3D Rendering Test

What this 3D test measures

This is a raster GPU test: a dense tetrahedral grid, per pixel lighting, and extra full draws each frame. So, vertex work and rasterization dominate. That is closer to how games and 3D apps spend GPU time than a full screen Mandelbulb. Volume Shader remains the ray marched scene if you want that clip. Open it from Run Volume Shader.

A 120 FPS reading on a 120 Hz screen often means the display is the cap, not the GPU. The HUD names that headroom. Raise the level until FPS falls. That drop is the useful score. Write down the verdict at Medium, then at Ultra. Falling FPS over a two minute run is heat. A fair timed compare is the GPU benchmark method. A long Ultra hold is a GPU stress test.

• FPS and frame time. Frame time is the honest number when FPS sticks to 60 or 120.
• Verdict: headroom (raise the level), holding, strained (the score), overloaded (drop a level)
• Stability: share of samples near the mean FPS
• Passes: extra draws per frame so a fast GPU cannot hide behind vsync

Levels and duration

Low is a smoke check. Medium is the comparison point: same browser, cooled device, write the average FPS and the verdict. High and Ultra add meshes, lighting loops, and extra passes. If Ultra still says headroom, this card is above the test. Compare two devices only at the same level. If shading looks like the bottleneck, switch to the shader test.

How to read it

A timed run prints a one line summary when the timer ends: average FPS, min FPS, and the verdict. 50 and up at Medium is holding. Strained is still a real score. Overloaded means drop a level. Headroom means the screen, not the GPU, stopped the counter.