We're currently in a big project using heavy cg-shaders and are thinking about how different texture formats actually work. The GL_LUMINANCE_ALPHA is a two-component texture, right? But is it really more efficient to use that one, than just using two channels in a, for example, GL_RGBA texture? We're using 3D-textures in GL_RGBA16F_ARB (using 3 channels) and seem to get better result with that format than using GL_RGB16F_ARB. How do the GPU handle these different formats?