3D Fluid Simulation Time by Resolution

3D Fluid Simulation Time by Resolution

A fluid solver chops the domain into a voxel grid and steps the physics one cell at a time. That makes total time scale as total = frames × cells × seconds_per_cell_per_frame, and for a cubic domain cells = resolution³. In production you'll see grids anywhere from 128³ for quick previews up to 1024³ and beyond on hero shots.

An HD smoke or liquid bake will usually run somewhere from 1 to 50 hours per 100 frames, and resolution, adaptive sampling and substeps decide where you land. Watch out for what doubling resolution does: cell count goes up 8× and the timestep budget roughly doubles on top of that, so the real cost jumps to around 16×.

Applications

This is the tech behind explosions, fire, smoke, oceans, splashes and viscous liquids in film and games. On the tools side, you've got Houdini FLIP/Pyro, Mantaflow inside Blender, and RealFlow when the shot is splash-heavy. NVIDIA Omniverse and GPU solvers like FluX and JangaFX EmberGen can speed up your iteration by 5 to 20 times.

FAQ

What is a domain box? It's the bounding volume that tells the solver where it's allowed to work. Nothing outside the box gets simulated, so keeping it snug around the action is how you hold cell counts down and, with them, your bake times.

What are emitters? These are the objects — meshes, particles, volumes — that feed fluid, smoke or temperature into the domain frame by frame, and that input is what keeps the simulation moving.

Why is FLIP popular for liquids? FLIP (Fluid-Implicit-Particle) splits the job: particles handle advection while a grid handles pressure. You end up with crisp splashes and you skip the smearing you'd normally get from a pure grid solver like Eulerian smoke.