I'm developing a flight simulator, and i'd like to have a very detailed terrain. My idea is to use 256 separate textures of 1024x1024 each (derived from a 16384x16384 texture). Before trying this i'd like some advices relative to performances and how the video card works. 1) Those 256 textures are loaded at startup so they are loaded in system memory right ? 2) Since i'm using quadtree culling, i'd bind only the textures which are in the frustum according to current camera position. So binding a texture means transferring it from system memory to video card memory ? 3) What happens if the amount of binded textures exceeds tha vga memory ? Thanks

1, if you upload your textures to opengl(glteximage2d) your texture can be placed in sys mem agp mem an vram and what not. the driver decides.

2. Best idea is to only use what is visible, but binding dosn't mean that its copying it from sys to vram. its just telling opengl you want to use the texture, and the driver will do what it thinks is best. probobly try to fix as much as it can in vram, for best proformance, as its drawn.

3.if you have more textures(in size) being drawn then you have vram for, probobly will have huge fps drops as it copies the data around trying to make room, out of mem errors, bandwidth bottle necks.

but most of whats going on is determined by the driver, some times you can nudge it in the right direction, like drawing with a texture durring loading will probobly upload it to vram.

If anything is wrong, I appreciate correction.