Physics Simulation Forum

If I were to limit myself to a X sized triangle mesh (world) and Y count of primitive sphere colliders (objects), is there some way to estimate the MAX memory usage of this system? Our current issue revolves around the fact that the memory usage of Bullet's objects seems predictable, but memory usage of the simulation itself can grow dramatically as more colliders are live/interacting. Our goal is to lock down some limits so that we can fit the simulation in memory.