Will I get performance improvment if combine small collision geometry objects into 1 large object? (With rendering that’s true to some extend but this is different case)
2 Answers
2Yes, You will lose precession but gain Performance.
As a side note, the performance hit on combining geometry comes from the additional uv seams - since the gpu splits vertices on seams, more uv seams increase the vertex count for the gpu.
I don’t have any theorical knowledge to back this up, so for what it’s worth: the simple shapes (sphers and boxes) are far easier to use for the physics engine than a simplified collision mesh. We did some projects on ps3 with havok and we had some trouble doing convex objects - I think we ended up creating collision geomtry from boxes instead of creating simple enclosed models of cups, bottles, etc.
Hope that helps 
I am not sure this is true, would you care to elaborate a little?
– Lovrenc