I have read some old threads but have not got a clear answer.
Anyone has the experience achieving GPU virtualization on a VMWare virtual machine and CUDA code can run on it? From nvidia’s description(Virtual GPU (vGPU) Monitoring and Management | NVIDIA) GPU virtualization seems only supported by Xenserver. I am wondering if it is also supported by VMWare? If yes, is it supported by VMWare workstation or only ESX/ESXi?
Answer to my own question: GPU virtualization is supported by VMWare. Need to install hypervisor on host machine that has graphics card. vDGA is achieved by setting device passthrough on the graphics card and assigning the card to individual virtual machine. Then CUDA code can run on the virtual machine.
I used tesla C1060 and K20c and both could be virtualized. Guest OS was Windows. The latter needs to download the latest windows driver (332.50) through.
Could you give a tutorial to do that? I have encountered the same problem, and I have two TITAN X cards. I would like to install a vmware virtual machine which can access the two cards. Many thanks!