Here are all the new additions and updates!

Core

• 16-bit float fields. Lower precision grids with half the memory footprint.
• Device + system memory usage. The solver can use system memory in addition to GPU memory. This is helpful for larger simulations that can't fit entirely into GPU memory. Only available when using Metal or Cuda.
• Larger neighbor tile cache sizes. Speeds up faster moving and/or higher resolution simulations.
• Faster advection tracing.
• Improved project non-divergent. Produces less artifacting and is more stable.
• Solver is anywhere from 8-32% faster overall.
• Updated to the latest NanoVDB 9.1.1 version.
• Many small performance optimizations.

API & OS Support

• Native Cuda support for Nvidia GPUs.
• Native Metal support on macOS with compatible devices.
• Native Arm support for Apple Silicon devices.

Forces

• 3rd disturbance and turbulence force.
• Disturbance seed parameter.
• Velocity drag.

Source Shapes