FPS and frame time
FPS is easy to read, but frame-time consistency is equally important. Spikes can reveal stalls even when the average FPS remains high.
WebGPU is not normally a separate download. Support comes from your browser, operating system, GPU driver and graphics hardware.
WebGPU is the modern successor to WebGL. It maps better to modern GPU APIs, supports compute workloads and can reduce CPU-side overhead in well-designed applications. It is not automatically faster in every test.
Load your graphics hardware from the browser and watch frame rate, frame pacing, artifacts and device stability.
This is a stability and behavior test, not a benchmark or GPU ranking.
Selected engine: WebGL 2. Results from different APIs should not be treated as identical benchmark scores.
GPU load can increase temperature, fan speed and battery drain. Phones and thin laptops may reduce performance by design.
Dynamic scale: small FPS changes are magnified so a real decline is easy to see.
If the Ramp stops at level 4, it means this scene stopped matching the display refresh there. It does not mean your GPU scored 4 out of 10.
At 60 Hz, the browser is normally synchronized near 60 FPS until the workload finally becomes too heavy. The zoomed scale makes the first decline much easier to spot.
WebGPU and WebGL 2 use different browser and driver paths. Compare repeat runs within the same API rather than treating the two as interchangeable benchmarks.
CamMicGo increases the graphics work required for each frame until the browser can no longer sustain the target fluidity.
Shader: a small program executed by the GPU to calculate colours, shapes, lighting and visual effects.
Full-screen pass: one complete processing pass over the test image. More passes mean more GPU work per frame.
Load is not watts, kilograms or a guaranteed GPU-usage percentage. It is a reproducible CamMicGo workload made from shader calculations and full-screen passes.
Estimated refresh rate: inferred from browser animation timing. It is not a guaranteed monitor specification read directly from the hardware. The drop threshold is currently about 88% of this estimated rate.
Current formula: shader iterations = 24 + (step × 14); full-screen passes = round(1 + step × 1.35). A step is a CamMicGo workload stage, not a physical unit or universal score.
A good test should not stop at a score. Use these practical checks to understand what the browser measured, what can affect the result, and what to change before your next call, stream or gaming session.
FPS is easy to read, but frame-time consistency is equally important. Spikes can reveal stalls even when the average FPS remains high.
The browser exposes only part of the graphics stack. Driver, browser, GPU switching and power policies can all affect the result, so this is a browser workload rather than a vendor-grade benchmark.