From the 2021.2.0a15 changelog:
I always wondered, why do you support LOD 0, LOD 1 and LOD 2 only? Why not allowing us to provide more than 3 LOD’s?
In my experience, using more than 3 LODs would be of little benefit
This goes well against the idea of optimizing a game, after all, the more LODs you use, it means that you are including more and more models that will only be consuming more storage space and that will only be visible at certain distances…
In the worst case scenario, you only need a highpoly model when the camera is close to the object, a second model with a little less detail when it is a little more distant, maybe a low poly for long distances, or you can jump straight to an imposter which will save a lot of hardware resources.
More LODs are very useful for characters, where you can’t use impostors and want to go smoothly from high poly cinematic to a few hundred triangles without popping. Skinned models are model expensive due to the skinning, so any savings is helpful there. Look at any AAA game and see how many LODs they have for characters.
More LODs will eventually also cost GPU performance, because you’ll reduce how many meshes can be rendered together using GPU instancing.
Really? From what I can tell, 8 LOD levels are supported.
You’re right, it does work.