Mouse DPI test

This measures mouse DPI, also written CPI (counts per inch): the number of movement counts your mouse reports for one inch of physical travel across your desk. It is the figure printed on the box and set in your mouse driver.

It has nothing to do with the DPI of an image or a print job. If you are trying to convert 300 DPI to pixels for a document, this is the wrong page and no tool here will help you.

What is mouse DPI?

Mouse DPI, dots per inch, is how many movement counts the sensor reports for one inch of travel across the desk; sensor makers call the same figure CPI, counts per inch. A mouse at 800 DPI moved one inch sends 800 counts, and with the operating system's pointer speed at its 1:1 setting those counts become 800 pixels of cursor travel. It is a resolution, not a speed: it sets how finely the mouse divides a movement, and only together with pointer speed or in-game sensitivity does it decide how far the cursor or your aim actually goes.

Why your mouse's DPI cannot be read out directly

A browser never learns what your mouse is. It receives movement counts, and those counts have already been through your operating system's pointer pipeline. Nothing in that stream says "1600 DPI." The only way to recover the number is to divide counts by a distance you actually know, which is why this test asks you to drag along something whose length is certain.

That is also why the number on the box can be wrong without anyone lying. Sensors vary between units, and the manufacturer's figure is a nominal setting, not a per-unit measurement.

How the test works

Lay a physical reference flat beside where the mouse will travel, press the left button with the mouse level with one end, drag alongside the reference to the other end, and release. The page adds up the raw movement reported during the drag and divides by the reference length. The reference is a ruler, not a track: the mouse stays on your mousepad or desk next to it. Dragging on top of a loose sheet reads low, because the sheet slips along under the mouse and the sensor never sees that part of the motion. On our bench that cost one mouse 20% of its counts across eleven runs against same-day controls, silently, with the passes inside every run still agreeing.

You do not need to drag in any particular direction, or keep the drag straight. Lay the card or sheet at whatever angle is comfortable. What is measured is the displacement from where you pressed to where you released, meaning the straight line between those two points, and that does not depend on the route your hand took to get there. Arc, wobble, overshoot and come back: only the two endpoints matter.

You do this three times, and you get a first number the moment the first pass ends. In a browser that reads the mouse directly (Chrome does), any comfortable speed is fine for all three. In a browser that cannot, the page asks for a deliberately slow, then normal, then fast pass instead; that speed variation is an acceleration test, explained below.

The references offered need no ruler, because their dimensions are fixed by standard:

• A bank card's long edge is 85.60 mm. Every credit and debit card in the world is ISO/IEC 7810 ID-1. Nothing to measure, nothing to look up.
• A4 paper's long edge is 297 mm, its short edge 210 mm (ISO 216). US Letter is 279.4 × 215.9 mm (ANSI A).
• Or a distance you measured yourself.

The card is the default because everyone has one within reach. A sheet of paper is a slightly better choice if one is to hand, for the reason given under "Alignment" below.

How to read your result

You get one number, the median of three passes, plus the spread between them. The spread is the useful part. Three passes landing within a percent or two of each other means the measurement was repeatable. A wide spread means something interfered, and this page will refuse to print a single number rather than average the mess into a confident-looking figure.

Why a DPI reading can be wrong

Pointer acceleration. Your operating system can move the cursor further when you move the mouse faster. That breaks the fixed relationship between counts and distance which this measurement depends on. This test deals with it in two ways. First it asks the browser to read the mouse directly, before the operating system adjusts anything; when the browser agrees (Chrome does), acceleration cannot reach the measurement at all, so no pass needs to be slow or fast. When the browser cannot do that, the three passes run deliberately slow, normal and fast instead. A mouse's DPI does not change with speed, so a fast pass reading far higher than the slow one means acceleration is scaling the movement before the browser sees it, and the page says that specifically instead of blaming your mouse.

If you need to switch it off at the system level, note that macOS has no checkbox for it. The Mouse settings only change tracking speed. Run defaults write .GlobalPreferences com.apple.mouse.scaling -1 in Terminal and log out and back in; reading that key back will show a value like 1.5 while acceleration is still active. On Windows it is the "Enhance pointer precision" tickbox in Mouse Properties → Pointer Options.

One honest caveat: a small slow-versus-fast difference is more often your hand than your operating system, because stopping precisely is harder when you are moving quickly, and the shorter the reference, the bigger that overshoot looks in percent. Overshooting the end of a bank card by three millimetres is already a 3.5% difference. Acceleration typically doubles or triples the gain, so only a gap far beyond any plausible overshoot is reported as a fault; a moderate one is shown as a note next to your result.

Browser. The measurement needs two things from the browser: to hold the locked pointer through three full drags, and to report movement without clamping it at the screen edge. Chrome does both and reads the mouse raw. Safari, at the time of writing, can drop the lock during a fast drag and can stop counting at the screen edge, which reads far low. If your passes disagree wildly or runs keep getting discarded, switch to Chrome before blaming the mouse.

Browser zoom. Zoom changes how movement is reported. Keep it at 100%; if it changes mid-measurement the result is discarded.

Surface. Optical sensors read some surfaces badly, glass and high-gloss finishes especially. A mousepad or bare desk is a fair test. Do not drag on the reference sheet itself: a loose sheet moves with the mouse and the counts it steals never reach the sensor. How much depends on the mouse, its feet, its weight and your desk, so the error is invisible in the spread: every pass loses the same share, all three agree, and the number is simply low.

Drag speed. Unhurried beats fast. On our bench, one budget mouse on a bare desk read at its full nominal figure on slow drags (about a second and a half per 10 cm) and a few percent lower as the drags got quicker. Fast passes reading below slow ones is a sensor losing counts at speed on that surface, which acceleration cannot cause. If you see that pattern, trust the slower passes, and try a mousepad.

Alignment, the error that hides. This is about where you start and stop, not about how straight the drag was. Three passes tell you whether your hand was steady. They cannot tell you whether it was right. If you stop two millimetres past each end every single time, all three passes agree beautifully and all three are wrong together, and no spread check can see it. That error is a fixed distance, so it is a much bigger share of a short reference than a long one: two millimetres at each end is nearly 5% of a bank card but under 1.5% of a sheet of A4. This is the one thing a longer reference genuinely helps with, and it is worth being precise about why: it does not make your passes agree more closely, it makes the same slip cost you less. So reach for the longest reference within easy reach, but do not go hunting for a longer one.

Better still, use the alignment check under the tool. Because the error is a fixed distance rather than a percentage, measuring on two references of very different lengths gives two equations with two unknowns, your DPI and your placement error, and both can be solved for. That turns the one confound this test structurally cannot see into a number you can read.

What this page does and does not claim

These are two different claims and it is worth keeping them apart.

The accuracy figures here are measured on real hardware, not assumed; how they were measured is written up separately. In repeated testing, repeated runs stayed within 5% of each other, and that same figure held on all three references we measured (a bank card, A4 paper and US Letter), even though the longest is three and a half times the shortest. That is why one figure is published rather than one per reference: the length of the thing you drag along turned out not to drive the spread. What drove it was how you drag, mostly speed, and the surface under the mouse. A hurried drag across a sheet repeats worse than a slow, careful one along a card.

The band measures repeatability, not correctness, and those are different things. A long reference still protects you from the error repeatability cannot see: stopping two millimetres off is nearly 5% of a bank card but under 1.5% of a sheet of A4, and every pass can make that same mistake together. So use a sheet if you have one, draw it slowly, and use the alignment check below, which measures your placement error instead of inferring it from agreement.

DPI covers the sensor. The button test and the double click test cover the switches, which is where most mice actually fail, and the polling rate test measures how often the sensor's reports actually arrive.

Frequently asked questions

What is my DPI?

Run the test above. There is no way for a website to read your mouse's DPI without measuring it against a known physical distance. A DPI checker or DPI analyzer that claims to detect it instantly is reading your screen resolution or guessing from your user agent, not measuring the mouse. If what you want is the number your mouse is set to rather than a measurement, the guide to checking mouse DPI lists where each brand's software shows it.

How do I change my mouse's DPI?

If your mouse has a DPI button, usually under or behind the scroll wheel, each press cycles a preset. Otherwise it is in the manufacturer's software, such as Logitech G HUB, Razer Synapse, SteelSeries GG or Corsair iCUE. Basic office mice, including most bundled and budget wireless models, have one fixed DPI and no way to change it. Your operating system's pointer speed slider is not DPI: it scales movement after the mouse has reported it, and this test can tell the difference.

Why doesn't my measured DPI match the number on the box?

A few percent is normal. Sensors vary between units and the box figure is a nominal setting. A large gap usually means pointer acceleration is on, the reference was misaligned, or browser zoom is not at 100%. Run the alignment check under the tool before concluding your mouse is mislabelled.

Is a higher DPI better?

No. Higher DPI means more counts per inch, which means the cursor travels further for the same hand movement. It is a sensitivity preference, not a quality measure, and very high settings can amplify sensor noise. What matters for aiming is effective DPI, your DPI multiplied by your game's sensitivity, which is why two players with very different DPI settings can aim identically.

Why not drag along an on-screen ruler?

Because a ruler drawn on the screen is measured in pixels, and a browser does not know how large a pixel is on your particular display, so an on-screen ruler cannot turn mouse counts into a physical distance by itself. Tools built that way still require you to move the mouse a real, separately measured distance across your desk; the picture is feedback, not the reference. Tracking a visible cursor also has two failure modes this page avoids: the counts pass through your operating system's pointer scaling unless it is set to exactly 1:1, a setting macOS does not offer, and a cursor that touches the edge of the screen stops counting while your hand keeps moving. A fixed-size physical object plus direct mouse readings needs no ruler and none of those assumptions.

Does this upload anything?

No. The measurement runs entirely in your browser and nothing about it leaves your machine. There is no account, no upload, and no server involved in the result.