GPU Benchmark

Stress the browser GPU with animated scenes, FPS meter and stability verdict.

Private by design: every scene is generated procedurally inside your browser. No model, texture, result or identifier ever leaves this page.

Graphics load ahead

This test renders thousands of particles and shapes for a few seconds. The device may warm up and the fan may spin — cancel any time with the button that stays on screen.

367Tools graphics score

Pick a length and press start. The frame above is a static preview — the run is fully animated.

Stress any browser GPU in half a minute

  1. Press start on Quick (~10 s) or Full (~30 s) — the page detects WebGL 1, WebGL 2 and WebGPU first.
  2. Watch seven scenes in a row: particles, growing geometry, procedural textures, transparency, lighting, simplified shadows and light post-processing.
  3. Weak devices drop quality automatically; dangerous frame times stop the test before the page freezes.
  4. Read average, min and max FPS plus p95/p99 frame times — percentiles expose stutter that averages hide.
  5. Compare the score (0–1000) across browsers on the same machine to find your fastest setup.
  6. Copy the summary or rerun after cooling down.

GPU benchmark questions

Why is my FPS lower than in games?

Browsers add sandbox and compositing overhead, and this test draws deliberately heavy scenes. Compare browsers against each other here, not against native games.

What do p95 and p99 mean?

The frame time that 95% (or 99%) of frames beat. A good average with a bad p99 means visible stutter — that is the number that predicts smoothness.

Why did quality drop by itself?

Below ~40 fps the test halves the load so weak devices still finish with a meaningful score. The final quality level is printed next to the result.

Can this overheat my device?

It is a short burst of shader math, like a demanding web game. The auto-stop guards the page, and cancel is always one tap away.