Some infos:
Blitz can only "hide" things that are (A) behind the camera or (B) out of the camera range IF they are individual entities. This means, if you have a big level that is loaded as one mesh and has say 20k triangles, then Blitz will always render 20k Triangles, no matter if you only watch a single wall or have a very low camera range.
So it may be important to cut the level in pieces. (See ClusterizeMesh in the archives)
While the fragmentation of a big level mesh may feel a little tricky, additional meshes in the scene will be hidden automaticly and efficiently, as long as they don't cover wide areas. Example: trees: not every tree should be an entity, nor should you add all trees to a single mesh. The golden middle way would be to add groups of trees to a grid of meshes using a polycount that makes sense.
If it's an indoor level then it's even easier, make sure the various level places are connected by fairly long tunnel systems, so you'll never render more than one place.
There are more systems to cull things and "optimize" the speed. VIS-Zones is one. BSP VIS Occlusion and more. A lot of people tried many things and there's a good chance that if you implement one of these complicated culling techniques then it will be slower than the unculled counterpart at the end of the day, so KISS, keep it simple.
If you use fragmentation and a low camera range then the speed is ok. All you have to do is to design the level with a low camerarange in mind.
Blitz can only "hide" things that are (A) behind the camera or (B) out of the camera range IF they are individual entities. This means, if you have a big level that is loaded as one mesh and has say 20k triangles, then Blitz will always render 20k Triangles, no matter if you only watch a single wall or have a very low camera range.
So it may be important to cut the level in pieces. (See ClusterizeMesh in the archives)
While the fragmentation of a big level mesh may feel a little tricky, additional meshes in the scene will be hidden automaticly and efficiently, as long as they don't cover wide areas. Example: trees: not every tree should be an entity, nor should you add all trees to a single mesh. The golden middle way would be to add groups of trees to a grid of meshes using a polycount that makes sense.
If it's an indoor level then it's even easier, make sure the various level places are connected by fairly long tunnel systems, so you'll never render more than one place.
There are more systems to cull things and "optimize" the speed. VIS-Zones is one. BSP VIS Occlusion and more. A lot of people tried many things and there's a good chance that if you implement one of these complicated culling techniques then it will be slower than the unculled counterpart at the end of the day, so KISS, keep it simple.
If you use fragmentation and a low camera range then the speed is ok. All you have to do is to design the level with a low camerarange in mind.