GPU, WebGL & WebGPU test
Verify the browser graphics path and inspect only the graphics details your browser chooses to expose.
Processed locally in this browserVerify the browser graphics path
This runs one tiny WebGL frame and reads it back. It checks API functionality—not gaming performance, GPU speed, temperature, VRAM, or a PassMark score.
Only the measurement summary is kept in this browser for 7 days and added to Device Check and the JSON report. No raw frame, cookie value or site content is stored.
What this test proves—and what it cannot
The graphics test creates a small WebGL drawing, reads it back, and reports browser-exposed graphics interfaces and limits. It is designed to verify the path used by web graphics, not to rank a GPU, estimate frame rate, or identify an exact chip when the browser hides it.
Last reviewed: 2026-09-01Signals measured
- Whether WebGL and WebGL 2 contexts can be created
- A controlled pixel render-and-read verification
- Renderer and vendor strings when the browser exposes them
- WebGPU exposure, texture limit, viewport limit, shader version, and extension count
How the test works
- A canvas requests a WebGL context without starting a sustained workload.
- The test draws a known color and reads one pixel to verify that the basic path completed.
- Known software-renderer labels are flagged as clues, not absolute proof that acceleration is disabled.
- No benchmark loop, thermal load, or comparative performance score is produced.
Interpreting the results
Measurement limits
- A passing render does not measure game, CAD, video-editing, or compute performance.
- Privacy protections may hide or generalize GPU and vendor names.
- Remote desktop, virtual machines, power mode, and driver blocklists can alter the path.
Missing or privacy-reduced detail is not scored as a fault by itself.
Questions about this test
Can this test tell me my exact GPU model?
Only when the browser exposes a sufficiently specific renderer string. DeviceTriage will not infer an exact model from a generic label such as Apple GPU.
Should I disable hardware acceleration after a failure?
Use one temporary comparison. If disabling it changes the result, investigate driver and browser status; do not treat permanent software rendering as a universal fix.