Solver

Solver controls

SubstepsDefault: 8

Number of simulation steps per frame. Higher values improve the resolution of fast motion and stiff constraints, at increased calculation time.

Constraint IterationsDefault: 3

Number of constraint-solving passes per substep. Higher values improve constraint convergence but increase calculation time.

Collider SubstepsDefault: 4

Number of intermediate mesh-collider poses used within each frame. Higher values resolve moving colliders more finely. Values above Substeps have no additional effect.

Collision PassesDefault: 1

Number of mesh collision passes per constraint iteration. Extra passes correct penetration after glue moves grains, at increased calculation time.

Smoothing IterationsDefault: 0

Additional grain-overlap correction passes after the main constraint solve. Higher values reduce remaining overlaps but increase calculation time.

Velocity MarginDefault: 1.0

Additional contact margin based on collider motion during a substep. Higher values help catch grains in a moving collider's path but can hold them farther from its surface.

CUDA Graph CaptureDefault: On

Reuse a captured GPU execution sequence to reduce launch overhead in compatible simulations. Unsupported configurations use the normal execution path.

Choosing what to increase

1. For fast motion or stiff constraints, increase Substeps first and compare a short simulation.
2. For bonds that remain too soft or fail to converge, increase Constraint Iterations.
3. For a fast animated mesh collider, review Collider Substeps; values above Substeps provide no additional interpolation.
4. If glue pushes grains into a mesh collider after contact correction, try more Collision Passes.
5. Use Smoothing Iterations for residual grain overlap. Each added pass costs computation.

Velocity Margin can help catch moving colliders, but too much can make particles sit away from the visible surface. It complements the per-collider Margin setting.

When Object Activation is enabled with animated colliders, collider poses are synchronized at every solver substep. In that case, Substeps determines the interpolation rate even if Collider Substeps is lower.

CUDA Graph Capture

This option reuses a captured execution sequence when the simulation is compatible. Unsupported configurations fall back to ordinary launches; enabling it does not guarantee a speedup for every scene. Animated values can invalidate the captured sequence. Keep the option at its default unless comparing performance or diagnosing a problem.