There are several ways to code for multiple GPUs, one is to rely on nVidia drivers (this was the only real option for DX11 platforms).  However, DX12 provides more alternatives to manage multiple GPUs (and their VRAM for that matter) and be able to leverage those GPUs in a far more efficient methods than DX11.  HOWEVER, that does require a significant change in how rendering is performed (lots of potential for breaking change, even more issues than we currently have).  I've posted this article m