Screen Test Pro

Performance test

A scored browser benchmark that finds the heaviest particle load your device can hold at a steady frame rate. Spectacular to watch, boring in its measurement — which is the point.

Run it

Hardware stats

Top scores

last 30 days
  1. 1
  2. 2
  3. 3
  4. 4
  5. 5
  6. 6
  7. 7
  8. 8
  9. 9
  10. 10
  11. yourun the benchmark — your score slots in here

What it runs

The Standard run is five Signature scenes on WebGL2 — each stressing a different part of the pipeline — plus one deliberately plain Canvas 2D scene as a CPU-bound control:

SceneStressesLooks like
Nebula DriftGPU particle update + fill rateA million-point nebula advected through curl-noise flow.
Neon MetropolisDraw calls + instanced geometryAn instanced emissive city flythrough.
Liquid ChromeFragment shader / raymarchingRaymarched metaball fluid with mirror shading.
Starfall CascadeCPU physics + per-frame uploadVerlet debris streams colliding with moving geometry.
HypernovaMixed finaleCore collapse, shockwave, and a half-million-particle burst.
Canvas 2D Basic CirclesCanvas 2D fill + simple physicsBouncing colored circles — the honest baseline.

The eight original effects from Screen Test Pro v1 live on in Classic mode — the button above runs them under the same measurement engine, scored in their own space and never submitted anywhere: pure nostalgia, honestly measured. And every scene, both tiers, can run unscored in the effects gallery — with the particle load under your control.

How the score works

Each scene ramps its particle load up and down, searching for the heaviest load your device sustains within a fixed frame budget of 16.67 ms — checked against the 95th percentile of measured frame times over windows of 0.7–1.6 seconds, with several consecutive windows required to accept or reject a level. That "max stable load" is normalized against per-scene reference values and combined into a weighted geometric mean, giving a total score and a tier (Entry / Balanced / Fast / Extreme).

Some details worth knowing, all covered in the full methodology:

What this can and can't tell you

This is a browser workload score. It reflects your GPU, CPU, browser, OS, thermals and settings acting together — which makes it a fair test of what your device actually delivers in a browser, and a poor substitute for a native benchmark when you're comparing silicon. We wrote up the difference honestly in What a browser benchmark can and can't measure. Two quick implications: close your other tabs before running, and treat small score differences (a few percent) as noise.

The anonymous sample

Finishing a Standard run submits one anonymous row — score, coarse hardware family, screen numbers — that feeds the public hardware benchmarks. Sharing is on by default; the start card shows the exact values your device would send, and unchecking the box there runs the benchmark without submitting anything — the choice is remembered. Details in the privacy policy.

Share this page
DiscussionOne thread per language, shared across the whole site. Sign in with Google to post.
Discussion