Three.js Post-Processing: 192 Desktop Trials Analyzed

Enabling post-processing effects often leaves a browser running at roughly 60 frames per second, but it does not leave the underlying rendering workload unchanged. Developer Alicia conducted a fresh test running 192 page trials across three distinct scenes, covering N8AO, bloom, tone mapping, volumetric beams, reflections, and render-buffer resolution to evaluate performance impacts beyond the standard FPS counter.

One of the most useful findings was a negative control: setting bloom intensity to exactly zero produced the exact same Reliquary image output as removing bloom entirely, yet it retained 16 extra draw calls. Furthermore, timing results carried an important warning that repeated desktop runs sometimes varied more than the small effect being measured, keeping those data ranges fully visible rather than offering a universal table of milliseconds.

Stock photo for illustration only, not from the actual event

The starting point for the experiments was pinned to commit 868539f of the Dioramas project. Its Reliquary and Kōen pages were adapted to freeze the initial hero camera and scene state. Reliquary provides reflective flooring and volumetric light, while Kōen features a textured bonsai and a contrasting bright background. A third owned fixture places 1,600 instances of a Houdini tree inside a separate WebGL scene while keeping diagnostic vertex colors.