GPU Benchmark
Test your graphics card in the browser with a GPU benchmark that renders a heavy GPU-bound 3D scene and measures the median GPU milliseconds per frame (derived FPS, an overall score and tier).
FPS over time
One bar per second of the run. Hover a bar to see that second's frame rate.
Hardware acceleration is unavailable in this browser. Neither WebGPU nor WebGL could be initialized, so the benchmark could not run.
- GPU model
- Unknown
- Vendor
- Unknown
- Engine
- —
vsync-limited — not comparable with true GPU scores
Mobile device — not comparable with desktop GPU scores
What the tiers mean
- Entry-level score 0-299 (under 20 FPS)
- Mid-range score 300-749 (20-49 FPS)
- Fast score 750-1499 (50-99 FPS)
- Very fast score 1500+ (100+ FPS)
Score = measured FPS x 15. On WebGL runs with GPU timer support, the FPS is derived from the median GPU milliseconds per frame (not limited by the monitor refresh rate). Where GPU timers are unavailable, the result is vsync-limited and not comparable with true GPU scores.
Run history
No runs saved yet. Your completed runs will appear here.
| Date | Sustained FPS | Per-frame ms | Overall | Tier | Engine | GPU model |
|---|
How to Use GPU Benchmark
Getting Started
- Click the Run Benchmark button to start the test.
- Watch the 3D preview while the GPU renders the test scene.
- Wait for the progress bar to reach 100%, then read your score, tier, and the sustained FPS.
Features
- Live 3D preview: the app renders a heavy animated scene so you can watch your GPU work, with a live FPS readout.
- Engine auto-detect: uses WebGPU when available and falls back to WebGL otherwise; the results show which engine ran.
- FPS & per-frame time: median GPU milliseconds per frame (derived FPS) via GPU-side timer queries, or the vsync-limited on-screen FPS when timers are unavailable. Warm-up frames are excluded.
- Overall score & tier: score = FPS x 15, mapped to entry-level, mid-range, fast, or very fast. The band boundaries are disclosed on the page.
- GPU info: the detected GPU model, vendor, and engine.
- Run history: past runs are saved on this device (up to 20) so you can compare over time.
- Copy results: copies a short text summary of the latest run.
Tips
- Keep the tab visible during the run so the frame timing is accurate.
- Close other heavy tabs for a cleaner reading.
- Re-run a few times to see normal run-to-run variance.
What you can do with GPU Benchmark
A few common ways people use GPU Benchmark, from a quick score to comparing runs over time.
-
Run the GPU benchmark
Click the Run Benchmark button to start the test. Watch the 3D preview and get sustained FPS, per-frame time, an overall score, and a tier within a few seconds.
-
Watch the live render and read sustained FPS
Watch the 3D scene while your GPU renders it and read a live FPS readout. The reported sustained FPS excludes a warm-up period, so it reflects steady performance.
-
Identify which GPU ran the test
See the detected GPU model, vendor, and the rendering engine in use (WebGPU or WebGL). When the browser hides the card's identity, an Unknown value is shown.
-
See where your GPU stands
Read a plain-language tier - entry-level, mid-range, fast, or very fast - mapped from your overall score, without interpreting raw numbers.
-
Track runs over time and share results
Every completed run is saved on this device (up to 20) so you can re-benchmark and compare. Copy a compact, human-readable summary of the latest run to share.
Frequently Asked Questions
- How does this GPU benchmark work?
- It renders a heavy, GPU-bound 3D scene in your browser using your graphics card and measures how long the GPU takes to render a frame. Where your browser supports it, the measured frames are wrapped in GPU-side timers (WebGL2 timer queries) and the MEDIAN GPU milliseconds per frame is computed; the FPS is derived from that median, so the result is not capped by your monitor's refresh rate. Where GPU timers are unavailable, it falls back to the on-screen (vsync-limited) FPS measurement and labels the result as such. Everything runs locally - nothing is sent to a server.
- Why does it use WebGL or WebGPU?
- We automatically use WebGPU when your browser supports it because it is the newer, more capable GPU API. If WebGPU is unavailable, we fall back to WebGL so the benchmark still works. The results panel shows which engine ran your test.
- What does the overall score and tier mean?
- We turn the measured FPS into an overall score (higher = faster, score = FPS x 15) and map it to a fixed tier: entry-level (under 20 FPS), mid-range (20-49 FPS), fast (50-99 FPS), or very fast (100+ FPS). The band boundaries are disclosed on the results page. Results measured with the vsync-limited fallback FPS are labelled and are not comparable with true GPU scores.
- Why do I see my GPU model as Unknown?
- Modern browsers often hide the exact GPU name for privacy, so they only report a generic renderer string. We show whatever the browser exposes and use Unknown when nothing detailed is available. This does not affect the score.
- Is my data safe?
- Yes. The benchmark runs entirely in your browser. Results are stored only on your own device in localStorage (if you keep history) and are never uploaded to any server.
Open data
The data behind this tool is open: see gpu-database.json in the AppDoesIt data repository.
gpu-database.json