Coupling high-resolution SPH fluids, soft bodies, and rigid bodies usually means routing data between separate solvers, and that hop is where most multi-physics pipelines lose their real-time budget.
Our latest Carbon Fluids Prototype takes a different approach: a single, unified CPU solver that evaluates fluids, deformables, and rigid bodies in one pass: No cross-solver handoff at all.
Performance:
• Multi-physics: 100k-particle SPH fluid interacting natively with a 12.8k-triangle soft body and rigid objects at 8-9ms per step (~110-125Hz).
• Precision robotics: a manually operated Franka arm with full two-way collision against a 6k-particle SPH fluid (live water/oil separation) and rigid lab equipment at 5ms per update (~200Hz).
Why CPU, not GPU?
GPU-accelerated ecosystems like NVIDIA Isaac Sim and Newton are excellent at what they do. But when fluid, cloth, and rigid-body solvers live on different devices or kernels, the coupling step between them is often where latency creeps in. Carbon keeps everything in one unified pass on optimized, multi-threaded CPU cores instead.
That trade-off buys three things:
• No cross-solver handoff lag: fluid-deformable-rigid interactions resolve in the same pass, not across a data bridge.
• 100% Deterministic execution: repeatable runs across identical inputs, which matters for physical AI training and Sim-to-Real pipelines.
• A free GPU: the physics budget stays off-GPU, so rendering, sensor simulation, and ML inference get the full card.
If data-transfer bottlenecks or determinism issues are slowing down your multi-physics pipeline, let's talk.