I was making some performance experiment and count actual triangels in the scene. Exactly same view for the both camera setups. One camera setup with no LOD and 3 cameras with LOD (I am using camera LODing method - first camera 0,3-45, second 45-200, third 200-1500 - so nothing is rendered twice). But that LODing is not important in this case. I turned all layers OFF and then started to turn them on, results (no occlussion):
- LOD setup (3 cameras):
347k triangels with all 3 LOD layers on (all meshes)
580k triangles with default layer (terrain)
4,5M triangles with water layer (there is very small piece of ocean - water3 + small lake in the view - pro water )
13,5M triangles with shadows on - directional light (3x more)
- Nod LOD and 1 camera
1,3M with objects (LOD works 347k vs 1,3m)
1,5M with default (+cca 200k polygons for this layer is same as before)
2,8M with water (3x less then before)
6,5M with shadows (cca 2x more)
- using LODs with this setup would me get on cca 4M
Conclusion:
Water layer is rendered as many times, as much cameras do you have, even if there are non redundant clipping planes. Shadows are dependand too, it increases triangels dramatically with number of cameras.
Is there any reason for this? Bug? Feaure? Am I doing something wrong?