kvm GPU implementation passthrough (GPU passthrough)

In order to facilitate artificial intelligence to automatically manage server or achieve a particular function, we need to virtualize the graphics card. This hardware for graphics, the NVIDIA GPU virtualization support a dedicated graphics card, such as GRID GPU series. In an example NVIDIA GRID K2, memory 8GB, the two may be virtual GPU. However, this comes with the virtualization capabilities of the graphics card, the price is very expensive.

For the GTX series graphics card, if Host host linux system, like the graphics card to a virtual machine (windows or linux) use, you need to use the GPU pass-through (GPU passthrough) technology. Currently very little information in this regard, two weeks off to explore, GTX TITAN X graphics transparent transmission finally succeeded, on the map:
kvm GPU implementation passthrough (GPU passthrough)

Supernumerary:
Detailed steps can refer to: https://mianbaoduo.com/o/bread/Y5ibmZk =

Guess you like

Origin blog.51cto.com/6627243/2480955