The general advice to improve performance is to reduce the number of meshes in your scene. If you search the net, they will tell you that is better to have a single mesh with 10,000 triangles than 10,000 meshes with a single triangle, for what I read there’s good reason for that, however none of them makes reference to occlusion culling.
I’ve been experimenting with occlusion culling and as far as I can tell it works on GameObjects not on triangles, so if the camera sees at least one triangle of a mesh the whole mesh is drawn, which means, in a scene with few large meshes it’s very likely that all meshes in front of the camera will be drown, so there’s little or no point in enabling occlusion culling.
On the other hand, if your scene is made of lot’s of small meshes, chances are that most of the meshes on the back get occluded by the objects in front, and thus not being drawn, this I believe can lead to a performance gain, if am I right, on a huge scene it should be better to have lots of small meshes instead of few big ones.
With this in mind, is there a point in increasing the number of triangles per mesh in order to reduce the number of meshes when you have occlusion culling enabled?
I suspect there is a way to make an educated guess about how many triangles per mesh is best for your scene based on the size of the scene and the average distance between objects.