x86 GPU Device Passthrough with Qemu

Overview

  • GPU passthrough allows a virtual machine (VM) to use a dedicated GPU, enabling near-native performance for 3D workloads, media decoding/encoding and other GPU-intensive tasks.
  • QEMU combined with libvirt is a common stack used for this on Linux.

1. Requirements

Hardware

  • CPU with IOMMU support:
    • Intel VT-d (Intel IOMMU)
    • AMD-Vi (AMD)

2. Enable IOMMU in the Bootloader

  • Make sure following parameters added to kernel command line.

    # For Intel CPUs
"intel_iommu=on"

# For AMD CPUs
"amd_iommu=on"

3. Identify the GPU

  • List devices

    lspci -nn

4. Isolate GPU with VFIO

  • Blacklist the GPU driver on the host system so it won't be used. Example 
    # For Intel GPUs
"module_blacklist=i915,xe"
  • Bind Devices to VFIO 
    options vfio-pci.ids=<VendorId>:<DeviceId>

5. Verify GPU Passthrough Works in QEMU

On Graphics VM (gui-vm) run the following commands.

  • To verifiy PCI device passthrough:

    lspci | grep VGA

  • Check the driver:

    $ lshw -c display

Limitation

With GPU passthrough to a specific VM, the GPU cannot be used for other VMs. However, sometimes we may need the GPU to be shared across different VMs, which is not possible with the current implementation.