Precise Convex vs Multiple Parts?

[MarmDev]

I’m attempting to optimise my game as much as I can now, but I was curious about performance regarding two options. Would setting a meshes collision fidelity to Precise be better than having a mesh with 1 or 2 extra parts with box collision fidelity? I need ray cast collision to be accurate since players can move furniture around and place stuff under and on objects. Below will be a badly drawn example of what I mean in regards of a table furniture:

Precise Collision (2 Parts):

Multiple Parts (3 Parts + Box/Hull Collision):

Of course there’s only a 1 part difference, but with some of the furniture I have right now it could be 3 to 5 part difference.

(Also note that the client renders these furniture, the server does not see any of the stuff players place down but instead data about it such as position for the player’s client to interpret this data)

[Aotrou (Shawn)]

Unfortunately it’s really difficult (actually nearly impossible) to accurately estimate which method is better for performance in your specific use case.

Like, I’ve seen lots of examples that are similar that have benefited from both methods, and those results only came from a lot of testing.

As you get more into performance optimization, you’ll find these “there’s no universal answer, but there is probably one that works best for you” situations happen often.

---

This is all to say that you should do a small sample of both methods and compare them. Maybe you’ll find that there’s such a small difference between the two that it isn’t worth it