Graphics Processing Units (GPUs) on Lenovo® ThinkSystem™ servers are typically used to offload tasks from the server CPU, such as AI, VDI, and rendering tasks. Customers who use a Linux virtual environment on their ThinkSystem server may want to assign the GPU to a virtual machine (VM), and thus allow the GPU to appear as if it was physically attached to the guest OS running in the VM. This functionality is called GPU passthrough.
This paper provides guidance on enabling GPU passthrough to a VM running in a Kernel Virtual Machine (KVM)-based OS. The paper is for Linux administrators wishing to use a GPU in a ThinkSystem server and to pass through the device to a VM.
Table of Contents
Enabling IOMMU in UEFI
Enabling IOMMU host kernel support
Unbinding the GPU device from host physical machine driver
Getting the GPU IOMMU configuration
Attaching a GPU device with virsh
Installing and enabling the NVIDIA driver in the guest OS
To view the document, click the Download PDF button.