How are textures dealt with internally?

Apparently Max adjusts texture sizes to the nearest power of 2 texture - like 128 X 128.

How does that effect LoadAnimImage with say two images within one 128 X 128 graphic? Are they treated as 64 X 128? Or is each image also sized-up - 128 X 128 X 2?

Because generally it's common practice to use a large graphic with multiple frames of animation and odd sized textues can be slower - or when they're sized-up as mentioned.

Basically I'm looking to optimize, and knowing how Max handles texture mapped quads (which is what they are internally I guess) would help.

On most hardware and under OpenGL1.2 without extra extensions, the textures in video memory have to be in powers of 2, starting at 64x64. It won't allow you to create a 64x128 texture, for example. Rectangular texture support was added later.

So if you load up an image as a 8x8 image it will sit on its own 64x64 texture, wasting lots of space. I don't know whether LoadAnimImage lets you grab stuff at non-power-of-two sizes, but if it does, it would grab them into power-of-two textures. So yes probably it would waste videoram.

If you are willing to learn OpenGL yourself you could store your full image containing all the anim frames as a single texture and then set up your own quads with texture mapping, adjusting the texture coordinates for each so that it shows only the frame you want.

Yes, and that's completely legitimate as it's not an odd texture size (64 and 128 are both powers of 2).

But that then runs into the other problem of using large textures which some cards deal with poorly - we can't win can we? ;-)

Someone posted that the Geforce2 slows down significantly using textures over 256 X 256 - any thoughts on optimum texture size?