Some performance questions about polycount

Hi, the objects only get culled as whole object. If you see even a small part of its bounding box, then the whole object is gonna be processed by the GPU and you gonna see the whole poly count in the stat view.

If you have 30-40k meshes then you need to use instancing (but keep in mind that then you won’t have dynamic occlusion culling for them, only distance and frustum culling. So if you got 10k instances behind a wall were you can’t see them, they will still all get processed by the GPU if you look at the wall). You can use the console command “FreezeRendering” to see what is being send to the GPU.

working on funfair stuff and for that
i need many many small lamp meshes
like arround 30-40k

Why not make some different meshes each with a couple of hundreds of lamps and use those? Cause that will give you better performance or less headaches and in the end it does not matter how you made it as long as it looks the same.