We are currently reworking our production pipeline for assets and their collision setups in UE5. One question that has come up during this process is regarding performance and efficiency:
Is it more efficient to use multiple simple collision shapes (e.g., boxes, spheres, capsules) compared to a single unique convex collision shape for an asset?
We are interested in understanding:
How the engine handles collision checks in both scenarios.
Any best practices or recommendations from your experience.
Data or benchmarks that might illustrate the performance impacts of these two approaches.
Specifically, we want to know if multiple simple collisions provide a tangible performance benefit over a single convex collision and under what circumstances one approach might be preferred.
Any insights, technical details, or references to existing documentation or studies would be greatly appreciated!
Unreal handles collision queries per collision shape.
Simple collision shapes are computationally cheaper for collision / physics as their shapes are optimized for fast checks. The main con behind simple collision shapes is the potential to be less accurate on complex geometry objects.
Unique convex collision shapes can be more precise. It may be more efficient in some cases where the object is complex and too many simple collision shapes causes more checks meaning more processing.
Unfortunately, I was unable to find any comparison between simple collision shapes vs a single unique convex collision shape - most of the information is simple vs complex collision. This video was informative though: How to Optimize Collision Performance | Community tutorial
You’d need to test it out on your end using profiling tools such as Unreal Insights to see what impacts the collision shapes will have on your project.
Ideally, you mix and match based on the needs of your project.