Three.js WebGPU (The New)
Dispatches GPU compute workgroups in parallel using WGSL and Storage Buffers. Positions, velocities, and fluid viscosity updates occur entirely within high-speed VRAM with zero CPU-to-GPU PCIe memory transfers.
- Parallel GPU compute workgroups (64–128 threads/group)
- Storage Buffers in VRAM (0 bytes PCIe bus bandwidth)
- Stateful fluid physics with true inertia & momentum
- Effortlessly scales to 1,000,000 particles at 60 FPS