Hello!
- Small background case/story to highlight the issue:
Our team is currently working on a project where we have an essential need to use ’per-poly collision’ for multiple characters during runtime. The project was created in UE4 (4.26) and after some months of development proof of concept was achieved. We were able to play runtime with multiple characters performing various animations with ’per-poly collision’ activated on skeletal meshes while retaining an tolerable/good performance overall.
Ever since we converted the project to UE5 (currently 5.4 in use) there has been issues with performance related to the ’per-poly collision’ active on characters during runtime. On the post attachments there is screenshot to provide comparison on a clean ’default 3rd person starting project’ where only ’per-poly collision’ is activated for the skeletal mesh asset & actors in the level (also setting the ’Collision Presets → Collision Enabled → Query Only’ and all channel response to ’ignore’ for both UE4 and UE5 examples.
-
This issue can be also repeated with any skeletal mesh when using ’per-poly collision’ & polygon resolution has a role to play, but by default in UE5 there is a huge performance drop on this.
-
The question or discussion of this topic is; ’How we would be able to get back similar performance in Unreal 5 we had during Unreal 4 by having the ’per-poly collision’ activated for ~several skeletal mesh actors during runtime?’
-
Example of things we have tested:
-Toggling Lumen from Global Illumination & Reflection Method Off (this didn’t had much impact on the new/clean test projects)
-Toggling some other project settings to see what would be most demanding for performance draining. (No discovery of the root cause yet)
-Using default UE4 & UE5 mannequins only for polygon resolution testing cases (example UE5 mannequin with ~90k triangles is equivalent for our project use cases with characters) and default clean starting projects such as 3rd person sample etc.
Thank you for responses!
-Mikko