F# ray tracer — benchmark reports

Rendered output and timings for the modernised renderer, measured against the frozen legacy port on the CPU and against the CUDA GPU backend.

CPU vs CPU — legacy port against the modernised renderer

Cold wall time, per-phase timings and linear-image differences across the benchmark scenes. Every repeat is retained.

Live progress page for the CPU run

GPU — detailed CUDA renders

Full-size images from the CUDA backend for the authored scenes. Refreshes itself while the run is still producing images.

GPU — material variants

The same scenes rendered across the material presets on the GPU.

Images: old versus new

Side-by-side renders, plus the plainly named PNG folders.

Supporting runs: