GPU Tester
The page detects which GPU the browser is using right now and which graphics APIs it can reach. Pick a workload preset and run the 15-second benchmark for average FPS and a score — only identical presets are comparable.
GPU and display
Driver-reported limits
Interactive graphics benchmark
Renders lit, textured 3D spheres with depth testing and per-pixel lighting for 15 seconds. Same preset, same scale — only identical settings are comparable.
Ready. Nothing runs until you press start.
How to read it
- The renderer string is what the browser chose to report — some browsers genericize it. On dual-GPU laptops it is the GPU the browser is using, not necessarily the fastest one installed.
- Renderer containing SwiftShader, llvmpipe or Basic Render Driver means software rendering — hardware acceleration is off or blocked and every graphics test will crawl.
- Score = average FPS × preset weight × render-scale²; the first 15 frames are warm-up and excluded. It is a browser-workload score, not a universal GPU ranking.
- On-screen rendering is vsync-capped: average FPS cannot exceed your display refresh rate. Repeat runs under the same conditions to compare.
- The test pauses itself when the tab loses focus — background-throttled numbers would be wrong.
Frequently asked questions
Why does the GPU name differ from what my OS shows?
The page reads the renderer string the browser exposes — some browsers genericize or mask it (e.g. only 'ANGLE' or 'Apple GPU'). On dual-GPU laptops it shows the GPU the browser is actually using; assign the high-performance GPU to the browser in OS settings and retest.
Is the score comparable to 3DMark or FurMark?
No. The score only reflects this page's WebGL2 workload in your browser — a fixed formula on an independent scale. Use it to compare browsers, hardware-acceleration toggles, or plugged-in vs battery on the same machine, not as a universal GPU ranking.
Why no temperature, power draw or VRAM numbers?
Browsers don't expose those to web pages, and this tool never guesses — it shows only what the browser reports plus measured frame rates. For thermals and power use the OS tools or utilities like GPU-Z / HWiNFO.
FPS capped at 60 — is something broken?
No. This benchmark renders on-screen, so it's capped by your display's refresh rate: a 60Hz panel never exceeds 60 average FPS. The score formula accounts for preset and render scale, so same-preset comparisons stay valid; for uncapped numbers use an offscreen benchmark app.
Two runs gave very different scores?
Usual suspects: power saving, busy background tabs, thermal throttling, or the browser switching GPUs mid-run. Plug in, close GPU-heavy apps, keep this tab in front, and take the middle of three same-condition runs.