Hi there, we’re currently experiencing some strange performance issues using nanite in 5.0.3.
Our current level design setup uses a large number of modular static meshes, all set with nanite enabled.
Strangely though, in a published build of the game, setting ‘r.nanite 0’, seems to lower the ms, even if the now non-nanite meshes have no LODs set up, when profiling on lower end PC’s as well as the Xbox One. Nanite does however seem to lower the ms on higher-end PC build although not by a very noticeable amount.
(In the editor though, nanite enable does have a significant performance gain but this isn’t important.)
- We are NOT using Lumen.
- We are using a dynamic movable light and skylight.
- Dynamic shadow distance movable set to 5000 with 3 cascades.
- Distance fields and far shadows are on.
- Global Illumination is set to ‘none’
*Reflections are set to Screen Space
*We are using standard ‘Shadow Maps’ and not ’ Virtual Shadow Maps’
*Virtual Texture Streaming is turned ON on all asset textures. - The Landscape is using World Partition
- HLODS are turned on using an Instanced HLOD Layer.
Are we better off using a traditional non-nanite, LOD-enabled setup?
Hopefully, this is enough information for you to somewhat determine what is going on but I am more than happy to provide more information.
Thanks a lot