I read that you can increase performance by limiting calls to glBindTexture(GL_TEXTURE_2D, TextureID) which means a sorting of polys by textures… But if you are dealing with many thousands of polys to consider for the graphics pipeline you would also need some other type of occulsion testing (BSP ,quad tree ect…)…Is this true and with the new graphics cards is the performance hit with the state changes still a factor… Thanks…
Yes it’s true, and yes state changes is still a factor, but if varies from card to card.
Obviously with occlusion culling you are going to draw less and that will improve performance IF you don’t waste too much time performing the occlusion tests. This is changing though at least for fill performance, cards like the Radeon and GeForce3 have coarse Z buffer testing. This means that if you draw front to back you will not have the same pixel fil penalty for drawing occluded triangles. YOu will still incur the transform cost though.