GPU Stress Test
Benchmark your graphics card right in the browser. WebGL renders thousands of rotating 3D cubes with heavy shader work, measures the frame rate your GPU sustains, and turns it into a comparable score. Pick a duration and stress level, then press Start.
Pick a duration and stress level, then start the benchmark. The scene renders right here.
Scores only compare when the run matches — same duration, stress level, and screen. Re-test with identical settings after driver updates.
Plug in, turn off battery saver, and close GPU-hungry tabs and video calls — otherwise you are benchmarking the leftovers.
WebGL is sandboxed and driver-protected: it cannot damage the card, and closing the tab stops the load instantly.
A GPU can look fine on the desktop and still be why games stutter — problems only show up under load. This test applies that load safely from a web page: no download, no install, nothing sent anywhere. Use it to check a new or second-hand card, confirm a driver update kept its performance, or verify your browser is using the GPU rather than software rendering. The workload is identical every run, so score differences point at your machine, not the test.
How the test works
The page fills a WebGL scene with rotating cubes — 100 on Low up to 10,000 on Ultra — each drawn with a fragment shader that does extra per-pixel lighting math, keeping both the geometry pipeline and the shader cores busy. A short warm-up lets the browser compile shaders and the GPU leave its idle state before measurement starts. You watch the live frame rate and frame time, and at the end get the average, minimum, and maximum FPS plus the 1% low — the average of your slowest frames, which is what stutter actually feels like. A high average with a much lower 1% low means the card is fast but not smooth.
How the score works
The score is your average FPS times the stress level’s weight, times ten — holding 60 FPS at 10,000 cubes is worth far more than the same 60 FPS at 100. Your display’s refresh rate caps the frame rate, so on lighter levels most GPUs sit pinned at the cap and score alike; the heavier levels are where cards separate. Holding the cap on High or above is a strong result — try Ultra to find the limit.
| Level | Cubes | Multiplier | Roughly equivalent to |
|---|---|---|---|
| Low | 100 | ×1 | Browsing, video playback, casual 2D games |
| Medium | 500 | ×2 | Light 3D games at modest settings |
| High | 2,000 | ×4 | Mainstream 3D gaming |
| Extreme | 5,000 | ×8 | Demanding titles, high settings |
| Ultra | 10,000 | ×16 | Maximum sustained load |
Is a browser GPU stress test safe?
Yes. WebGL runs in the browser sandbox, behind the same driver-level protections as any game: the GPU throttles itself when hot, and the OS resets the driver if anything hangs. A web page cannot disable those safeguards or touch fan curves and voltages, and it stops the instant you close the tab. It also cannot push a high-end card as hard as a native benchmark, so treat the score as a comparable number, not an absolute limit.
Why is my score lower than expected?
Usually the browser never asked your real GPU for help. Check the GPU name shown above the test: if it reads “SwiftShader”, “llvmpipe”, or “Software”, hardware acceleration is off and the CPU is doing the drawing. On laptops with two GPUs, browsers often pick the power-saving integrated one — the name tells you which card actually ran. Other common causes:
| Cause | What happens | Fix |
|---|---|---|
| Hardware acceleration disabled | The CPU renders in software — scores collapse to a fraction of normal | Enable it in browser settings, then restart the browser |
| Battery saver / unplugged laptop | The GPU is clocked down to save power | Plug in and switch to a performance power mode |
| Integrated GPU selected | The browser runs WebGL on the slower of two GPUs | Set the browser to the high-performance GPU in OS graphics settings |
| Other tabs or apps using the GPU | Video calls, videos, and games share the same card | Close them and re-run for a clean number |
| Thermal throttling | FPS starts strong, then sags as heat builds | Clean vents, improve airflow; the 60s run makes this visible |
| High-resolution display | More pixels per frame means more shading work | Expected — compare scores on the same screen |
FPS, frame time, and 1% lows
Frame rate is measured like the Frame Rate Tester does — by timing every frame the browser presents — and capped by your display, so a 60 Hz panel tops out near 60 FPS; check yours with the Refresh Rate Tester. Frame time is the per-frame view (16.7 ms ≈ 60 FPS), and the 1% low averages your worst frames: the difference between a benchmark that looks good and a game that feels good.