How to Run a Browser GPU Stress Test and Interpret Its Limits
Run the shader stress test, compare repeated FPS samples, and understand why browser results cannot certify GPU health or measure temperature.
A browser stress test is a quick way to observe a graphics workload without installing an application. On GPUtest.org, the shader stress test runs mathematical work in a WebGL 2 fragment shader. It can reveal changes in frame delivery under that workload. It cannot certify a graphics card as healthy, read its temperature, or guarantee full utilization.
Start with a controlled short run
- Save other work and close competing games, video editors, and tabs.
- Open the test guide. Review its limitations before selecting Start test.
- Begin with the default stress level and a short duration. Keep the tab visible and the window size fixed.
- Observe FPS for 30–60 seconds. The duration is a practical observation window, not a hardware certification threshold.
- Use Stop test and return to guide or Escape to end the workload. Reduce settings if interaction becomes difficult.
The configuration panel changes shader stress level and run duration. Change only one setting between comparisons. More demanding settings may reduce FPS without indicating a fault.
What the result actually contains
| Reading | Meaning | Important limit |
|---|---|---|
| Sampled FPS | Animation callbacks per second | Includes browser scheduling and display refresh limits |
| Minimum and maximum FPS | Extremes among collected samples | A single dip needs investigation, not a diagnosis |
| Average FPS | Mean of collected FPS samples | Compare only the same workload and conditions |
| Frame consistency index | An index derived from FPS variation | Not a probability that hardware is stable |
The consistency index is clamped to the range 0–100 and computed as 100 - 10 × standard deviation of FPS samples. A value of 100 means the samples did not vary; a constant slow workload can also receive 100. Exported JSON contains these observations and settings. It contains no real temperature or utilization reading.
Investigate a change rather than guessing its cause
If FPS falls, repeat the same settings after other workloads close. Keep power mode, viewport, and browser version fixed. If the issue persists, compare another browser and a native application that exercises the same device.
A blank canvas or WebGL context loss can reflect browser memory limits, unsupported features, a driver reset, or software rendering. It does not alone prove a broken GPU. Unexpected glitches or a system restart are reasons to stop and gather evidence, not to keep increasing the load.
Temperature requires a separate tool
Use a hardware monitor supported by your device to read temperature, utilization, and clocks. Consult the manufacturer's guidance for that exact model. There is no universal temperature threshold or test duration that establishes safety for every graphics card.
See our measurement methodology and the WebGL API documentation for the browser interface used here.