What these tests actually measure
Every check on this site runs inside the browser tab you already have open. There is nothing to install, no driver to replace and no account to create. A test starts when you allow it and stops when you close the page.
That also sets the boundary. A browser can only read the values the operating system and the browser choose to expose. It can report the polling rate a mouse actually sends, the keys a keyboard actually transmitted, or the resolution of the track a camera delivers in this session. It cannot read a sensor's rated specification, measure a cable, or explain why a value is lower than the one printed on the box.
How to run a check
- Pick the test that matches the symptom, not the part name. "My clicks feel delayed" is a latency question; "my keyboard types twice" is a chatter question.
- Allow access, keep the tab focused and let the test run for a few seconds before reading a number. A reading taken in the first second is usually warm-up.
- Change one condition — another port, another browser, another device — and run the same test again. Two readings taken under the same conditions simply repeat the same answer.
Start from the symptom
- Keys repeating, doubling or dropping characters: use the keyboard chatter test.
- Clicks that feel late or arrive in the wrong order: start with the keyboard latency and mouse latency tests.
- A cursor that stutters in a game but is fine on the desktop: read the polling rate test.
- A screen that looks wrong at one fixed colour: run the screen and dead pixel tests.
- A camera that looks softer than the box promised: open the webcam test, then the webcam resolution guide.
- Sound from one ear only, or a microphone nobody can hear: the headphone and microphone tests separate the device from the call app.
Reading a result without over-claiming
A single number is an observation about one session, not a verdict about the hardware. If a mouse reports 125 Hz here and 1000 Hz in the manufacturer's own utility, both readings can be correct — they were taken through different layers, and neither one proves the other wrong. Record what you measured, where you measured it and what you changed between runs. That record is what support asks for, and it is the only thing that turns a number into a case.
Privacy: nothing is uploaded
No measurement on this page is stored on a server. The tests read values the browser already exposes and print them in the tab; there is no account, no upload and no profile built from your readings. The privacy page lists what each test reads and what it never asks for.
When a browser test is not enough
Some faults live below the browser: a failing port, a firmware bug, a display driver, a cable that only misbehaves at one refresh rate. When a test cannot separate those layers, the honest next step is the same test on known-good hardware, and then the vendor's own diagnostic tool. Each tool page states which layer it can see and which one it cannot.