I’ve been using vbo’s to render meshes with jogl for awhile and I’ve noticed some odd behavior.
First off, this odd behavior isn’t so odd, and is quite nice: When using buffer objects for glDrawElements() calls, the graphics card (or at least ati x1300 and geforce 8600) do very efficient frustrum culling, tripling my framerates.
Secondly, on a windows computer running xp pro sp2 w/ a ati x1300, I had 2.5 million triangles at 25 fps using BufferARB calls, I switched to using the core Buffer calls and the fps dropped to 1. This is probably the oddest since they should be mapped to the same thing (unless jogl treats them differently before calling opengl stuff)
Thirdly, when using vbo’s for glDrawElements() I see no peformance gain compared to just sending the buffer downstream unless the drawn objects would be offscreen. Only then do the vbo’s perform some form of culling while the standard usage does nothing.
Someone please tell me that I’m not going crazy?