Blender Simulation Nodes Bake vs Legacy Physics
This article explores the mechanics of Geometry Node simulation bakes in Blender and examines how they fundamentally differ from Blender’s legacy physics caching systems. While traditional physics caches rely on dedicated, fixed-algorithm solvers for cloth, particles, and fluids, Geometry Node simulation bakes record custom procedural attribute data frame by frame directly inside the node graph. Understanding these differences clarifies how modern simulation baking provides unprecedented flexibility, better pipeline integration, and granular control compared to the rigid, legacy modifier-based point caches.
What Are Geometry Node Simulation Bakes?
Geometry Node simulation bakes, introduced with the Simulation Zone workflow, store the evaluated state of procedural geometry and custom attributes over time. Instead of relying on a pre-defined physical engine (like a dedicated rigid body or cloth engine), users design the rules of the simulation by manipulating fields and attributes between a Simulation Input and Simulation Output node.
When baked, Blender calculates the node network iteratively frame by frame and saves the resulting point, edge, face, or volume data to disk or memory. Because the system is entirely attribute-driven, you can bake arbitrary data—such as custom velocities, age counters, color values, or dynamic weights—directly into the mesh cache.
What Are Legacy Physics Caches?
Legacy physics caches belong to Blender’s traditional physics solvers, which include Soft Bodies, Cloth, Rigid Bodies, Dynamic Paint, Fluid (Mantaflow), and the original Particle System.
These tools operate as black-box modifiers. When you run a
simulation, the engine calculates physics based on hardcoded equations
and writes point cache files (.bphys or OpenVDB for fluids)
either to disk or to the blend file's temporary memory. The user
interacts primarily with fixed exposed parameters, such as mass,
damping, or turbulence, with limited access to modify the intermediate
mathematical state of the simulated components.
Key Differences
1. Data Structure and Attribute Access
- Legacy Physics: Stores fixed transformation or deformation data specifically mapped to the solver's predefined schema. You cannot easily inject arbitrary data channels or transfer custom attributes between frames.
- Geometry Node Bakes: Operates on generalized
attributes. Any standard or custom attribute (e.g.,
uv_map,temperature,custom_vector) can be created, manipulated inside the loop, and written to the bake.
2. Modularity and Flexibility
- Legacy Physics: Solvers are tied directly to specific object modifiers. Combining different types of physics (e.g., cloth tearing based on particle collisions) requires complex workarounds or is entirely impossible natively.
- Geometry Node Bakes: Completely modular. Particle dynamics, deformation, procedural growth, and custom constraint solvers can live in the same node tree, interact seamlessly, and be baked through a single unified Bake node.
3. Pipeline Integration and Baking Control
- Legacy Physics: Caching is typically an all-or-nothing process bound to the object or the scene timeline. Modifying the upstream geometry invalidates the entire cache, often requiring manual resets.
- Geometry Node Bakes: Supports standalone Bake nodes placed at any point within the node tree, not just around the simulation zone. This enables partial caching, branching caches, or freezing static procedural elements before feeding them into an active simulation, significantly improving performance and iterative workflows.
4. Determinism and Playback
- Legacy Physics: Older systems often exhibit instability or non-deterministic behavior if frames are dropped during viewport playback before a full bake is completed.
- Geometry Node Bakes: Designed around strict frame-step evaluation. The Simulation Zone explicitly defines how the state at frame N-1 transitions to frame N, providing reliable, reproducible simulation states across local renders and distributed render farms.