GPU Benchmark
A GPU benchmark is a repeatable score. You lock the test, the complexity, the browser, and the thermal state, then write down average FPS. Extreme on a hot phone next to Standard on a cooled PC is not a compare. Use a timed run on the 3D rendering test, or Volume Shader at Standard if you want the classic volume load.
A fair GPU benchmark method
Compares go wrong when the variables drift. Prep the device first. Close extra tabs. Turn off battery saver. Let both devices cool. Note whether a laptop is plugged in. Then run the same URL at the same level. If you want heat more than a score, use a GPU stress test instead of this protocol.
- Same test URL and same complexity. Standard or Medium is the default.
- Same browser family if you can. Chrome and Safari differ.
- Turn off battery saver. Note plugged in versus battery on a laptop.
- Run 60 seconds on the 3D test. Write down average FPS, min FPS, and the verdict.
- Repeat twice after a short cool down and average the averages.
What to write down
This site does not upload a global leaderboard. A useful note is average FPS, min FPS if the HUD shows it, test name, complexity, browser, and whether the device was plugged in. Screenshot the HUD if you want a clip.
Do not mix Volume Shader FPS with 3D test FPS. They are different pipelines. Fair compare conditions are in how to interpret and compare results.
What the number means
Higher average FPS is better. Stability and min FPS matter as much as the peak. A 120 FPS reading on a 120 Hz screen at Medium often means headroom. Raise Ultra before you call the card unlimited. Volume Shader Standard is the usual comparison for the volume scene. The 3D test is better when you want a clocked duration, frame time, and a headroom or strained verdict.
Bands for the volume counter are in what FPS means. A falling number during a long hold is throttling, not a benchmark result.
When to use 3DMark or Furmark instead
Published desktop databases, driver regression, and power limit hunting still belong to native benches. 3DMark Time Spy, Steel Nomad, and similar suites exist so reviewers can compare cards with a shared scene. Furmark exists to hold a desktop GPU at the wall. A browser GPU benchmark is a fast, cross device check, including phones that cannot install those tools.
Use both when you have both. Use this site when you only have a browser. See browser vs Furmark and 3DMark.
Questions
This site does not upload a global leaderboard. Write down average FPS, test name, complexity, browser, and whether the device was plugged in.
Yes. The web tests do not require payment.
No. Extreme and Ultra are stress levels. Use Standard on Volume Shader or Medium on the 3D test unless both devices already look easy at those levels and you agree to raise them together.