i’m not sure if this make any sense to mix LOD for distant trees or objects.
I have just found out that the current mental-ray’s have a huge problem with the render proxy’s (Assembly) in certain conditions.
I posted before what I was trying to do, and it backfired on me. But with some help from bart from LAmrUG in the mental images forum, the problem was tracked down and is caused by the renderer to use only 1 cpu/core in certain conditions.
In my case: using many unique smaller assembly proxy’s. With current versions 3.7 and 3.8, it’s advised to rather use the mip_binaryproxy nodes.
In my situation, the triangle count for 1 of 4 groups reached something of 132 milion triangles. And yes, at one point i managed to get these all in maya with my 8800gt.
But the unique triangle tessilation was very low, the whole thing was made up of around 70 unique assembly’s.
The assembly’s were used to keep all the detail and complexity outside of maya and the viewport. And it worked wonders.
The point is: i can get so much detail to render (using the assembly’s will only be usable from 3.9 on) with such speed that it doesn’t matter to use different detail models. The high detail ones work as well for distant things and render very fast. So i don’t have to work extra time for lower-res versions.
Only problem that remains with the assembly’s is: how does one work with these and be able to include/exclude from lights? It doesn’t work for me. They get lit my any light even if they exclude all.