How to read your test results: dBFS, FPS, polling rate and more

Microphone: dBFS, peak, clipping, noise floor

The mic test and the webcam and mic test measure your signal in dBFS: decibels relative to digital full scale. 0 dBFS is the loudest value the system can store, so every real reading is negative. Each 6 dB down roughly halves the signal; 20 dB down is a tenth.

• Level is the average (RMS) loudness right now. Our convention: a full-scale sine wave reads −3.0 dBFS, a full-scale square wave 0.0 dBFS. Other tools may label the sine 0 dB; the difference is the convention, not your mic.
• Peak is the highest level in the last moment, held for 1.5 seconds so you can read it.
• Clip lights when any sample hits full scale (99.9 % or more). Clipping sounds crackly and cannot be repaired afterwards. Lower the input gain until the lamp stays dark when you speak loudly.
• −∞ means pure digital silence: the device sends zeros. Usually a muted or wrong device.

Rough guide for speech, as our rule of thumb: an average around −30 to −15 dBFS with peaks staying below about −6 dBFS is a comfortable speaking level for calls. Around −18 dBFS is healthy. Much below −40 dBFS while you talk normally, the mic is too quiet or too far away.

The verdict line uses fixed thresholds: anything above −60 dBFS counts as "we hear you"; five seconds below −80 dBFS while you are asked to speak counts as silence.

Noise floor

Press Stay quiet for 3 seconds and the test takes the median level of the silence:

• below −60 dBFS: a very quiet room or mic, great for calls and recording;
• −60 to −45 dBFS: normal for a room with a computer running;
• above −45 dBFS: a noisy room or a high input gain.

The gap between your speaking level and the noise floor is what listeners notice. As a rule of thumb, 30 dB or more sounds clean.

Microphone: sample rate, channels, processing, latency

• Sample rate (for example 48 kHz) is how many times per second the sound is measured. 44.1 and 48 kHz are both full quality for speech and music. It says nothing about how good the mic itself is.
• Channels: most mics are mono. A stereo reading on a laptop often means two built-in mics.
• Processing: Raw turns off echo cancellation, noise suppression and automatic gain; Processed (like Zoom/Discord) turns them on. Processed audio is often quieter between words and can change your level by itself, which is why your mic can look different here than in a call app.
• Latency is the browser’s estimate of delay in the audio path, in ms, where the browser reports it.

Recordings you make stay in the tab’s memory until you download them, and nothing is uploaded.

Webcam: reported vs measured FPS and resolution

A camera tells the browser what it intends to do; the test also counts what actually arrives.

• Reported FPS is the frame rate the camera driver promises, often 30.
• Measured FPS counts frames that really reach the page over a few seconds. In dim light many webcams lengthen the exposure, so 30 can drop to 15 even though 30 is still "reported". More light usually fixes it.
• Resolution shows what you get now; Max supported is the largest size the camera offers. Where the browser cannot say, Find max resolution tries 3840×2160 down to 640×480, restarting the camera a few times.
• Lighting: "Too dark", "Good" or "Overexposed", from the average brightness of the picture.

A camera can be "1080p" and still look soft: resolution is pixel count, not sharpness. Light matters more.

Keyboard and mouse: polling rate in Hz and ms

Polling rate is how often a keyboard or mouse reports to the computer. Hz and ms are two views of the same thing: interval in ms = 1000 ÷ Hz.

• 125 Hz = every 8 ms
• 250 Hz = every 4 ms
• 500 Hz = every 2 ms
• 1000 Hz = every 1 ms
• 8000 Hz = every 0.125 ms

The keyboard polling rate test looks at the gaps between your key presses: on a 125 Hz keyboard they fall on multiples of 8 ms. It reports the largest step that fits, as an estimate. The mouse polling rate test counts movement events in the busiest second while you move in circles, in buckets of 125, 250, 500, 1000 and 2000+ Hz.

Why a browser can show less than the box says

• Event merging: browsers bundle fast movement into fewer events, often one per screen frame. Chrome and Edge offer raw pointer updates; other browsers may report at most about the screen’s refresh rate. The test tells you which method it used.
• Timer rounding: to stop timing attacks and fingerprinting, browsers coarsen their clocks. Some, notably Firefox and Safari, round event times to whole milliseconds or more, which hides anything finer than 1 ms. That is why the keyboard estimate tops out at ≥ 1000 Hz (browser limit).
• Slow movement: a mouse only reports when it moves. Gentle movement gives low numbers on any mouse.

So an 8000 Hz mouse that reads "1000 Hz" in a browser is not broken. A 1000 Hz mouse that reads 125 Hz in Chrome is worth checking in the maker’s software.

Screen: refresh rate

The screen test times how often the browser draws a frame over 2 seconds and shows the nearest common rate: ≈ 60, 120 or 144 Hz. Frame time = 1000 ÷ Hz, so 60 Hz is 16.7 ms per frame and 144 Hz is 6.9 ms.

If a 144 Hz monitor reads 60 Hz, check the refresh rate setting in your system’s display settings, the cable (some older cables cap the rate), and battery saving on laptops. A busy computer or a background tab can also lower the measurement; press Measure again with the tab in front.

Gamepad: stick values and drift %

In the gamepad tester, each stick reports X and Y from −1.000 to 1.000; 0, 0 is the centre. Triggers go from 0.000 (released) to 1.000 (fully pressed).

• Drift is the largest distance from the centre while you do not touch the stick, in % of full tilt. Our verdict flags anything above 3 %.
• Suggested deadzone is drift + 3 percentage points, rounded up: drift 5.0 % gives a deadzone of 8 %.
• Circularity is the average error at the edge while you turn the stick at full tilt: 0 % is a perfect circle.

Fixes for drift are in controller not detected and stick drift.

Skill tests: CPS and reaction time

• CPS (clicks per second) = clicks ÷ seconds: 47 clicks in 5 seconds is 9.4 CPS. Try the CPS test.
• Reaction time in the reaction time test is in milliseconds; lower is faster. We show the Average, Median, Best and Spread (SD). The median is the steadiest number to compare, because one slow try moves it less than the average.
• Every reaction result includes the delay of your screen, mouse or touch screen and browser, usually a few tens of ms. Compare results on the same device only.
• The ruler drop test works without a screen: time = √(2 × distance ÷ 9.81). A catch at 20 cm means about 202 ms.

The tiers we show on skill tests are our own fun labels, not population statistics.

For problems rather than numbers, see laptop keyboard keys not working, mic and camera in Zoom, Teams and Discord or microphone not working on Windows.