Oculus Quest performance optimisation

Unity Engine
, , ,
1

Hi,
A couple of questions about building for Oculus Quest (v2 but also trying to remain compatible with Quest 1 if possible):

  1. We’re looking at MeshBaker and MeshCombine2 to batch meshes/textures. It says to combine “all” meshes in the scene that use the same material but say there are 2 combinable meshes, one visible and the other not visible, do we still combine these? i.e. if they are combined, does the 2nd mesh get processed (when we could normally use occlusion culling and just forget about it)?
  2. Given #1 do we still use any static/dynamic batching feature in Unity or just turn off? Any benefit?
  3. Any definitive info on how to reduce mesh complexity and draw calls? We read a LOT but it’s confusing. Blender, “Stats” window and “PolyCount” asset give very different numbers (I know a “vert” in Blender is different to one in Unity but our meshes/mats/UVs are all quite simple). We are baking everything (no realtime lighting) so we don’t have that “lighting/shadows doubles the vertex count” issue. But if we want to hit 100K tris, which number do we trust? Even as a rough guide to work towards.

We’d love to see an end-to-end workflow for this kind of optimisation - any good videos?

Thanks!

2

There’s plenty of things to look out for when optimizing for mobile.

is a good start.

3

I watched that a few days ago - very good info.

As I say though, not too much info on my questions above. I suspect that we should bake groups of objects together so that they can be culled/occluded while still reducing draw calls where possible.

4

I am not answering directly to your question, but yet what helped me optimize my Diploma thesis which was on Oculus Rift VR with each scene >~ 10 mil triangles was “addressables”.

Will take some time to properly set up, but you’ll see a great difference in performance.