Originally Posted by
Syx
Unfortunately, the VRAM for both cards isn't pooled like when extra RAM is added to mobo, the VRAM in an SLI configuration works in parallel. So the assets to be rendered are sent to the respective card, which then only utilizes the VRAM available on board to store data. The GPUs cannot access data stored in VRAM on a different physical card. It's a damn shame. The benefit is that less VRAM it utilized, which means more room in VRAM to store computational data for graphical effects like supersampling, ambient occlusion, tessellation and more buffered frames.