Foliage clusters and lods

I’m struggling with that foliage/grass LOD behavior for a while and I’m also very interested in understanding how foliage clustering is working on 4.15 and newer.

It was working differently somehow on 4.14 and older versions. I presume they’ve changed the way the engine clusters instances together but now it’s so strange that it’s really hard to predict its behavior (as you’ve well described).

Because of that, migrating projects with lots of foliage instances from 4.14 to newer versions will cause performance loss. In my case I haven’t figured out a way to get the same performance back again even after hours of tweaking… In older versions the clustering process happened in a way that LOD screensize parameters had similar behavior on both single static meshes and foliage instances. It was probably clustering instances together in small packs based on distance and numbers.

Link for my unanswered question here: https://answers.unrealengine.com/questions/658509/inconsistent-lod-behavior-on-foliage-and-grass.html

Do we have any control over how the engine clusters foliage instances? How is that calculated?

The following [documentation][1] says…

Currently the entire cluster of instances change LODs simultaneously. We may add support for distance-based fading per instance in the future.

So I painted a handful of trees on my map. They are updated individually, NOT simultaneously. So either each individual one is it’s own cluster or the documentation is incorrect.

However, I also painted a few stumps on my map. The LODs for these ARE all updated simultaneously.

Why are they behaving differently? Are they batched based on tri count or what is going on?
Very little is exposed for foliage and there is a ton of mystery around it…

So I went ahead and painted a lot more stumps and now their LODs are updated at different times, still clustered though for sure. Is the engine perhaps calculating the cluster size based on the tri count of the mesh? And this is why my trees seem to be clustered in groups of 1. But these stumps are clustered in much larger groups?

Some insight on how these are calculated would be great.

Thank you!

Hey, have you found any answer to this yet? They’ve just released 4.17 and it keeps the same… Also no answers on any of those LOD related questions on answerhub…

Just to back you up, I am facing the same problem.