ScreenTest.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

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 — same measurement engine, separate score space, pure nostalgia. And every scene can run unscored in the gallery.

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 ~1.6-second windows, 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 inthe 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 inWhat 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.