wslconfig.

Docker Desktop’s requirements table lists Windows 11 Enterprise, Pro and Education, version 23H2 or later; Home isn’t listed, and this article hasn’t verified whether Docker Desktop works on Home.

Used or Windows 10 laptops: pick by Windows 11 requirements

Windows 10 support ended on October 14, 2025, and consumer Extended Security Updates (ESU) run only until October 12, 2027. If you buy used, choose a machine that meets Windows 11’s requirements (a compatible 64-bit CPU with 2 cores or more, TPM 2.0, UEFI Secure Boot).

Per Microsoft’s documentation, ESU for businesses and schools runs for up to 3 years after Windows 10’s end of support.

Picks by condition

One pick per condition where a listing qualifies, from Amazon.com. When more than one product checked for this article meets these criteria, the pick is the cheapest. Check prices on the seller site.

Sold new by Amazon.com or the maker, not a gray-market import, with a model number that matches the maker’s spec
At least 100 ratings and 4 stars or more

Connecting an external monitor at 4K 60 Hz while charging the laptop

A USB-C hub with HDMI at 4K 60 Hz and up to 85 W of PD pass-through.

Sold by UGREEN GROUP LIMITED. Whether the rating count is shared with other models on the listing wasn't checked.

A 1 TB external SSD for a PC

1 TB, USB-C, and the maker lists it as compatible with PC and Mac.

A Windows laptop that meets the emulator’s RAM, SSD and x64 conditions

At least 16 GB of RAM, 256 GB of SSD and an x64 CPU. In this calculation, whether its GPU (listed by ASUS as Intel® Arc™ Graphics, with no VRAM figure) meets the emulator’s VRAM condition isn’t judged.

FAQ

How much RAM does a Windows laptop need for programming?

The minimum RAM is 8 GB for Docker Desktop, 8 GB for Android Studio, and 16 GB with the emulator. With 16 GB of memory in a programming laptop, everything except the Android emulator clears the official minimum with room to spare.

With the emulator it’s exactly the minimum, so it’s borderline, and Android Studio recommends 32 GB with the emulator. The official minimums are per app, and running several at once isn’t judged.

Can you program on a laptop with 8 GB of RAM?

Yes. With 8 GB, web development, WSL 2 and Visual Studio meet the official minimum.

Docker and Android Studio (without the emulator) are exactly at the official 8 GB minimum, so it’s borderline. With the Android emulator the minimum is 16 GB, so 8 GB isn’t enough.

VS Code has no official RAM figure, so web development and WSL 2 are judged against Windows 11’s 4 GB minimum. Rounded up to a memory tier, the laptop to buy starts at 8 GB.

Visual Studio’s minimum RAM is 4 GB, and 16 GB is recommended for typical professional solutions.

Is a 256 GB SSD enough for programming?

Setups between typical and full, and your own files and Docker images, aren’t calculated.

Can you program on a Snapdragon (Arm) Windows laptop?

Android Studio needs an x64 CPU, while VS Code and WSL 2 also run on Windows on Arm. Visual Studio supports Arm64 too, but Python development, data science and other workloads aren’t supported.

Before choosing Arm, check the workloads you use against Microsoft’s list. Per Android Studio’s official requirements, Windows machines with Arm-based CPUs (such as Snapdragon) aren’t currently supported.

Per the vendors’ official documentation as of October 2026, Docker calls its Arm build Early Access, but Microsoft says it runs natively, so the two differ.

Do WSL2 and Docker work on Windows 11 Home?

WSL 2 works on Windows 11 Home and supports both x64 and Arm64 CPUs. Docker Desktop’s requirements table lists Windows 11 Enterprise, Pro and Education, version 23H2 or later; Home isn’t listed, and this article hasn’t verified whether Docker Desktop works on Home.

What’s the difference between the Ryzen AI 5 330 and 340?

Two chips both called Ryzen AI 5 differ: the 330 has 4 cores and the 340 has 6 cores. Check the core count on the spec page, not the name.

Data and formulas used (for verification)

How this was checked:

Sources: Windows 11 specifications and Windows 10 end of support ／ Install WSL, WSL FAQ and .wslconfig settings ／ Docker Desktop for Windows and Windows on Arm overview ／ Visual Studio 2026 system requirements and Visual Studio on Arm devices ／ Android Studio install requirements and .NET MAUI Android emulator ／ Dell 14 specs and ASUS Vivobook 16 specs

Method

Memory needed = the official minimum RAM for the chosen kind of development (Docker Desktop for Windows 8 GB, Visual Studio 2026 4 GB, Android Studio 8 GB without the emulator and 16 GB with it). VS Code (web development, WSL 2) has no official RAM figure, so it’s judged against the Windows 11 minimum of 4 GB. Official recommendations (Visual Studio 16 GB, Android Studio 32 GB) aren’t used for the conclusions. Each official figure is given per app and no page gives a combined figure, so running several at once isn’t judged.
Minimum memory = the smallest of 8, 16, 32 and 64 GB (tiers chosen for this article as common laptop sizes) at or above the memory needed. If the headroom (tier − memory needed) is under 10% of the memory needed, it’s marked borderline (a threshold set for this article).
SSD needed = Windows 11’s minimum storage 64 GB plus the tool’s official free space (Visual Studio: the top of the typical 20-50 GB, or 210 GB with every feature; Android Studio 8 GB without the emulator and 16 GB with it), rounded up to 256 GB, 512 GB, 1 TB or 2 TB. Docker Desktop and VS Code publish no free-space figure. Your own files, Docker images and what’s inside WSL aren’t calculated. WSL 2’s default virtual disk size of 1 TB is a cap, not usage.
x64 or Arm = whether the tool officially runs on Windows on Arm (Snapdragon). Google states Android Studio doesn’t support Windows on Arm (Microsoft Learn’s .NET MAUI page also says the Android emulator needs x64). Visual Studio 2026 supports Windows 11 on Arm64, but some workloads such as Python development and data science aren’t supported there. Docker labels the Arm version of Docker Desktop Early Access, while Microsoft’s Windows on Arm overview says it runs natively (Docker’s label is used). WSL supports x64 and Arm64, and VS Code runs natively on Arm. Windows 10 consumer ESU runs until 2027-10-12 (business and education: up to 3 years after end of support).
By default WSL 2 can use 50 % of Windows’ memory (changeable in .wslconfig). Speed and comfort weren’t tested on real hardware and aren’t judged.jp or Amazon.com as of October 8, 2026.
Products (USB-C hubs, an external SSD, and a US laptop) must have 100+ ratings, 4+ stars, a seller that is Amazon or the maker (checked on the seller page’s business name and description), and be new with a model number matching the maker’s spec page (checked October 11, 2026). The hubs’ 4K HDMI at 60 Hz and PD pass-through (Anker 85 W, UGREEN 85 W) and the WD My Passport SSD’s “compatible with PC and Mac” (Windows isn’t named) come from the makers’ spec pages. The Zenbook Duo’s GPU is listed as “Intel® Arc™ Graphics” with no VRAM figure, so whether it meets the emulator GPU condition (4 GB of VRAM) isn’t judged.

Official values used

Item	Value	Source	Checked
Windows 11 minimum RAM	4 GB	microsoft.com	2026-10-08
Windows 11 minimum storage	64 GB	microsoft.com Storage: 64 GB or larger storage device	2026-10-08
Windows 11 minimum CPU cores	2 cores	microsoft.com Processor: 1 GHz or faster with 2 or more cores on a compatible 64-bit processor	2026-10-08
Windows 11 TPM requirement	2.0	microsoft.com TPM: version 2.0。System firmware: UEFI, Secure Boot capable	2026-10-08
Copilot+ PC RAM requirement	16 GB	microsoft.com	2026-10-08
Windows 10 end of support date	2025-10-14	microsoft.com Windows 10 reached end of support on October 14, 2025	2026-10-08
Windows 10 consumer ESU end date	2027-10-12	microsoft.com	2026-10-08
WSL supported Windows versions	(stated in a Japanese-language source)	learn.microsoft.com Prerequisites: Windows 10 version 2004 and higher (Build 19041 and higher) or Wi	2026-10-08
WSL 2 works on Windows 11 Home	true	learn.microsoft.com	2026-10-08
WSL supports Arm64 CPUs	true	learn.microsoft.com What processors does WSL support? WSL supports x64 and Arm64 CPUs.	2026-10-08
WSL 2 default memory limit (share of Windows memory)	50 %	learn.microsoft.com	2026-10-08
WSL 2 default swap (share of memory, rounded up to GB)	25 %	learn.microsoft.com	2026-10-08
WSL 2 default virtual disk size limit	1 TB	learn.microsoft.com	2026-10-08
WSL 2 default processors	(stated in a Japanese-language source)	learn.microsoft.com	2026-10-08
Docker Desktop for Windows minimum RAM	8 GB	docs.docker.com	2026-10-08
WSL version required by Docker Desktop	2.1.5	docs.docker.com	2026-10-08
Docker Desktop supported Windows 11	(stated in a Japanese-language source)	docs.docker.com	2026-10-08
Docker Desktop for Windows Arm status (per Docker)	Early Access	docs.docker.com	2026-10-08
Current Visual Studio release (Stable channel)	Visual Studio 2026	learn.microsoft.com	2026-10-08
Visual Studio 2026 minimum RAM	4 GB	learn.microsoft.com Hardware: Minimum of 4 GB of RAM	2026-10-08
Visual Studio 2026 recommended RAM (typical professional solutions)	16 GB	learn.microsoft.com We recommend 16 GB RAM for typical professional solutions; works best with 64 GB	2026-10-08
Visual Studio 2026 typical free disk space	20-50 GB	learn.microsoft.com	2026-10-08
Visual Studio 2026 maximum disk space (all features)	210 GB	learn.microsoft.com Minimum of 2.5 GB up to 210 GB of available space, depending on features install	2026-10-08
Visual Studio 2026 CPU recommendation	(stated in a Japanese-language source)	learn.microsoft.com ARM64 or AMD64/x64 processor; Quad-core or better recommended; works best with 1	2026-10-08
Visual Studio 2026 minimum display resolution	(stated in a Japanese-language source)	learn.microsoft.com	2026-10-08
Visual Studio 2026 on Arm64	(stated in a Japanese-language source)	learn.microsoft.com	2026-10-08
Workloads not supported by Visual Studio 2026 on ARM64	(stated in a Japanese-language source)	learn.microsoft.com The following workloads and components are not supported by Visual Studio ARM64	2026-10-08
Android Studio minimum RAM (without emulator)	8 GB	developer.android.com	2026-10-08
Android Studio + Emulator minimum RAM	16 GB	developer.android.com	2026-10-08
Android Studio + Emulator recommended RAM	32 GB	developer.android.com	2026-10-08
Android Studio minimum free disk (without emulator)	8 GB	developer.android.com Disk space: Studio: 8 GB of free space	2026-10-08
Android Studio + Emulator minimum free disk	16 GB	developer.android.com	2026-10-08
Extra storage per additional AVD (up to)	6 GB	developer.android.com You might need up to 6GB of additional storage for each additional AVD created	2026-10-08
Approx. memory per simultaneously running AVD	4 GB	developer.android.com Running multiple AVDs simultaneously requires approximately 4GB of memory per AV	2026-10-08
Minimum GPU VRAM with Emulator	4 GB	developer.android.com	2026-10-08
Android Studio does not support Windows on ARM	Windows machines with ARM-based CPUs aren’t currently supported.	developer.android.com	2026-10-08
Android Studio CPU note	(stated in a Japanese-language source)	developer.android.com Please be aware that Intel Core N-Series and U-Series processors are not recomme	2026-10-08
Windows 11 on Arm x64 emulation	(stated in a Japanese-language source)	learn.microsoft.com Windows 11 adds the ability to run unmodified x64 Windows apps on Arm devices.	2026-10-08
Visual Studio Code natively supports Arm	true	learn.microsoft.com Visual Studio Code natively supports Arm and can be installed on Arm devices.	2026-10-08
Ryzen AI 5 330 core count (per Dell spec page)	4 cores	dell.com	2026-10-08
Ryzen AI 5 340 core count (per ASUS spec page)	6 cores	asus.com Processor: AMD Ryzen AI 5 340 Processor 2.0GHz (22MB Cache, up to 4.8GHz, 6 core	2026-10-08
Windows 10 business/education ESU maximum length (after end of support)	3 years	learn.microsoft.com	2026-10-08
Windows 11 processor requirement	(stated in a Japanese-language source)	learn.microsoft.com Processor: 1 gigahertz (GHz) or faster with two or more cores on a compatible 64	2026-10-08
How to check Windows 11 processor eligibility	(stated in a Japanese-language source)	learn.microsoft.com	2026-10-08
Anker Nano 7-in-1 HDMI refresh rate at 4K	60 Hz	ankerjapan.com	2026-10-11
Anker Nano 7-in-1 maximum PD pass-through	85 W	ankerjapan.com	2026-10-11
UGREEN Uno 6-in-1 HDMI refresh rate at 4K	60 Hz	ugreen.com	2026-10-11
UGREEN Uno 6-in-1 maximum PD pass-through	85 W	ugreen.com	2026-10-11
WD My Passport SSD compatibility statement	compatible with PC and Mac	sandisk.com	2026-10-11
ASUS Zenbook Duo (UX8406CA) GPU listing	Intel® Arc™ Graphics	asus.com	2026-10-11

Minimum Windows laptop spec for your development work: quick reference

The calculator in this article shows the result for your own settings. Settings can be set in the URL; each link below opens the calculator with those settings.

Settings	Result	Calculator
Web / WSL 2 (VS Code), Not decided	✓ 8 GB memory, 256 GB SSD, x64 or Arm. Memory needed: 4 GB (official minimum for Windows 11). VS Code has no official RAM figure, so this uses the Windows 11 minimum. Memory: 8 GB or more. Memory headroom: 4 GB. SSD: 256 GB or more (from Windows 11’s 64 GB minimum; this tool publishes no free-space figure). CPU: works on either x64 or Arm. By default WSL 2 can use up to 4 GB (50% of Windows’ memory; change it in .wslconfig). Space for your own files, Docker images and what’s inside WSL isn’t calculated. Pick one kind of development. Each official RAM figure is given per app and no page gives a combined figure, so running several together isn’t judged	Open with these settings
Web / WSL 2 (VS Code), Arm (Snapdragon)	✓ 8 GB memory, 256 GB SSD, Arm OK. Memory needed: 4 GB (official minimum for Windows 11). VS Code has no official RAM figure, so this uses the Windows 11 minimum. Memory: 8 GB or more. Memory headroom: 4 GB. SSD: 256 GB or more (from Windows 11’s 64 GB minimum; this tool publishes no free-space figure). CPU: works on either x64 or Arm. By default WSL 2 can use up to 4 GB (50% of Windows’ memory; change it in .wslconfig). Space for your own files, Docker images and what’s inside WSL isn’t calculated. Pick one kind of development. Each official RAM figure is given per app and no page gives a combined figure, so running several together isn’t judged	Open with these settings
Docker, Not decided	△ 8 GB memory (borderline), 256 GB SSD, x64 or Arm. Memory needed: 8 GB (official minimum for Docker Desktop). Memory: 8 GB or more. Memory is exactly the official minimum with nothing to spare (borderline). SSD: 256 GB or more (from Windows 11’s 64 GB minimum; this tool publishes no free-space figure). CPU: either x64 or Arm. Docker labels the Arm version of Docker Desktop Early Access (Microsoft’s Windows on Arm overview says it runs natively). By default WSL 2 can use up to 4 GB (50% of Windows’ memory; change it in .wslconfig). Docker’s requirements list Windows 11 Enterprise, Pro and Education, version 23H2 or later (Home isn’t listed and wasn’t verified). Space for your own files, Docker images and what’s inside WSL isn’t calculated. Pick one kind of development. Each official RAM figure is given per app and no page gives a combined figure, so running several together isn’t judged	Open with these settings
Docker, Arm (Snapdragon)	△ 8 GB memory (borderline), 256 GB SSD, Docker on Arm is Early Access. Memory needed: 8 GB (official minimum for Docker Desktop). Memory: 8 GB or more. Memory is exactly the official minimum with nothing to spare (borderline). SSD: 256 GB or more (from Windows 11’s 64 GB minimum; this tool publishes no free-space figure). CPU: either x64 or Arm. Docker labels the Arm version of Docker Desktop Early Access (Microsoft’s Windows on Arm overview says it runs natively). By default WSL 2 can use up to 4 GB (50% of Windows’ memory; change it in .wslconfig). Docker’s requirements list Windows 11 Enterprise, Pro and Education, version 23H2 or later (Home isn’t listed and wasn’t verified). Space for your own files, Docker images and what’s inside WSL isn’t calculated. Pick one kind of development. Each official RAM figure is given per app and no page gives a combined figure, so running several together isn’t judged	Open with these settings
Visual Studio (.NET, C++, etc.), Not decided	✓ 8 GB memory, 256 GB SSD, x64 or Arm. Memory needed: 4 GB (official minimum for Visual Studio 2026). Memory: 8 GB or more. Memory headroom: 4 GB. SSD: 256 GB or more (114 GB from Windows 11’s 64 GB minimum plus the tool’s official free space). CPU: works on either x64 or Arm. Visual Studio on Arm64 supports Windows 11 only (.NET desktop, C++ desktop, ASP.NET and web, and others are supported). Visual Studio typically needs 20–50 GB. Installing every feature takes up to 210 GB, which needs a 512 GB SSD. Visual Studio recommends a quad-core or better CPU. Visual Studio recommends 16 GB (this check uses the minimum). Space for your own files, Docker images and what’s inside WSL isn’t calculated. Pick one kind of development. Each official RAM figure is given per app and no page gives a combined figure, so running several together isn’t judged	Open with these settings
Visual Studio (.NET, C++, etc.), Arm (Snapdragon)	✓ 8 GB memory, 256 GB SSD, Arm OK. Memory needed: 4 GB (official minimum for Visual Studio 2026). Memory: 8 GB or more. Memory headroom: 4 GB. SSD: 256 GB or more (114 GB from Windows 11’s 64 GB minimum plus the tool’s official free space). CPU: works on either x64 or Arm. Visual Studio on Arm64 supports Windows 11 only (.NET desktop, C++ desktop, ASP.NET and web, and others are supported). Visual Studio typically needs 20–50 GB. Installing every feature takes up to 210 GB, which needs a 512 GB SSD. Visual Studio recommends a quad-core or better CPU. Visual Studio recommends 16 GB (this check uses the minimum). Space for your own files, Docker images and what’s inside WSL isn’t calculated. Pick one kind of development. Each official RAM figure is given per app and no page gives a combined figure, so running several together isn’t judged	Open with these settings
Visual Studio (Python, data science, etc.), Not decided	✓ 8 GB memory, 256 GB SSD, x64. Memory needed: 4 GB (official minimum for Visual Studio 2026). Memory: 8 GB or more. Memory headroom: 4 GB. SSD: 256 GB or more (114 GB from Windows 11’s 64 GB minimum plus the tool’s official free space). CPU: x64 (Intel or AMD) required. On Arm64, Visual Studio doesn’t support workloads such as Python development, data science and Azure (except the container development tools). Visual Studio typically needs 20–50 GB. Installing every feature takes up to 210 GB, which needs a 512 GB SSD. Visual Studio recommends a quad-core or better CPU. Visual Studio recommends 16 GB (this check uses the minimum). Space for your own files, Docker images and what’s inside WSL isn’t calculated. Pick one kind of development. Each official RAM figure is given per app and no page gives a combined figure, so running several together isn’t judged	Open with these settings
Visual Studio (Python, data science, etc.), Arm (Snapdragon)	✕ Officially not supported on Windows on Arm. Memory needed: 4 GB (official minimum for Visual Studio 2026). On Arm64, Visual Studio doesn’t support workloads such as Python development, data science and Azure (except the container development tools). Switch the CPU type to x64 (Intel or AMD)	Open with these settings
Android (no emulator), Not decided	△ 8 GB memory (borderline), 256 GB SSD, x64. Memory needed: 8 GB (official minimum for Android Studio). Memory: 8 GB or more. Memory is exactly the official minimum with nothing to spare (borderline). SSD: 256 GB or more (72 GB from Windows 11’s 64 GB minimum plus the tool’s official free space). CPU: x64 (Intel or AMD) required. Android Studio officially says Windows machines with Arm CPUs aren’t currently supported. Android Studio says Intel Core N-series and U-series processors aren’t recommended due to insufficient performance. Space for your own files, Docker images and what’s inside WSL isn’t calculated. Pick one kind of development. Each official RAM figure is given per app and no page gives a combined figure, so running several together isn’t judged	Open with these settings
Android (no emulator), Arm (Snapdragon)	✕ Officially not supported on Windows on Arm. Memory needed: 8 GB (official minimum for Android Studio). Android Studio officially says Windows machines with Arm CPUs aren’t currently supported. Switch the CPU type to x64 (Intel or AMD)	Open with these settings
Android (with emulator), Not decided	△ 16 GB memory (borderline), 256 GB SSD, x64. Memory needed: 16 GB (official minimum for Android Studio). Memory: 16 GB or more. Memory is exactly the official minimum with nothing to spare (borderline). SSD: 256 GB or more (80 GB from Windows 11’s 64 GB minimum plus the tool’s official free space). CPU: x64 (Intel or AMD) required. Android Studio officially says Windows machines with Arm CPUs aren’t currently supported. Android Studio says Intel Core N-series and U-series processors aren’t recommended due to insufficient performance. Android Studio recommends 32 GB (this check uses the minimum). With the emulator, the official minimum GPU is one with 4 GB of VRAM. Running several virtual devices (AVDs) at once takes about 4 GB of memory each, and each extra AVD can take up to 6 GB of storage. Space for your own files, Docker images and what’s inside WSL isn’t calculated. Pick one kind of development. Each official RAM figure is given per app and no page gives a combined figure, so running several together isn’t judged	Open with these settings
Android (with emulator), x64 (Intel / AMD)	△ 16 GB memory (borderline), 256 GB SSD, x64. Memory needed: 16 GB (official minimum for Android Studio). Memory: 16 GB or more. Memory is exactly the official minimum with nothing to spare (borderline). SSD: 256 GB or more (80 GB from Windows 11’s 64 GB minimum plus the tool’s official free space). CPU: x64 (Intel or AMD) required. Android Studio officially says Windows machines with Arm CPUs aren’t currently supported. Android Studio says Intel Core N-series and U-series processors aren’t recommended due to insufficient performance. Android Studio recommends 32 GB (this check uses the minimum). With the emulator, the official minimum GPU is one with 4 GB of VRAM. Running several virtual devices (AVDs) at once takes about 4 GB of memory each, and each extra AVD can take up to 6 GB of storage. Space for your own files, Docker images and what’s inside WSL isn’t calculated. Pick one kind of development. Each official RAM figure is given per app and no page gives a combined figure, so running several together isn’t judged	Open with these settings
Android (with emulator), Arm (Snapdragon)	✕ Officially not supported on Windows on Arm. Memory needed: 16 GB (official minimum for Android Studio). Android Studio officially says Windows machines with Arm CPUs aren’t currently supported. Switch the CPU type to x64 (Intel or AMD)	Open with these settings

Best MacBook for Programming: Smallest Config by Use

Sat, 27 Mar 2021 00:00:00 GMT

For the October 2026 lineup, the smallest configuration that meets the official minimums is 8 GB of memory and 256 GB of storage. That holds with no Android emulator and up to 1 display. It’s borderline against Android Studio’s 8 GB minimum.

With up to 2 displays, an M5 configuration with 16 GB of memory and 512 GB of storage meets the official minimums, though it’s borderline with the emulator. With 3 displays you need the M5 Pro chip, and with 4 displays the M5 Max.

This covers the MacBooks Apple sells now, checked as of October 2026, without hands-on testing.

The smallest config depends on what you develop and how many displays you use

The table shows the first config, from Neo to Pro, that meets the official minimum RAM.

External displays	Web and Docker (minimum 4 GB)	Android Studio only (minimum 8 GB)	With the Android emulator (minimum 16 GB)	iOS (Xcode)
Up to 1 display	MacBook Neo, 8 GB / 256 GB	MacBook Neo, 8 GB / 256 GB, borderline	13-inch MacBook Air (M5), 16 GB / 512 GB, borderline	MacBook Neo, 8 GB / 256 GB, memory not judged
2 displays	13-inch MacBook Air (M5), 16 GB / 512 GB	13-inch MacBook Air (M5), 16 GB / 512 GB	13-inch MacBook Air (M5), 16 GB / 512 GB, borderline	13-inch MacBook Air (M5), 16 GB / 512 GB, memory not judged
3 displays	14-inch MacBook Pro (M5 Pro), 24 GB / 1 TB	14-inch MacBook Pro (M5 Pro), 24 GB / 1 TB	14-inch MacBook Pro (M5 Pro), 24 GB / 1 TB	14-inch MacBook Pro (M5 Pro), 24 GB / 1 TB, memory not judged
4 displays	14-inch MacBook Pro (M5 Max), 36 GB / 2 TB	14-inch MacBook Pro (M5 Max), 36 GB / 2 TB	14-inch MacBook Pro (M5 Max), 36 GB / 2 TB	14-inch MacBook Pro (M5 Max), 36 GB / 2 TB, memory not judged

“Borderline” means the spare RAM (the configuration’s RAM minus the minimum) is under 10% of the minimum.

Minimum MacBook configuration for your development work

Pick your kind of development, how many external displays you use and how much weight you'll carry. It rounds Docker's and Android Studio's official minimum RAM up to Apple's memory tiers and shows the smallest configuration that fits, plus one MacBook you can buy that fits.

Is the MacBook Neo enough?

The MacBook Neo (8 GB) meets the official minimums only if you skip the Android emulator and use at most 1 display. For Android Studio alone it’s exactly the 8 GB minimum, so it’s borderline. Running Docker Desktop and Android Studio together isn’t assessed, since the official pages give no combined figure.

The Neo has two USB-C ports, one USB 3 and one USB 2. The spec lists DisplayPort support for the USB 3 port.

For iOS work, Xcode has no official RAM figure, so the Neo isn’t judged for it either.

How much RAM and storage to pick

The minimum RAM is 4 GB for Docker Desktop, 8 GB for Android Studio, and 16 GB with the emulator.

With the Android emulator you need at least 16 GB. The standard 16 GB on the Air and the M5 14-inch Pro is exactly that minimum, so it’s borderline; the M5 Pro starts at 24 GB.

Android Studio recommends 32 GB of RAM. That is the largest option with the M5 chip. This article’s checks use the minimum, not the recommendation.

Each Android Virtual Device running at the same time needs about 4 GB of memory. The official pages give no total for several, so running several isn’t assessed.

Storage

Per Docker’s and Android Studio’s official requirements, Android Studio needs 8 GB of free space, 16 GB with the emulator, and recommends 32 GB or more on an SSD. Even the smallest current drive, 256 GB, is above those figures. Space taken by macOS and your own files isn’t calculated.

Chips and external displays

Per Apple’s spec pages, the Neo drives up to 1 display, and the Air and the M5 14-inch Pro drive up to 2 displays. The M5 Pro drives up to 3 displays and the M5 Max up to 4 displays, so 3 or more displays need an M5 Pro or better.

The most memory you can order is 8 GB on the Neo and 32 GB on the Air and the M5 14-inch Pro. The M5 Pro goes to 64 GB and the M5 Max to 128 GB, but only with the higher-core GPU.

Which screen size to pick

For the lightest, pick the Air 13 at 1.23 kg; for a bigger screen, the Air 15 or the Pro 16.

Model	Screen	Resolution	Weight	External displays	Max RAM
Neo	13.0 in	2,408 × 1,506	1.23 kg	1 display	8 GB
Air 13	13.6 in	2,560 × 1,664	1.23 kg	2 displays	32 GB
Air 15	15.3 in	2,880 × 1,864	1.51 kg	2 displays	32 GB
Pro 14 M5	14.2 in	3,024 × 1,964	1.55 kg	2 displays	32 GB
Pro 14 M5 Pro	14.2 in	3,024 × 1,964	1.60 kg	3 displays	64 GB
Pro 14 M5 Max	14.2 in	3,024 × 1,964	1.62 kg	4 displays	128 GB
Pro 16 M5 Pro	16.2 in	3,456 × 2,234	2.14 kg	3 displays	64 GB
Pro 16 M5 Max	16.2 in	3,456 × 2,234	2.15 kg	4 displays	128 GB

The M5 Pro’s 64 GB needs the 20-core GPU, and the M5 Max’s 128 GB the 40-core GPU.

Picks by condition

One pick per condition, from Amazon.com:

Sold new by Amazon.com, with a configuration that matches Apple’s spec
At least 100 ratings and 4 stars or more

Among the current models checked, when more than one listing qualifies, the pick is the cheapest one, shipping included.

No Android emulator, up to 1 display

For Android Studio alone it’s exactly the 8 GB minimum, so it’s borderline.

Other colors are separate listings. Whether ratings are pooled was not checked.

With the emulator, up to 2 displays, 1.3 kg or less

It’s exactly the 16 GB emulator minimum, so it’s borderline.

15.3 in screen, with the emulator, up to 2 displays

Also exactly the 16 GB minimum, so borderline.

3 displays and the emulator

Its standard 24 GB is 8 GB above the minimum.

FAQ

Is the MacBook Neo good enough for programming?

For the October 2026 lineup, the smallest-memory configuration, 8 GB, drives up to 1 display. It meets the official minimums only if you skip the Android emulator.

For Android Studio alone it’s exactly the 8 GB minimum, so it’s borderline. Running Docker Desktop and Android Studio together isn’t assessed, since the official pages give no combined figure.

Models that fit are in the picks section.

How much RAM does a MacBook need for programming?

The minimum RAM is 4 GB for Docker Desktop, 8 GB for Android Studio, and 16 GB with the emulator. Android Studio recommends 32 GB of RAM.

That is the largest option with the M5 chip. This article’s checks use the minimum, not the recommendation.

How much RAM does Xcode need for iOS development?

Not judged. Apple’s Xcode system requirements list no RAM figure.

Is 256 GB of storage enough for a programming MacBook?

Android Studio needs 8 GB of free space, 16 GB with the emulator, and recommends 32 GB or more on an SSD. Even the smallest current drive, 256 GB, is above those figures.

Space taken by macOS and your own files isn’t calculated. Docker and Xcode give no official disk figure.

How many external displays can a MacBook run?

Per Apple’s spec pages, on current MacBooks the display count depends on the chip. The M5 drives up to 2 displays, the M5 Pro 3 displays and the M5 Max 4 displays.

The 8 GB configuration drives 1 display.

How long does a MacBook battery last?

Apple’s video playback figures range from 16 hours to 24 hours across current configurations other than the M5 Max, from Apple’s own tests. The M5 Max wasn’t checked.

Data and formulas used (for verification)

How this was checked:

Sources: Apple specs for Neo, Air and Pro, and Compare Mac ／ Android Studio install requirements and Docker Desktop for Mac ／ Xcode system requirements (Apple)

Method

Memory needed = the official minimum RAM for the chosen kind of development (Docker Desktop on Mac 4 GB, Android Studio on Mac 8 GB for Studio only and 16 GB with the emulator). Apple lists no RAM figure for Xcode, so memory isn’t judged for it. The official recommendation (Android Studio 32 GB) isn’t used for the conclusions.
Minimum configuration = the smallest memory tier on Apple’s spec pages at or above the memory needed (Neo 8 GB; Air and 14-inch Pro with M5: 16, 24, 32 GB; M5 Pro: 24, 48, 64 GB; M5 Max from 36 GB). If the headroom (configured memory − memory needed) is under 10% of the memory needed, it’s marked borderline (a threshold set for this article).
External display counts are Apple’s spec-page maximums (Neo 1, Air 2, Pro with M5 2, M5 Pro 3, M5 Max 4). Weights are from Apple’s spec pages.
Storage is only compared with Android Studio’s official free-space figures (Studio only 8 GB, with the emulator 16 GB, recommended 32 GB or more on an SSD); space used by macOS and your own files isn’t calculated. Docker Desktop and Xcode publish no disk figure.
Speed and comfort weren’t tested on real hardware and aren’t judged. Picks are new listings sold by Amazon with 100+ ratings, 4+ stars, and a configuration matching Apple’s specs, cheapest including shipping.

Official values used

Item	Value	Source	Checked
MacBook Air 13-inch display size (diagonal)	13.6 in	apple.com	2026-10-08
MacBook Air 13-inch native width	2,560 px	apple.com	2026-10-08
MacBook Air 13-inch native height	1,664 px	apple.com	2026-10-08
MacBook Air 13-inch weight	1.23 kg	apple.com	2026-10-08
MacBook Air 13-inch video streaming battery life (up to)	18 hours	apple.com	2026-10-08
MacBook Air 13-inch wireless web battery life (up to)	15 hours	apple.com	2026-10-08
MacBook Air (M5) maximum external displays	2	apple.com	2026-10-08
MacBook Air 13-inch base memory	16 GB	apple.com	2026-10-08
MacBook Air 13-inch base storage	512 GB	apple.com	2026-10-08
MacBook Air (M5) maximum memory option	32 GB	apple.com	2026-10-08
MacBook Air (M5) maximum storage option	4 TB	apple.com	2026-10-08
MacBook Air 15-inch display size (diagonal)	15.3 in	apple.com	2026-10-08
MacBook Air 15-inch native width	2,880 px	apple.com	2026-10-08
MacBook Air 15-inch native height	1,864 px	apple.com	2026-10-08
MacBook Air 15-inch weight	1.51 kg	apple.com	2026-10-08
MacBook Air 15-inch video streaming battery life (up to)	18 hours	apple.com	2026-10-08
MacBook Air 15-inch wireless web battery life (up to)	15 hours	apple.com	2026-10-08
MacBook Pro 14-inch display size (diagonal)	14.2 in	apple.com	2026-10-08
MacBook Pro 14-inch native width	3,024 px	apple.com	2026-10-08
MacBook Pro 14-inch native height	1,964 px	apple.com	2026-10-08
MacBook Pro 14-inch (M5) weight	1.55 kg	apple.com	2026-10-08
MacBook Pro 14-inch (M5 Pro) weight	1.6 kg	apple.com	2026-10-08
MacBook Pro 14-inch (M5) video streaming battery life (up to)	24 hours	apple.com	2026-10-08
MacBook Pro 14-inch (M5) wireless web battery life (up to)	16 hours	apple.com	2026-10-08
MacBook Pro 14-inch (M5 Pro) video streaming battery life (up to)	22 hours	apple.com	2026-10-08
MacBook Pro 14-inch (M5 Pro) wireless web battery life (up to)	14 hours	apple.com	2026-10-08
MacBook Pro (M5) maximum external displays	2	apple.com	2026-10-08
MacBook Pro (M5 Pro) maximum external displays	3	apple.com	2026-10-08
MacBook Pro (M5 Max) maximum external displays	4	apple.com	2026-10-08
MacBook Pro 14-inch (M5) base memory	16 GB	apple.com	2026-10-08
MacBook Pro 14-inch (M5) base storage	1 TB	apple.com	2026-10-08
MacBook Pro 14-inch (M5 Pro) base memory	24 GB	apple.com	2026-10-08
MacBook Pro (M5 Pro) maximum memory option	64 GB	apple.com	2026-10-08
MacBook Pro 16-inch display size (diagonal)	16.2 in	apple.com	2026-10-08
MacBook Pro 16-inch (M5 Pro) weight	2.14 kg	apple.com	2026-10-08
MacBook Pro 16-inch (M5 Pro) wireless web battery life (up to)	17 hours	apple.com	2026-10-08
Android Studio (Mac) minimum RAM (Studio only)	8 GB	developer.android.com	2026-10-08
Android Studio (Mac) minimum RAM (with the emulator)	16 GB	developer.android.com	2026-10-08
Android Studio (Mac) recommended RAM	32 GB	developer.android.com	2026-10-08
Android Studio (Mac) free disk space (Studio only)	8 GB	developer.android.com	2026-10-08
Android Studio (Mac) free disk space (with the emulator, minimum)	16 GB	developer.android.com	2026-10-08
Android Studio (Mac) recommended free disk space (SSD)	32 GB	developer.android.com	2026-10-08
Memory per running Android Virtual Device (approx.)	4 GB	developer.android.com	2026-10-08
Extra storage per additional AVD (up to)	6 GB	developer.android.com You might need up to 6GB of additional storage for each additional AVD created	2026-10-08
Docker Desktop on Mac minimum RAM	4 GB	docs.docker.com	2026-10-08
MacBook Neo display size (diagonal)	13.0 in	apple.com	2026-10-08
MacBook Neo native width	2,408 px	apple.com	2026-10-08
MacBook Neo native height	1,506 px	apple.com	2026-10-08
MacBook Neo weight	1.23 kg	apple.com	2026-10-08
MacBook Neo video streaming battery life (up to)	16 hours	apple.com	2026-10-08
MacBook Neo wireless web battery life (up to)	11 hours	apple.com	2026-10-08
MacBook Neo memory (the only configuration listed on the spec page)	8 GB	apple.com	2026-10-08
MacBook Neo storage (model 1)	256 GB	apple.com	2026-10-08
MacBook Neo storage (model 2)	512 GB	apple.com	2026-10-08
MacBook Neo maximum external displays	1	apple.com	2026-10-08
MacBook Neo USB 3 (USB-C) ports	1	apple.com	2026-10-08
MacBook Neo USB 2 (USB-C) ports	1	apple.com	2026-10-08
MacBook Pro 16-inch native width	3,456 px	apple.com	2026-10-08
MacBook Pro 16-inch native height	2,234 px	apple.com	2026-10-08
MacBook Pro 16-inch (M5 Max) weight	2.15 kg	apple.com	2026-10-08
MacBook Pro 16-inch (M5 Pro) maximum external displays	3	apple.com	2026-10-08
MacBook Pro 16-inch (M5 Max) maximum external displays	4	apple.com	2026-10-08
MacBook Pro 16-inch base memory (M5 Pro)	24 GB	apple.com	2026-10-08
MacBook Pro 16-inch base storage (M5 Pro)	1 TB	apple.com	2026-10-08
MacBook Pro 16-inch (M5 Pro) video streaming battery life (up to)	24 hours	apple.com	2026-10-08
MacBook Air 15-inch base memory	16 GB	apple.com	2026-10-08
MacBook Air 15-inch base storage	512 GB	apple.com	2026-10-08
MacBook Pro 14-inch (M5 Max) weight	1.62 kg	apple.com	2026-10-08
MacBook Pro (M5 Max) smallest memory option	36 GB	apple.com	2026-10-08
MacBook Pro (M5 Max) base storage	2 TB	apple.com	2026-10-08
RAM figure in Xcode’s system requirements	No RAM or disk space figure is listed (the page lists macOS versions and SDKs only)	developer.apple.com	2026-10-08
MacBook Pro 14-inch (M5) maximum memory option	32 GB	apple.com	2026-10-08
MacBook Pro (M5 Max) maximum memory option	128 GB	apple.com	2026-10-08

Minimum MacBook configuration for your development work: quick reference

The calculator in this article shows the result and a matching product for your own settings. Settings can be set in the URL; each link below opens the calculator with those settings.

Settings	Result	Calculator
Web / Docker, None, Doesn’t matter	✓ The MacBook Neo meets the official minimums. Memory needed: 4 GB (official minimum for Docker Desktop). Minimum configuration: MacBook Neo, 8 GB memory, 256 GB storage. Memory headroom: 4 GB. The MacBook Neo drives up to 1 external display. The MacBook Neo meets the official minimums only if you skip the Android emulator, use at most one external display, and run one of these apps at a time. Other models that fit: 13-inch MacBook Air (M5), 15-inch MacBook Air (M5), 14-inch MacBook Pro (M5), 14-inch MacBook Pro (M5 Pro), 16-inch MacBook Pro (M5 Pro), 14-inch MacBook Pro (M5 Max), 16-inch MacBook Pro (M5 Max). Pick one kind of development. Each official minimum is for that app on its own; running Docker and Android Studio together has no official figure and isn’t judged	Open with these settings
Web / Docker, 1, Doesn’t matter	✓ The MacBook Neo meets the official minimums. Memory needed: 4 GB (official minimum for Docker Desktop). Minimum configuration: MacBook Neo, 8 GB memory, 256 GB storage. Memory headroom: 4 GB. The MacBook Neo drives up to 1 external display. The MacBook Neo meets the official minimums only if you skip the Android emulator, use at most one external display, and run one of these apps at a time. The Neo outputs video only from its single USB 3 port, so with a display connected you’re left with one USB 2 port. Other models that fit: 13-inch MacBook Air (M5), 15-inch MacBook Air (M5), 14-inch MacBook Pro (M5), 14-inch MacBook Pro (M5 Pro), 16-inch MacBook Pro (M5 Pro), 14-inch MacBook Pro (M5 Max), 16-inch MacBook Pro (M5 Max). Pick one kind of development. Each official minimum is for that app on its own; running Docker and Android Studio together has no official figure and isn’t judged	Open with these settings
Web / Docker, 2, Doesn’t matter	✓ The 13-inch MacBook Air (M5) meets the official minimums. Memory needed: 4 GB (official minimum for Docker Desktop). Minimum configuration: 13-inch MacBook Air (M5), 16 GB memory, 512 GB storage. Memory headroom: 12 GB. The 13-inch MacBook Air (M5) drives up to 2 external displays. Other models that fit: 15-inch MacBook Air (M5), 14-inch MacBook Pro (M5), 14-inch MacBook Pro (M5 Pro), 16-inch MacBook Pro (M5 Pro), 14-inch MacBook Pro (M5 Max), 16-inch MacBook Pro (M5 Max). Pick one kind of development. Each official minimum is for that app on its own; running Docker and Android Studio together has no official figure and isn’t judged	Open with these settings
Web / Docker, 3, Doesn’t matter	✓ The 14-inch MacBook Pro (M5 Pro) meets the official minimums. Memory needed: 4 GB (official minimum for Docker Desktop). Minimum configuration: 14-inch MacBook Pro (M5 Pro), 24 GB memory, 1 TB storage. Memory headroom: 20 GB. The 14-inch MacBook Pro (M5 Pro) drives up to 3 external displays. Other models that fit: 16-inch MacBook Pro (M5 Pro), 14-inch MacBook Pro (M5 Max), 16-inch MacBook Pro (M5 Max). Pick one kind of development. Each official minimum is for that app on its own; running Docker and Android Studio together has no official figure and isn’t judged	Open with these settings
Android (no emulator), 1, Doesn’t matter	△ The MacBook Neo just meets them. Memory needed: 8 GB (official minimum for Android Studio). Minimum configuration: MacBook Neo, 8 GB memory, 256 GB storage. Memory is exactly the official minimum with nothing to spare (borderline). The MacBook Neo drives up to 1 external display. The MacBook Neo meets the official minimums only if you skip the Android emulator, use at most one external display, and run one of these apps at a time. The Neo outputs video only from its single USB 3 port, so with a display connected you’re left with one USB 2 port. Storage is above Android Studio’s official free-space figures (8 GB, 16 GB with the emulator). Space for macOS and your own files isn’t calculated. Other models that fit: 13-inch MacBook Air (M5), 15-inch MacBook Air (M5), 14-inch MacBook Pro (M5), 14-inch MacBook Pro (M5 Pro), 16-inch MacBook Pro (M5 Pro), 14-inch MacBook Pro (M5 Max), 16-inch MacBook Pro (M5 Max). Pick one kind of development. Each official minimum is for that app on its own; running Docker and Android Studio together has no official figure and isn’t judged	Open with these settings
Android (no emulator), 2, Doesn’t matter	✓ The 13-inch MacBook Air (M5) meets the official minimums. Memory needed: 8 GB (official minimum for Android Studio). Minimum configuration: 13-inch MacBook Air (M5), 16 GB memory, 512 GB storage. Memory headroom: 8 GB. The 13-inch MacBook Air (M5) drives up to 2 external displays. Storage is above Android Studio’s official free-space figures (8 GB, 16 GB with the emulator). Space for macOS and your own files isn’t calculated. Other models that fit: 15-inch MacBook Air (M5), 14-inch MacBook Pro (M5), 14-inch MacBook Pro (M5 Pro), 16-inch MacBook Pro (M5 Pro), 14-inch MacBook Pro (M5 Max), 16-inch MacBook Pro (M5 Max). Pick one kind of development. Each official minimum is for that app on its own; running Docker and Android Studio together has no official figure and isn’t judged	Open with these settings
Android (with emulator), None, Doesn’t matter	△ The 13-inch MacBook Air (M5) just meets them. Memory needed: 16 GB (official minimum for Android Studio with the emulator). Minimum configuration: 13-inch MacBook Air (M5), 16 GB memory, 512 GB storage. Memory is exactly the official minimum with nothing to spare (borderline). The 13-inch MacBook Air (M5) drives up to 2 external displays. Android Studio recommends 32 GB (this check uses the minimum). Storage is above Android Studio’s official free-space figures (8 GB, 16 GB with the emulator). Space for macOS and your own files isn’t calculated. Other models that fit: 15-inch MacBook Air (M5), 14-inch MacBook Pro (M5), 14-inch MacBook Pro (M5 Pro), 16-inch MacBook Pro (M5 Pro), 14-inch MacBook Pro (M5 Max), 16-inch MacBook Pro (M5 Max). Pick one kind of development. Each official minimum is for that app on its own; running Docker and Android Studio together has no official figure and isn’t judged	Open with these settings
Android (with emulator), 2, Doesn’t matter	△ The 13-inch MacBook Air (M5) just meets them. Memory needed: 16 GB (official minimum for Android Studio with the emulator). Minimum configuration: 13-inch MacBook Air (M5), 16 GB memory, 512 GB storage. Memory is exactly the official minimum with nothing to spare (borderline). The 13-inch MacBook Air (M5) drives up to 2 external displays. Android Studio recommends 32 GB (this check uses the minimum). Storage is above Android Studio’s official free-space figures (8 GB, 16 GB with the emulator). Space for macOS and your own files isn’t calculated. Other models that fit: 15-inch MacBook Air (M5), 14-inch MacBook Pro (M5), 14-inch MacBook Pro (M5 Pro), 16-inch MacBook Pro (M5 Pro), 14-inch MacBook Pro (M5 Max), 16-inch MacBook Pro (M5 Max). Pick one kind of development. Each official minimum is for that app on its own; running Docker and Android Studio together has no official figure and isn’t judged	Open with these settings
Android (with emulator), 3, Doesn’t matter	✓ The 14-inch MacBook Pro (M5 Pro) meets the official minimums. Memory needed: 16 GB (official minimum for Android Studio with the emulator). Minimum configuration: 14-inch MacBook Pro (M5 Pro), 24 GB memory, 1 TB storage. Memory headroom: 8 GB. The 14-inch MacBook Pro (M5 Pro) drives up to 3 external displays. Android Studio recommends 32 GB (this check uses the minimum). Storage is above Android Studio’s official free-space figures (8 GB, 16 GB with the emulator). Space for macOS and your own files isn’t calculated. Other models that fit: 16-inch MacBook Pro (M5 Pro), 14-inch MacBook Pro (M5 Max), 16-inch MacBook Pro (M5 Max). Pick one kind of development. Each official minimum is for that app on its own; running Docker and Android Studio together has no official figure and isn’t judged	Open with these settings
Android (with emulator), 4, Doesn’t matter	✓ The 14-inch MacBook Pro (M5 Max) meets the official minimums. Memory needed: 16 GB (official minimum for Android Studio with the emulator). Minimum configuration: 14-inch MacBook Pro (M5 Max), 36 GB memory, 2 TB storage. Memory headroom: 20 GB. The 14-inch MacBook Pro (M5 Max) drives up to 4 external displays. Android Studio recommends 32 GB (this check uses the minimum). Storage is above Android Studio’s official free-space figures (8 GB, 16 GB with the emulator). Space for macOS and your own files isn’t calculated. Other models that fit: 16-inch MacBook Pro (M5 Max). Pick one kind of development. Each official minimum is for that app on its own; running Docker and Android Studio together has no official figure and isn’t judged	Open with these settings
iOS (Xcode), 1, Doesn’t matter	△ MacBook Neo (memory not judged). Apple lists no RAM figure for Xcode, so memory isn’t judged. The configuration below is chosen only by display count and weight. Minimum configuration: MacBook Neo, 8 GB memory, 256 GB storage. The MacBook Neo drives up to 1 external display. Whether the Neo’s 8 GB is enough for Xcode isn’t judged, since there’s no official figure. If you also run the Android emulator, it falls short of the official 16 GB minimum. The Neo outputs video only from its single USB 3 port, so with a display connected you’re left with one USB 2 port. Other models that fit: 13-inch MacBook Air (M5), 15-inch MacBook Air (M5), 14-inch MacBook Pro (M5), 14-inch MacBook Pro (M5 Pro), 16-inch MacBook Pro (M5 Pro), 14-inch MacBook Pro (M5 Max), 16-inch MacBook Pro (M5 Max). Pick one kind of development. Each official minimum is for that app on its own; running Docker and Android Studio together has no official figure and isn’t judged	Open with these settings
iOS (Xcode), 3, Doesn’t matter	△ 14-inch MacBook Pro (M5 Pro) (memory not judged). Apple lists no RAM figure for Xcode, so memory isn’t judged. The configuration below is chosen only by display count and weight. Minimum configuration: 14-inch MacBook Pro (M5 Pro), 24 GB memory, 1 TB storage. The 14-inch MacBook Pro (M5 Pro) drives up to 3 external displays. Other models that fit: 16-inch MacBook Pro (M5 Pro), 14-inch MacBook Pro (M5 Max), 16-inch MacBook Pro (M5 Max). Pick one kind of development. Each official minimum is for that app on its own; running Docker and Android Studio together has no official figure and isn’t judged	Open with these settings
Android (with emulator), 2, 1.3 kg or less	△ The 13-inch MacBook Air (M5) just meets them. Memory needed: 16 GB (official minimum for Android Studio with the emulator). Minimum configuration: 13-inch MacBook Air (M5), 16 GB memory, 512 GB storage. Memory is exactly the official minimum with nothing to spare (borderline). The 13-inch MacBook Air (M5) drives up to 2 external displays. Android Studio recommends 32 GB (this check uses the minimum). Storage is above Android Studio’s official free-space figures (8 GB, 16 GB with the emulator). Space for macOS and your own files isn’t calculated. Pick one kind of development. Each official minimum is for that app on its own; running Docker and Android Studio together has no official figure and isn’t judged	Open with these settings
Web / Docker, 3, 1.3 kg or less	✕ No model fits. Memory needed: 4 GB (official minimum for Docker Desktop). No model in this weight range supports 3 external displays. Set portability to “Doesn’t matter” to see one	Open with these settings
Android (with emulator), 4, 1.7 kg or less	✓ The 14-inch MacBook Pro (M5 Max) meets the official minimums. Memory needed: 16 GB (official minimum for Android Studio with the emulator). Minimum configuration: 14-inch MacBook Pro (M5 Max), 36 GB memory, 2 TB storage. Memory headroom: 20 GB. The 14-inch MacBook Pro (M5 Max) drives up to 4 external displays. Android Studio recommends 32 GB (this check uses the minimum). Storage is above Android Studio’s official free-space figures (8 GB, 16 GB with the emulator). Space for macOS and your own files isn’t calculated. Pick one kind of development. Each official minimum is for that app on its own; running Docker and Android Studio together has no official figure and isn’t judged	Open with these settings

Aulvem Blog

What Size Tablet for Reading? 8, 10 or 11 Inch

Quick reference by what you read

See how big a book page shows on a tablet

Screen size for novels and manga

How big a screen PDFs and magazines need

Portrait or landscape: which is better for reading?

Weight difference and eye comfort

Picks by what you read

Novels and shinsho manga on an 8-inch tablet

B6 manga at paper size on a budget

A4 PDFs on one portrait page, on a 10-11-inch tablet

A4 PDFs and magazines near paper size

FAQ

What size tablet is best for reading?

Is an 8-inch tablet big enough for novels?

How big does a tablet need to be for manga?

What size tablet can read PDF textbooks and papers?

Is a cheap Android tablet fine for reading?

Best Earbuds Mic for a Noisy Office: Look Past ANC

ANC and call noise reduction are separate

Will it remove chatter, keyboards, air conditioning, street noise and wind?

Call features and wind measures the makers publish

Does Teams certification tell you how well noise is removed?

Why call audio sounds muffled on Windows

How to check for your own office

Check which earbuds publish call features for a noisy office

Picks by situation

For calls in a noisy office

For calls on an iPhone and a Mac

FAQ

Does strong ANC mean the earbuds’ mic also cuts the noise around me on calls?

Do more mics mean cleaner calls?

Will earbuds cut keyboard clatter or air-conditioning hum on calls?

Do Teams-certified earbuds also remove more noise?

What Size Power Bank Do You Need for a 3-Day Trip?

Power bank size for a 3-day trip at a glance

Why a power bank charges less than its label says

Charging at the hotel, and not charging

How to check your own devices

Calculate the power bank size for your trip

Taking a power bank on a plane

Picks for each situation

Phone only, and you can charge at the hotel each night

A tablet or handheld console besides your phone, and you can charge at the hotel each night

A laptop too, and you can charge at the hotel each night

FAQ

Is 10,000mAh enough for a 3-day trip?

Do I need 20,000mAh for a 3-day trip?

What size power bank do you need if you can’t charge for 3 days?

How many times can a 10,000mAh power bank charge my phone?

What size power bank should two people share?

Can I bring a power bank on a plane?

What Monitor Does VALORANT Need? 144Hz vs 240Hz vs 360Hz

New frames on screen, by average fps and Hz

What the pros use

How to measure your fps in VALORANT

VALORANT fps-to-monitor calculator

Where 24 inches ends and 27 begins

How to choose between 1080p and 1440p

Response time and panel type

Variable refresh rate and Reflex

Picks by situation

Casual play

Serious ranked play

Competitive play, 300Hz+ over DisplayPort, average 300 fps+, from 51.59 cm (20.4 in)

FAQ

Is 144Hz enough for VALORANT? Does a high-Hz monitor help if my fps is low?

What’s the difference between 240Hz and 360Hz? Do you need 360Hz?

Pros use 360Hz. Should regular players copy them?

Should you get a 24-inch or a 27-inch monitor?

1080p or 1440p?

Do you need G-SYNC, FreeSync or NVIDIA Reflex? What if fps goes above the Hz?

My top country in GA4 was a bot — identifying Singapore traffic without ASN data

What GA4 showed

The screen size gave it away

Cloudflare’s free plan without ASN

The beacon’s Chrome versions against real readers

Which posts the bot was reading

What I changed, and what I left