Today I did some testing on my card (GF7900GS) and found out that NPOT textures (TEXTURE_2D but not POT) are actually 1-2% faster then the ARB_texture_rectangle ones. I wouldn’t call that a performance hit It is aviable for cards like Geforce FX and up and also for ATi cards with recent drivers, but I am not sure which Radeons do offer support it.
Also, there is no real reason to avoid the 0…1 coords range, while it is true, tha only few cards support real NPOT textures.
I was advised to use GL_TEXTURE_RECTANGLE_ARB for avoiding the 0…1 texture co-ord range.
No, that was a piece of advice for avoiding it.
i would like to know if there are any performance hits with using this
Yes. Perhaps.
On cards that actually natively have the concept (ie, anything nVidia makes), then there can be a performance penalty. On cards that emulate the functionality (ie, by using a regular texture and modifying the shader to convert the texture coordinate range back to [0,1]), then the only penalty is in the conversion, which should be fairly negligable.
In either case, I would strongly suggest not bothering with texture rectangles simply because you don’t like using the [0,1] texture coordinate range.