I have an application that requires a GPU card to do some statistical calculations. Instead of buying a dedicated server to host this application I'd like to host on an existing VMWare infrastructure. To do this I understand that I'd need to use pcipassthrough to give the guest OS (CentOS 6.6, in this case) access to the GPU. This card would be used solely for it's computational capabilities through the NVIDIA CUDA toolkit, no graphics will be required. Because of the reliance on CUDA I have to use an NVIDIA GPU. After searching these and other forums it looks like NVIDIA Quadro 4000 will fit the bill for my purposes, but I wanted to ask: Has anyone successfully used a Quadro 4000 or other NVIDIA GPU (I'm not wedded to the Quadro 4000, just anything that will work) with pcipassthrough as a computation engine?
