Click the pad. Every press is timestamped against the next paint frame to estimate end-to-end latency. We run 30 trials, then report the distribution.
This measures click-to-paint within the browser — not the full hardware path. Real input lag also includes USB report time, OS scheduling, and display response. Expect 8–20 ms on a typical 144 Hz setup.
End-to-end latency is the time from the physical switch closing to the resulting visual change appearing on screen. This tool stamps the mousedown event using event.timeStamp and then schedules a paint via requestAnimationFrame, measuring the delta when the new frame is committed. The reported number is the browser-visible slice of the full pipeline. It does not include the USB transit (~0.5 ms), the display's own response time (1–15 ms), or scan-out — so true real-world latency is roughly 8–15 ms higher than the displayed figure on a typical 144 Hz setup.
On a 240 Hz monitor with a 1000 Hz mouse and a tuned PC, the browser-measured number should land around 4–8 ms median with a 95th percentile under 14 ms. On a 60 Hz display the floor rises to roughly 12–16 ms simply because frames are 16.7 ms apart. If the median is above 25 ms, suspect background CPU contention — closing other tabs and turning off browser hardware acceleration toggles often cuts it in half. Compare against /mouse/polling to confirm the mouse itself is not the bottleneck.
Two effects. First, browser GC pauses can spike the late-percentile numbers, which is why the tool reports min, median, and p95 separately. Second, thermal throttling on laptops reduces CPU clock under sustained load. If the median is stable but p95 climbs from 20 ms to 60 ms over 30 trials, the CPU is throttling — open Task Manager or htop and watch clock speed during the run. Plugging in a laptop charger usually restores stable numbers.
LDAT is a hardware probe attached to the screen that measures true photon latency — what a player actually sees. NVIDIA Reflex Analyzer does the same with compatible monitors and mice. Both are several milliseconds more honest than any in-browser test because they include the display panel response. This tool is a useful relative measurement (Mouse A vs Mouse B on the same setup) but should not be quoted as absolute end-to-end latency for marketing or competitive claims.
Yes. V-Sync gates frames to the display refresh and adds up to one frame of latency. G-Sync and FreeSync remove that gating and typically lower median by 3–5 ms. Disabling browser hardware acceleration shifts compositing back to the CPU and usually increases latency by 5–10 ms but stabilises it. For a clean baseline run in fullscreen on a fixed-refresh monitor with G-Sync on. The /mouse/polling tool shows the report-rate side; combined they explain most of the click-to-pixel budget.
Modern 2.4 GHz gaming dongles (Logitech Lightspeed, Razer HyperSpeed, Pulsar Xlite) add roughly 1 ms over wired, which is inside the noise floor of this test. Bluetooth LE adds 7–15 ms and will appear clearly on the histogram as a higher floor. If wireless latency reads identical to wired here, the test is working correctly — wireless really is that close now. If wireless reads 20 ms higher, the receiver is probably plugged into a USB hub or behind a metal case panel.