Rendering and Collision Usage

Let me create a scenario to explain this. Let’s say I create an anchored sphere in studio, and set can collide off. Then I create a block that has the same size, position, and rotation of that sphere, and the block is fully transparent, with can collide on. Since the CPU handles collision, would it even detect the sphere at all collision wise, and would the GPU recognize the block at all?

1 Like

Huh, nice question. I would assume that parts are loaded on multiple levels, so that collision and appearance are loaded separately, this doesn’t affect the entire entity though, it still exists. In your case, the sphere’s collision would just not load, which would optimize the game by a little, and your cube would not load the appearance (or render it), which would also slightly optimize the game.

If instancing didn’t exist, I would have suggested making thousands of duplicates of both (alternate between the methods) and comparing the performance in the performance stats.

2 Likes

I’d need an update since my info is dated, but the last I heard is that there’s still a collision calculation made (albeit a small one?) even for CanCollide = false parts when something passes through it. I was told it has something to do with how rendering and collisions aren’t on the same track - though I don’t know to what extent that entails.


If you’re doing this to make a collision under layer for a MeshPart that has really inaccurate collision geometry, it’s certainly one of the better solutions.


Sorry I couldn’t answer your question completely, if you still need help let me know and I’ll dig more :slight_smile:

(And if someone who knows can confirm or deny my above info, I’d appreciate the update and fix this post!)

1 Like