GPU Rendering

I had a few questions about GPU rendering. First, I was wondering what a float texture was. I know you can only have 5 but how are they different from other textures? Second, it said that each texture takes up more memory as it increases in size. So does that mean they all add up, or do they count individually? Like if you had two 200k textures and your GPU only had 200k of memory, would it add them up together or do each one on it’s own? I would think it adds them up but I just wanna be sure. Third, and this is gonna be more difficult to answer, does anyone know when subsurface scattering is gonna be available in GPU rendering? You don’t need to know the exact date, I was just wondering if it was gonna be months or years. Thanks.

I know that in Cycles the SSS works! at least I think?

Float textures are high bit depth textures… hdr textures… normal textures are made up of individual pixels, each pixel is made up of a channel R G and B (red green and blue)… each pixel is of a value between 0 and 255… the reason is because this fits nicely into 8bits worth of data… high bit depth textures, instead of having 8 bits allocated per channel, have 16bits, or 32 bits… which gives much higher precision…

the other thing about these high bit depth textures is that they can go outside the range of black to white, you have blacker then black and whiter then white, on a numeric scale this just means negative numbers and numbers greater then one.

It needs to store all textures at the same time when rendering… cpu or gpu. it doesnt matter… not just that, they need to be stored uncompressed… not compressed… so jpg images may only be a few kilobytes, but may be hundreds of mb on the gpu

Ok, that makes sense because I read that textures use more memory with GPU rendering. Is there any way to tell what the file size will be uncompressed? I’m asking all this because I wanted to get a new graphics card for my computer and I’m trying to figure out which kind to get.