DeviceBenchFree online screen test
Screen test that walks ten patterns across your display
Ten test patterns take over your whole screen one after another — five solid fills, a 17-step gray staircase, a shadow and highlight clipping strip, a geometry grid, one-pixel sharpness targets and a square moving at 960 px/s — and each one asks a single question you answer with Pass or Fail. Nothing here grades the panel for you; the run guarantees coverage and keeps the record, so at the end you have a plain-text list of what failed alongside your screen size and pixel ratio. Every failure hands you the specialist page that measures that fault properly, because this is the index of display testing rather than a replacement for it.
- 100% free
- No signup
- 10 patterns
- Pass/fail per pattern
- Copyable result
Ten patterns, about two minutes. Mark each one as you look at it, or jump straight to the one you came for.
- 1not markedIs the whole sheet one shade, with no dark speck, no patch of pink or green, and no dust you have been mistaking for a fault?
- 2not markedDoes the panel go dark all over, or is one edge glowing and one pixel still lit?
- 3not markedAny pixel sitting black on this field has a red subpixel that no longer lights — count them.
- 4not markedGreen carries most of the brightness you see, so a weak patch here is the one your eye will keep noticing later.
- 5not markedBlue is the dimmest of the three and hides nothing: mura, blotches and pressure marks show up here first.
- 6not markedCan you count all 17 steps, and does each one look neutral rather than tinted?
- 7not markedHow many of the ten bars can you separate at each end, and how gray is the black they sit on?
- 8not markedDo the hairlines reach all four edges, is the outer frame complete, and is the circle round?
- 9not markedAre the hairlines separate lines, and is the 9-pixel text legible without leaning in?
- 10not markedDoes the square stay a square, or does it smear, tear along a horizontal seam, or stutter?
What the patterns are drawn on
Reading the screen…
Before you trust step 9
A one-pixel line only survives the trip to the glass at a whole-number pixel ratio and 100% browser zoom. Yours reads 1, so the sharpness step is being drawn honestly; change the zoom and it stops being.
A browser can put a known pattern on the glass and nothing more. Which panel is behind it, how many bits it really drives and what it left the factory calibrated to are all invisible from inside a tab, so the ten marks below are your own judgment recorded in order — which is also how a shop assesses a return.
How to run a full screen test
Two minutes, ten patterns, one question each. The order matters: the patterns that need a dark room come first, while your eyes are still adapted.
Kill the reflections, then press Start
Turn off the lamp that is pointing at the glass and give your eyes about a minute to settle, because the first two patterns are a white sheet and a black one and both of them are ruined by a window behind you. Press Start and the first pattern takes the whole display; if your browser will not hand a page the full screen — iOS Safari never will — the pattern still covers the viewport and the address bar simply stays above it.
Answer the one question on the bar, then mark it
Each pattern has exactly one thing it is on screen to settle, printed on the bar at the bottom, and the bar fades after three and a half seconds so it stops tinting the field you are grading. Press P for Pass, F for Fail, arrow keys to move between patterns, or use the buttons if you are on a phone; the Exit control in the top corner is the one thing that never disappears.
Read the failures as a list of next pages
Back on the page, every row you marked Fail swaps its question for the tool that measures that particular fault — a failed black field points at the backlight bleed test, a failed grid points at your resolution, a failed moving square points at the refresh rate. Copy the result before you close the tab: it is fourteen lines of plain text with your screen size and pixel ratio at the top, which is a far better opening line for a support ticket than a photograph of a monitor.
Technical specifications
| Patterns in the run | 10, in a fixed order: white, black, red, green, blue, a gray staircase, a clipping strip, a geometry grid, a sharpness block and a moving square |
|---|---|
| Gray staircase | 17 patches from 0 to 255 in steps of 16, each one 1/17th of the screen width, edge to edge with no gaps |
| Clipping strip | Levels 0, 1, 2, 3, 4, 6, 8, 12, 16 and 24 on black; 255, 254, 253, 252, 251, 249, 247, 243, 239 and 231 on white |
| Geometry grid | One-pixel hairlines every 48 px, a 2-px frame on the outermost pixels, a center cross and a circle 40% of the shorter screen edge |
| Moving object | A 120-px white square crossing a 50% gray field at 960 px/s, driven from requestAnimationFrame so a reduced-motion setting cannot flatten it |
| What the run cannot see | Panel type, native bit depth, factory calibration and backlight luminance — a page can put a known pattern on the glass and grade nothing itself |
| Result you can keep | 14 lines of plain text: date, screen size in CSS and device pixels, pixel ratio, color depth, window size and a verdict per pattern |
| Data kept | None — the ten verdicts live in this tab's memory and a reload starts an empty run |
Frequently asked questions
Every pattern passed but the picture still looks wrong. What did I miss?
A pass/fail sweep catches faults that are visible in a still frame, and three big categories are not. Color accuracy is one: a screen with a green cast and a crushed gamma curve will sail through all ten patterns because nothing here asks you to compare a shade against a reference. Motion is another: the moving square shows you a smear exists, it does not tell you whether 12 ms of overshoot is normal for the panel. And a fault that only appears after twenty minutes — a backlight that drifts as it warms, a cable that starts dropping a link — cannot show up in a two-minute run at all. Run the color page next, and if the complaint is 'it looks bad in games' the refresh rate and frame rate pages are where that argument gets settled.
The sharpness pattern is gray mush instead of lines. Is the monitor soft?
Usually not — it is almost always the browser resampling before the panel ever sees the pattern. A one-pixel line survives the trip only when the device pixel ratio is a whole number and zoom is at 100%; at 125% scaling or a ratio of 1.5, each line is spread across one and a half device pixels and comes out gray. The panel underneath the run reports your actual ratio, so check that first. If the ratio is clean and the lines are still soft, the next suspects are a monitor running at something other than its native resolution, which the resolution page will show you, and the monitor's own Sharpness control, which most panels ship set above the neutral midpoint and which adds haloes rather than detail.
Can I run this on a television?
Yes, and it is worth doing on a TV precisely because a TV does more to the signal than a monitor does. Open the page in the set's own browser or cast the tab, then check three things before you trust a result: turn off any Overscan or Zoom setting — on Samsung it is Screen Fit, on LG it is Just Scan — because the geometry grid losing its outer frame on all four sides is overscan and not a panel fault. Switch the input's label or picture mode to PC or Game, which is what stops the set from applying edge enhancement and chroma subsampling to what is really desktop content. And turn off motion interpolation, or the moving square will be smoothed by the TV's own processing rather than by the panel.
Do I have to be in a dark room?
For four of the ten patterns, yes. Black, the clipping strip, blue and to a lesser extent the staircase all depend on your eye adapting to low light, and a lamp reflecting off the glass raises the apparent black level far more than any panel fault does. The other six — white, red, green, the grid, the sharpness block and the moving square — are fine in a normally lit room and are actually easier to judge there. If you can only do one pass, do it with the lights low and the screen at its usual brightness rather than at maximum, since maximum brightness hides shadow detail problems that you will meet again at your normal setting.
The fault only appears in one corner. Is that the screen or the computer?
Move the pattern rather than guessing: if you can drag the browser window so the affected area lands somewhere else on the panel and the fault stays with the screen position, it is the display; if it follows the window, it is the rendering path. The second way to split it is to change the route entirely — a different cable, a different port, or another device driving the same panel. A corner problem that survives a new cable and a new source is the panel or its backlight, and on a black field that is very often edge bleed rather than anything wrong with the pixels.
Why does the control bar disappear while I am looking at a pattern?
Because a bar sitting on the field changes what you are grading. A dark strip across the bottom of a white sheet makes the area next to it look brighter than it is, which is the same simultaneous-contrast effect that makes a gray square look different on black than on white, and on the uniformity patterns that is enough to invent a fault. It comes back the moment you move the mouse, touch the screen or press a key. If you want a field with genuinely nothing on it, the black screen and white screen pages exist for that: they draw one color and no controls at all.
I am testing a screen I might return. What should I record?
Copy the result before you close the tab, then photograph the failing patterns with the room lights off. The text gives the shop your screen resolution, your device pixel ratio and which of the ten patterns failed, which is a much stronger opening than a description, and the photographs give them the thing itself. One caution about pixel faults specifically: manufacturers judge those against published fault-class thresholds rather than against zero, so a single dark subpixel is very often within the specification the screen was sold under — the dead pixel page sets out how those classes are counted before you spend a week on a claim.
About test patterns and what a sweep like this is for
Test patterns are old technology used for a good reason: a display fault is almost impossible to see in ordinary content and trivial to see against a field that should be uniform. Broadcast engineers had SMPTE color bars, print had the step wedge, and the reason both survive is that each one isolates a single variable. A solid fill removes every cue except uniformity, so a patch of mura or one unlit subpixel has nothing to hide behind. A staircase removes hue and leaves only the transfer curve, which is why a crushed shadow becomes two steps that look identical instead of a vague impression that films are murky. A one-pixel line removes size, so the only question left is whether the display resolves its own smallest addressable unit. Nothing in a photograph or a desktop wallpaper does any of that, which is why a screen can look fine for months and then fail eight of these in two minutes.
The thing this run cannot do is grade anything for you, and it is worth being clear about why. Nothing in a browser reaches the display’s own description of itself — not the EDID block that names the model, not whether it drives eight bits or six with dithering, not whether the technology behind the glass is IPS, VA or OLED. All a page controls is which value goes out and all it collects is what you say came back, which answers the question most people actually arrive with — is this normal, or should it go back — and falls a long way short of a specification. For the questions that need a number rather than an opinion, the specialists take over — the monitor color test turns the staircase into a measured gamma figure and a banding verdict, the burn-in test tells a ghosted taskbar apart from permanent wear, and the FPS test counts what the moving square only suggests.
Two mistakes account for most false alarms here. The first is grading a pattern that the browser has already destroyed: at 125% Windows scaling or 110% browser zoom a one-pixel line is spread across a fraction of a device pixel and comes back as gray, and people conclude the monitor is soft when nothing has reached the monitor intact. The panel beside the run prints your device pixel ratio for exactly this reason. The second is grading in the wrong room — black level and shadow separation are judged by a dark-adapted eye, and a ceiling light reflecting off the coating will add more apparent black-level lift than any real panel defect. If you find light pooling along an edge on the black field, that is the backlight rather than the pixels, and the backlight bleed test is the page that gives you a way to judge how much is too much. If instead you found a single stubborn dot, take its coordinates to the dead pixel test, which can also flash a stuck one in the hope of freeing it.
Where the ten verdicts go
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.
The verdicts you mark exist as ten entries in this tab’s memory and nowhere else. They are assembled into text only when you press Copy, at which point they go to your clipboard and still not to us, and reloading the page starts an empty run because there was never anywhere for the old one to be stored.