DeviceBenchFree online keyboard test
Keyboard test that lights the key you actually pressed
Click the layout, press keys, and each one lights its own box on a full 104-key board with a counter for how many times it has fired. Keys are matched by switch position rather than by the character they produce, so the same box lights on QWERTY, AZERTY and QWERTZ, and the left and right Shift are two different boxes rather than one. A key that never lights is dead, a key that stays lit is stuck, and a counter that climbs by two on one deliberate tap is switch chatter — the three faults a keyboard actually has, told apart on one screen, with nothing to install and nothing leaving the tab.
- 100% free
- No signup
- 104 keys
- Any layout
- Counts every press
Your browser shortcuts still work. Press Start, then walk the board.
Nothing pressed yet.
None yet. A count here that runs ahead of your presses is switch chatter.
These never reach a web page and will stay gray however hard you press them: MetaLeft, MetaRight, PrintScreen, F11, F12, along with Fn, Alt+Tab and Cmd combinations. A phone or tablet has no keyboard to test unless you have paired a physical one.
All 68 device tests
Everything on DeviceBench runs in this tab: keyboard and mouse, display and pixels, microphone and camera, sound, controllers, reflexes, network and the calculators for a build.
Keyboard
5 toolsKeyboard Test
Press every key and watch it light up on a full keyboard layout, so a dead or sticky key gives itself away — the first of 66 hardware checks that run entirely in this tab.
You are on this page
Keyboard Checker
A full 104-key checker that confirms every switch is working: press each key and see it light up, count its presses, and spot dead, stuck or chattering keys without installing anything.
Online Keyboard
Click or tap keys on a full QWERTY layout to type text in the browser, then copy it — for phones, tablets or any moment a physical keyboard is broken or missing.
Keyboard Latency Test
Time the gap between a keypress and the frame that reacts to it, averaged over many presses — the browser's share of the delay, not the switch's own.
Key Rollover & Ghosting Test
Hold several keys down at once to find how many your keyboard reports together, and which combinations drop a key or invent one you never pressed.
Mouse
6 toolsMouse Test
Check every button in one pass — left, right, middle, the two side buttons and the wheel click — and see which ones report nothing when you press them.
Double Click Test
Click once at a time and catch the presses that register as two; every unintended double is timed in milliseconds, which is the evidence a warranty claim needs.
Scroll Wheel Test
Roll the wheel and watch each notch counted in both directions, so a wheel that jumps back a step or skips one is easy to see.
Mouse DPI Test
Drag a measured distance across the pad, enter how far the mouse really moved, and get a DPI figure — pointer acceleration keeps this an approximation, not a firmware readout.
Cursor Test
Trace the pointer's path across the screen to spot stutter and jumps, and see how much extra distance pointer acceleration adds when you move the same inch quickly.
Polling Rate Test
Count the position updates your browser receives while you move the mouse steadily, which lands near 125, 500 or 1000 Hz — the rate the page sees, not the sensor's own.
Click speed
5 toolsCPS Test
Click as fast as you can for 1, 5, 10 or 60 seconds, or the Kohi 10-second standard, and get clicks per second with your best run kept alongside.
Click Counter
Keep a running tally with no timer and no score: press to add one, step back if you overcount, and run several counts side by side.
Spacebar Counter
Tap the space bar and count the hits, timed or untimed, which is also the quickest way to find out whether a worn key is dropping presses under speed.
Right Click CPS Test
Score the right button on its own, with the context menu suppressed while the timer runs so a burst of right clicks is not interrupted by a pop-up.
Jitter, Butterfly & Drag Click Test
Measure jitter, butterfly and drag clicking as three separate runs, and watch how many clicks one drag across the button registers before your finger leaves it.
Display
15 toolsBlack Screen
Fill the display with pure black, one click and nothing else — for blanking a second screen, wiping the glass clean, or seeing how much light an LCD still emits when it is showing you nothing.
White Screen
Turn the display into one sheet of white and set its color temperature and brightness — a soft key light for a video call, and the flattest field there is for spotting dust on the glass.
FPS Test
Watch a moving object and read the frame rate your browser is actually drawing, with dropped frames counted — this is the browser's rate, capped by the panel, not your game's.
Dead Pixel Test
Cycle red, green, blue, white and black full-screen to hunt single bad pixels, then flash a fast color loop over a stuck one to try to free it.
Monitor Test
Run every test pattern back to back — solid fills, grayscale steps, geometry grids and sharpness targets — marking each one pass or fail as you go.
Screen Resolution
Read your display's resolution alongside the browser window's own size and the pixel ratio between them, which is why a 4K laptop reports 1920 to a web page.
PPI Calculator
Enter a diagonal size and a resolution to get pixels per inch, dot pitch, and the distance at which those pixels stop being separately visible.
Refresh Rate Test
Count the frames your monitor actually delivers in a second to find its true Hz — the check that catches a 144 Hz panel still running at 60.
Monitor Size Comparison
Lay two screen sizes over each other to scale — a 27-inch 16:9 next to a 34-inch ultrawide — and get the real width and height in inches and centimeters.
Monitor Color Test
Judge color by eye against gradient ramps, a gamma target and neutral gray steps, then nudge your monitor's own settings until banding and color cast disappear.
Burn-in Test
Look for a ghost of the taskbar or a channel logo burned into an OLED panel, then run a scrolling wash that can fade light image retention.
Screen Ruler
Drag a ruler across the page to measure anything on screen in pixels, and in centimeters or inches once you calibrate it against a bank card's known width.
Input Lag Test
Press the pad and see how long a click takes to reach a painted frame, split into machine wait and display wait — your own reaction time falls outside the figure, and so does everything past the browser.
Backlight Bleed Test
Dim the room and fill the screen with black to see light leaking past the panel edge, then switch to gray fields to check whether brightness is even across it.
Ghosting Test
Slide test objects across the panel at set speeds and look for the smear trailing behind them, which is slow pixel response rather than anything burned in.
Camera & mic
5 toolsMic Test
Speak and watch the level meter move, then record a few seconds and play them back to hear what the other end of the call actually gets.
Webcam Test
See the live picture at its real resolution and frame rate, switch between connected cameras, and judge focus and lighting without the video ever leaving this tab.
Video Call Test
Run camera, microphone and speakers through the checks a meeting app makes before it lets you in, plus the connection steadiness that decides whether you freeze.
Webcam Mirror
Your camera, full screen and flipped the way a real mirror flips, with no buttons or readouts in the way while you check your hair before a call.
Headset Test
Put the boom mic and both earcups through one run: left and right channels, mic level, and the silent-input warning that usually means the inline mute switch is on.
Audio
7 toolsSound Test
One button, one reference clip: confirm sound is reaching your ears at all, and see which output device the browser is sending it to.
Tone Generator
Hold a steady tone at any frequency you type in — sine, square, triangle or sawtooth — and stack a second one beside it to hear two close pitches beat.
Hearing Range Test
Step up through the frequencies and note where the tone disappears — an upper limit for your ears and your headphones together, which is not a medical hearing test.
Speaker Test
Send sound to the left speaker only, then the right, then both — enough to catch a dead channel, a swapped pair or one side playing quieter than the other.
Headphone Test
Tell left from right without guessing, hear a sound pan cleanly from one cup to the other, and catch a driver that rattles when a low note pushes it hard.
Bass Test
Sweep down through the low end to find where your speakers stop reproducing bass at all and start distorting instead — the point most desktop pairs give up.
Surround Sound Test
Move a test sound through each 5.1 and 7.1 position so a rear channel plugged into the wrong socket is obvious, on any system that exposes more than stereo.
Controllers
4 toolsGamepad Tester
Plug in any pad the browser can see and watch every button, trigger and stick report live values, whichever console or brand it came from.
Stick Drift Test
Let go of both sticks and watch what they still report — the resting offset from center, held over a minute, and the dead zone it would take to hide it.
Xbox Controller Tester
The same readout mapped to an Xbox pad's own labels — A, B, X, Y, bumpers, triggers, View and Menu — so you can tell which physical button is failing.
Joystick Test
Built for flight sticks and arcade sticks rather than pads: throttle, twist, hat switch and every extra axis a HOTAS reports, each on its own bar.
Reflex & memory
6 toolsTyping Speed Test
Type a passage for one, three or five minutes and get words per minute, accuracy, and a map of which keys cost you the most time.
Reaction Time Test
Wait for the color to change and click — five rounds, a median in milliseconds so one blink cannot move it, and a note of the false starts where you jumped early.
Aim Trainer
Hit targets as they pop up and get the numbers a shooter cares about: time to acquire each one, accuracy, and where on the screen you tend to miss.
Visual Memory Test
Squares light up in a pattern for a moment and you repeat it back, one more square added every level until you miss three times.
Chimp Test
Numbers appear scattered across the screen, then vanish behind blocks the moment you touch the first one — click them in order, and the count climbs each round.
Number Memory Test
A number appears for one second per digit, then you type it back from memory — one digit longer each round, until the span where your recall gives out.
Network
3 toolsPing Test
Measure the round trip to a nearby server repeatedly and read best, worst and median in milliseconds — done over web requests, so it reads higher than a command-line ping.
Packet Loss Test
Burst hundreds of small requests at a server and count what fails or comes back late — a browser cannot see dropped packets directly, only the delay they cause.
Jitter Test
Watch how much your latency wobbles from one probe to the next, which is what breaks a call or a game even when the average ping looks fine.
PC build calculators
6 toolsBottleneck Calculator
Pick a CPU, a GPU and a resolution to see which runs out of headroom first — an estimate from published benchmarks, with the method shown rather than hidden.
PSU Calculator
Add up what your parts actually draw under load and get a wattage with headroom for transient GPU spikes, not just the sum of the numbers on the boxes.
Mouse Sensitivity Converter
Carry the same centimeters per 360° turn from one game to the next, across the different sensitivity scales each engine uses, with eDPI worked out alongside.
FOV Calculator
Match a field of view between horizontal, vertical and diagonal at any aspect ratio, so a number copied from one game means the same thing in the next.
Bitrate Calculator
Work out the bitrate a resolution and frame rate need, the upload speed that supports it, and how many gigabytes an hour of recording will cost you.
Hz to ms
Convert a refresh rate into the milliseconds each frame gets — 144 Hz is 6.94 ms — and back again, with a table of the rates people actually run.
Device & browser
6 toolsTouch Screen Test
Draw with several fingers at once to see how many touch points the panel tracks together, and to find the dead strip where a swipe stops being picked up.
Battery Health
Read the charge level, charging state and time estimates a browser is allowed to see, plus where to look for the cycle count and wear figures it is not.
Browser Info
See the user agent your browser sends, the Client Hints brands beside it and which web features answer, in one block you can paste straight into a bug report.
Connection Status
A live indicator that flips the instant your connection drops or comes back, with the connection type and effective speed where the browser reports them.
Phone Sensor Test
Tilt, spin and shake the phone to watch the accelerometer, gyroscope and compass respond, buzz the vibration motor, and see which sensors a browser is not allowed to read.
Device Info
The specs a web page can see — CPU cores, memory, screen and platform — with a note on which of them the browser rounds off or hides for privacy.
How to test a keyboard
Three passes, in this order, because each one rules out a different fault.
Click the keyboard, then press keys
The test only starts capturing after you click the layout, which is deliberate: a page that swallows keystrokes the moment it loads takes your browser shortcuts with them. Once the border turns cyan it is listening, and every key you press lights up its box and increments the counter on it. Press Stop, or click outside the layout, and your shortcuts come back.
Walk the whole board, including the duplicates
Go row by row rather than typing a sentence — a sentence never touches Scroll Lock, the right Alt or the numpad's own Enter. Press both Shift keys, both Ctrl keys and both Alt keys separately: they are different switches and the test tracks them as different boxes, so a laptop with a dying right Shift shows up here and nowhere in normal typing.
Read the three failures off the board
A box that never lights is a dead switch. A box that stays lit after you release is stuck, or the key-up event never arrived. A counter that jumps by two on one clean press is switch chatter, the same fault a mouse has when it double-clicks on its own — the counters are there so you can see it rather than guess.
Technical specifications
| Keys on the layout | 104 — ANSI full-size, including numpad, both modifier pairs and the six-key navigation cluster |
|---|---|
| Key identification | KeyboardEvent.code, the physical switch position, not the character it types |
| Modifier handling | Left and right Shift, Ctrl, Alt and Meta tracked as separate switches |
| Auto-repeat | Excluded from the counters — a held key adds one press, not forty |
| Keys the browser cannot see | Fn, most laptop media keys, and OS-reserved combinations such as Alt+Tab and Ctrl+W |
| Per-key counter | Unbounded, shown on the key itself, cleared by Reset or a reload |
| Data kept | None — no keystroke is stored, buffered into text or transmitted |
| Price | Free, no signup, no limit on how long you test |
Frequently asked questions
Why does my Fn key not light up?
Because Fn never reaches the browser. It is handled inside the keyboard's own controller, which uses it to change what the other keys send — Fn plus F5 becomes a brightness key before anything leaves the device. The same is true of most laptop media keys and of layer keys on programmable boards. If Fn plus a key produces the wrong result, that is firmware or a BIOS setting, not something a web page can see.
Some keys do nothing here — is the keyboard broken?
Not necessarily: the operating system and the browser take a handful of keys before the page gets them. Alt+Tab, Cmd+Tab, the Windows key on its own, Ctrl+W and F11 are the usual ones, and on macOS the Cmd combinations belong to the system. Everything else should register. If an ordinary letter, number or arrow does nothing, that one really is the switch.
A key I never pressed lit up. What is that?
That is ghosting, and it means the keyboard's matrix cannot tell which of several closed switches in the same row and column you actually pressed. It shows up when three or more keys are held at once, which is why it surprises gamers rather than typists. The full version of that test, including how many keys your board reports at once, is on the key rollover page.
One press counted as two — is that the keyboard or the browser?
It is almost always the keyboard, and it has a name: switch chatter. A worn contact bounces as it closes and the controller reports two presses a few milliseconds apart. Held keys are excluded from the count here — the browser marks an auto-repeat with a flag and the counter ignores it — so a double on a single deliberate tap is real. It is the keyboard version of a mouse double-clicking on its own.
Does this work with a non-English layout?
Yes, and it is the reason the layout on screen is drawn by position rather than by letter. Every event carries a physical code — the key next to the left Shift is always KeyZ, whatever it types — so the same box lights on QWERTY, AZERTY and QWERTZ. What the key produces is shown separately, so you can also see that your layout is doing what you expect.
Can I test a keyboard before I decide whether to keep it?
That is what this page is for: plug it in, click the layout and walk the board, and you have an answer within a minute without installing a vendor utility. If you are testing a keyboard you intend to return, the counters are worth a screenshot — a shop argues with “it sometimes types double” and does not argue with a key that has fired 40 times for 20 presses.
Does the page record what I type?
No. Keys arrive as events, light a box, add one to a counter and are then gone; nothing is assembled into text, nothing is written to storage and nothing is sent anywhere. That is also why the counters reset when you reload the page.
About testing a keyboard from a browser
A browser sees a keyboard through two fields on every event, and the difference between them is the whole reason a web page can test hardware at all. KeyboardEvent.key is what the key produced — the letter, after the operating system has applied your layout, your modifiers and your dead keys. KeyboardEvent.code is which switch closed, as a fixed name tied to the physical position: the key to the right of the left Shift reports KeyZ whether it types z on QWERTY, y on QWERTZ or w on AZERTY. This page draws its layout from code, which is why it needs to know nothing about your language, and shows key beside it so you can also confirm the layout itself is behaving.
What the browser cannot see is worth knowing before you conclude anything. Fn is resolved inside the keyboard’s own controller and never becomes an event. A handful of combinations belong to the window manager — Alt+Tab, the Windows key alone, F11 in most browsers — and are taken before the page is asked. Auto-repeat, on the other hand, does arrive, forty events a second from one held key, which is why the counters here ignore any event the browser has flagged as a repeat. Without that filter a leaning elbow would look exactly like a chattering switch.
Three faults account for nearly every keyboard someone comes here to test, and each has its own signature on this board. A switch that never lights is dead — the contact no longer closes, and no amount of software changes that. A box that stays lit after your finger leaves means the key-up never arrived, which is a physically stuck keycap or debris under it. A counter that advances by two on a single deliberate press is chatter, a worn contact bouncing as it closes, and it is the fault that gets a keyboard replaced under warranty. If instead keys go missing only when you hold several at once, nothing is worn — you have hit the limit of the matrix, and the key rollover test measures how many keys your board can report together. If everything registers but feels late, the keyboard latency test times the gap between a press and the frame that reacts to it.
What happens to the keys you press
Every number on this page is worked out by JavaScript running in the tab you are reading it in. Nothing you type, paste or open is uploaded, logged or kept, which is also why the tools carry on working after you disconnect from the network.
Keystrokes here are read one event at a time to light a box and add one to a counter. They are never joined back into text, never written to storage and never sent anywhere, which is why reloading the page loses your counts — there is nowhere for them to have been kept.