Anything above 32000 Tris per surface may load nicely, but as soon as Renderworld is excuted, it's MAVed. I've just tested a mesh with 1.60 Million tris, saved it as B3D (28 Megs, without UV Coords), and then loaded it again. It's pretty slow, but it works. You may use 64k Vertices, not sure of that, but since most meshes are welded and use shared vertices, the number of tris is about the same as the number of vertices. So I'd suggest to use a 32000 limit.
But it's funny, not so long ago computer architecture used to hit the 64k Ram segmentation border and they had to fix a lot of problems due to the lack of imagination in the times of the development of the 16 Bit adressing medthod. And right now the same happens with 3D stuff. We're doomed to work with segmented Maps... How does OGL handle this?
But it's funny, not so long ago computer architecture used to hit the 64k Ram segmentation border and they had to fix a lot of problems due to the lack of imagination in the times of the development of the 16 Bit adressing medthod. And right now the same happens with 3D stuff. We're doomed to work with segmented Maps... How does OGL handle this?