Controller Latency Test

Press Start, wait for the panel to turn orange, and press any button as fast as you can. The test records the milliseconds from the colour change to the browser receiving your press, repeats it for five rounds and reports the median. Switch to session B, plug in a cable or pair over Bluetooth, run five more rounds, and the tool tells you which connection was faster and by how many milliseconds.

The number for a single round is mostly your reaction time, and that is the point to understand before reading anything else. A human takes roughly 150 to 250 ms to see a colour change and move a thumb; a controller and its link add somewhere between one and a few tens of milliseconds on top. This test cannot pull those apart in one run, so it is built for comparison: same person, same screen, two setups, and the difference between them is the controller input lag you were looking for.

How to run the controller latency test

1. Connect the controller the way you want to measure first, for example over USB, and press any button so the browser reveals it. The P1 to P4 slots at the top show which pad is selected.
2. Make sure Session A is selected next to the instruction line, then press Start. The panel reads Wait for it… for a random one to three seconds. Keep every button released during the wait.
3. The moment the panel turns orange and says Press now!, press any button. The panel then shows that round's time in milliseconds and Round 1 of 5. Press Next round or click the panel for the next one.
4. Finish all five rounds. The Session A card fills with the five times, the best round and the median, which is the number to use; the median ignores one lucky or distracted round.
5. Change one thing: unplug the cable and pair over Bluetooth, or switch to a different controller and pick it in the slots. Select Session B and run five more rounds.
6. Read the comparison line under the two cards, which says which session is faster and by how many milliseconds. Reset clears both sessions when you want to start a new comparison.

A press during the wait shows Too early and cancels that round, so you cannot improve your score by mashing. A round only ends when a button goes down after the colour change, and the test also waits until every button is released before it will accept the press, which stops a held trigger from ending a round instantly.

What does the controller latency test actually measure?

The timer starts when the page switches the panel to orange and stops when the browser's Gamepad API first reports a button as pressed. Everything between those two moments is counted: the display drawing the orange frame, your eyes and brain noticing it, your thumb moving, the switch closing, the controller packaging the report, the cable or radio carrying it, the operating system handing it to the browser, and the browser noticing on its next check.

Most of that span is you. Reaction time for a visual cue is in the 150 to 250 ms range for most adults and varies from round to round, which is why the test takes the median of five rounds rather than a single reading. The device part, controller plus link plus browser, is far smaller and far more consistent. Change only the device part between two sessions and the human part cancels out, leaving the difference you care about.

The browser adds its own granularity. The test checks the controller once per screen frame, so a single round is rounded to the nearest frame: about 17 ms steps on a 60 Hz display, 7 ms on a 144 Hz one. Over five rounds the median smooths this, but it is one more reason not to read a single round as an absolute figure.

USB vs Bluetooth: which is faster for a controller?

A cable is almost always faster, and this test shows you by how much on your own hardware. Over USB the controller reports at its full rate and the operating system receives every packet directly. Over Bluetooth the controller sends reports less often, the adapter buffers them, and interference from Wi-Fi, other Bluetooth devices or a USB 3 port next to the adapter can delay or drop packets. The polling rate test measures the report rate side of this directly; the latency test measures the end result in milliseconds.

What you should expect depends on the pad. An Xbox Series or DualSense over a good Bluetooth adapter typically lands within a few milliseconds of USB, close enough that most players cannot feel it. A cheap adapter, a laptop with the antenna behind a metal lid, or a desk full of 2.4 GHz devices can push the gap past 20 or 30 ms, and that is noticeable in rhythm and fighting games. Proprietary dongles, such as the Xbox Wireless Adapter or the 2.4 GHz receivers that ship with many PC pads, usually sit between the two.

If the Bluetooth session comes out slower by more than 10 ms and you care, there are three things to try before buying a cable long enough to reach the couch: move the adapter to a USB 2 port on an extension lead away from the computer, switch the controller to its proprietary dongle if it has one, and turn off Bluetooth devices you are not using. Re-run session B after each change.

Software between the controller and the browser shows up here too. Steam Input, DS4Windows, reWASD and similar tools read the real controller and present a virtual one, and that extra hop costs a few milliseconds on every press. If you suspect one of them, run session A with the tool running and session B with it closed, same controller and same cable, and the comparison line tells you what it costs. The same trick works for a Bluetooth adapter you are thinking of replacing: measure with the old one, swap, measure with the new one.

How to read the results

The median, not the best. The best round is the one where you happened to be focused; the median is what the connection delivers in normal play. Compare medians between sessions. The tool uses the median for the comparison line for the same reason.

The spread tells you about stability. Five rounds that land within 20 ms of each other on a stable connection are normal. One round far above the rest is you blinking; several rounds scattered across 100 ms on one session but tight on the other point at dropped or delayed packets on the loose one, which is a connection problem even if the median looks fine.

The difference, not the absolute. A session reading 230 ms says nothing about your controller on its own. Session A at 230 ms and session B at 262 ms says the setup in B costs you 32 ms, and that is a figure you can act on. Keep the display, the lighting and your posture the same between sessions so the human side stays constant.

Run it twice if the gap is small. With five rounds per session, a difference under 10 ms is within the noise of your own reaction. If the comparison line reports a few milliseconds either way, swap the order, run both sessions again and see whether it holds.

Testing input lag on different controllers

The same A/B method compares two controllers over the same connection. Connect both, run session A with the first pad selected in the P1 to P4 slots, pick the second pad and run session B. Older pads and controllers in a compatibility mode often report less frequently, which shows up as a small but consistent penalty here; the Xbox controller test and PS5 controller test pages explain the modes for each brand.

Keep the conditions equal when comparing pads. Charge both controllers, use the same port or adapter, and run the two sessions one after the other on the same screen; a comparison made a week apart with a different monitor or a different chair says more about the room than about the controllers.

What the test cannot tell you

It cannot give you the absolute latency of the controller in milliseconds, because your reaction time is in every number and no web page can read the moment your thumb actually moved. It cannot see latency inside a game, where the engine's input polling and the render pipeline add their own delay on top of what the browser measures. And it measures button presses only; stick movement is not timed here, and the joystick test covers stick accuracy rather than speed.

For a general check that every button and both sticks work, the gamepad tester on the home page is the place to start; this test assumes the controller already passes it.