I've been trying to sort this out withouth asking but I can't. I'm playing around with raycasting on the GPU, currently on a GeForce GTX 285 with 1024 MB VRAM. I would very much like to be able to play around with a lot more data.

When trying to figure out which card to get, if any, I seem to come back to the GTX 680. One of them with 4096 MB of VRAM would probably make my raycasting skyrocket from already decent performance. I tried it out on a GTX 470 and that was very pleasant experience. But that card didn't have more VRAM, just more pure power.

My question: If I got a GTX 680 with 4096 MB VRAM, would I be able to access all of it from a single OpenGL context? I have a very clear memory of reading somewhere (somewhere on the big, big internet) that those cards have two GPUs and 2x2096 MB of VRAM.

Can anyone help me sort this one out? I have one little guy sitting on my left shoulder saying "Don't ask, of course you'll get full access to all that lovely VRAM; let's head over to the store in another browser tab and order one right now!" and on the other one there sits another guy saying "4096 MB of VRAM in a single context? That would be awesome. So much awesome, would that be available for such money?". Which one is right?