Is it quicker to sample a 64*64 texture than a 128*128 texture (arbitary numbers) of the same colour depth and both on the graphics memory? If there is a difference in time, why is that? I was thinking, if you have a model loaded and it is close to the camera, if you load the same model further away is there any point in loading a lower resolution texture? the way i see it, it should take the same amount of time to sample a low res texture as it does for a high res texture. you still have to fetch texture data per pixel for a 2*2 texure? or does it optimize? how is texture data managed on the graphics card? cheers

Take a look at MIP mapping. It automatically uses smaller versions of the texture for far away models.

Smaller textures are more efficient because they allow better use of the texture cache, which is pretty small (just a few K).