Is it faster to make a box collider equal size in each axis?

Questions & Answers
1

There are cases where I want to make a box collider different sizes in each direction. For example, suppose I have a stairway that is twice as wide as it is high, and I want to place a box collider on it like a ramp. It would naturally be wide, have some length from bottom to top, and only a small thickness (like the floor).

In some cases, such as if there’s nothing under the stairs, and it’s in a hallway so there’s nothing beside it, I could just make the entire box collider use the largest dimension (i.e. the width) in all three axes.

My question: is there any performance reason for doing so? Will the box collider’s uniform size make calculations any faster?

2

No, you can have box colliders of any dimensions and get same performance results.
Note that you can stretch the collider with both transform and its own parameters. Can’t say this for sure,but perhaps using the box params instead of transform might be slightly faster, but that’s a blind guess.

In unity, you get a penalty for non-uniform scaling of meshes, because unity shaders don’t recalculate normals after such scaling. Therefore unity duplicates the mesh data and applies stretching to it.