Skip to content

Support us

Created: 2021-08-27 | Last update: 2022-07-17

Proxmox GPU Passthrough to VM

Introduction

GPU passthrough is a technology that allows the Linux kernel to present the internal PCI GPU directly to the virtual machine. The device behaves as if it were powered directly by the virtual machine, and the virtual machine detects the PCI device as if it were physically connected. We will cover how to enable GPU passthrough to a virtual machine in Proxmox VE.

Your mileage may vary depending on your hardware.

Proxmox Configuration for GPU Passthrough

The following examples uses SSH connection to the Proxmox server. The editor is nano but feel free to use any other editor. We will be editing the grub configuration file.

Find the PCI address of the GPU Device. The following command will show the PCI address of the GPU devices in Proxmox server:

lspci -nnv | grep VGA

Find the GPU you want to passthrough in result ts should be similar to this:

01:00.0 VGA compatible controller [0300]: NVIDIA Corporation TU104 [GeForce RTX 2080 SUPER] [10de:1e81] (rev a1) (prog-if 00 [VGA controller])

What we are looking is the PCI address of the GPU device. In this case it's 01:00.0.
01:00.0 is only a part of of a group of PCI devices on the GPU.
We can list all the devices in the group 01:00 by using the following command:

lspci -s 01:00

The usual output will include VGA Device and Audio Device. In my case, we have a USB Controller and a Serial bus controller:

01:00.0 VGA compatible controller: NVIDIA Corporation TU104 [GeForce RTX 2080 SUPER] (rev a1)
01:00.1 Audio device: NVIDIA Corporation TU104 HD Audio Controller (rev a1)
01:00.2 USB controller: NVIDIA Corporation TU104 USB 3.1 Host Controller (rev a1)
01:00.3 Serial bus controller [0c80]: NVIDIA Corporation TU104 USB Type-C UCSI Controller (rev a1)

Now we need to get the id's of those devices. We can do this by using the following command:

lspci -s 01:00 -n

The output should look similar to this:

01:00.0 0300: 10de:1e81 (rev a1)
01:00.1 0403: 10de:10f8 (rev a1)
01:00.2 0c03: 10de:1ad8 (rev a1)
01:00.3 0c80: 10de:1ad9 (rev a1)

What we are looking are the pairs, we will use those id to split the PCI Group to separate devices.

10de:1e81,10de:10f8,10de:1ad8,10de:1ad9

Now it's time to edit the grub configuration file.

nano /etc/default/grub

Find the line that starts with GRUB_CMDLINE_LINUX_DEFAULT by default they should look like this:

GRUB_CMDLINE_LINUX_DEFAULT="quiet"
intel_iommu=on
amd_iommu=on

Then change it to look like this (Intel CPU example) and replace vfio-pci.ids= with the ids for the GPU you want to passthrough:

GRUB_CMDLINE_LINUX_DEFAULT="quiet intel_iommu=on pcie_acs_override=downstream,multifunction video=efifb:off video=vesa:off vfio-pci.ids=10de:1e81,10de:10f8,10de:1ad8,10de:1ad9 vfio_iommu_type1.allow_unsafe_interrupts=1 kvm.ignore_msrs=1 modprobe.blacklist=radeon,nouveau,nvidia,nvidiafb,nvidia-gpu"

Save the config changed and then update GRUB.

update-grub

Next we need to add vfio modules to allow PCI passthrough.

Edit the /etc/modules file.

nano /etc/modules

Add the following line to the end of the file:

# Modules required for PCI passthrough
vfio
vfio_iommu_type1
vfio_pci
vfio_virqfd

Save and exit the editor.

Update configuration changes made in your /etc filesystem

update-initramfs -u -k all

Reboot Proxmox to apply the changes

Verify that IOMMU is enabled

dmesg | grep -e DMAR -e IOMMU

There should be a line that looks like DMAR: IOMMU enabled. If there is no output, something is wrong.

[0.000000] Warning: PCIe ACS overrides enabled; This may allow non-IOMMU protected peer-to-peer DMA
[0.067203] DMAR: IOMMU enabled
[2.573920] pci 0000:00:00.2: AMD-Vi: IOMMU performance counters supported
[2.580393] pci 0000:00:00.2: AMD-Vi: Found IOMMU cap 0x40
[2.581776] perf/amd_iommu: Detected AMD IOMMU #0 (2 banks, 4 counters/bank).

Check that the GPU is in a separate IOMMU Group by using the following command:

#!/bin/bash
shopt -s nullglob
for g in $(find /sys/kernel/iommu_groups/* -maxdepth 0 -type d | sort -V); do
    echo "IOMMU Group ${g##*/}:"
    for d in $g/devices/*; do
        echo -e "\t$(lspci -nns ${d##*/})"
    done;
done;

Now your Proxmox host should be ready to GPU passthrough!

Windows Virtual Machine GPU Passthrough Configuration

For better results its recommend to use this Windwos 10/11 Virutal Machine configuration for proxmox.

Limitations & Workarounds

  • In order for the GPU to to function properly in the VM, you must disable Proxmox's Virutal Display - Set it none.

  • You will lose the ability to conect to the VM via Proxmox's Console.

  • Display must be conected to the physical output of the GPU for the Windows Host to initialize the GPU properly.

  • You can use a HDMI Dummy Plug as a workaround - It will present itself as a HDMI Display to the Windows Host.

  • Make sure you have alternative way to connect to the VM for example via Remote Desktop (RDP).

Find the PCI address of the GPU.

lspci -nnv | grep VGA

This should result in output similar to this:

01:00.0 VGA compatible controller [0300]: NVIDIA Corporation TU104 [GeForce RTX 2080 SUPER] [10de:1e81] (rev a1) (prog-if 00 [VGA controller])

If you have multiple VGA, look for the one that has the Intel in the name.
Here, the PCI address of the GPU is 01:00.0.

Proxmox lspci vga

For best performance the VM should be configured the Machine type to q35.
This will allow the VM to utilize PCI-Express passthrough.

Open the web gui and navigate to the Hardware tab of the VM you want to add a vGPU.
Click Add above the device list and then choose PCI Device

Windows VM Add PCI Device

Open the Device dropdown and select the GPU, which you can find using it’s PCI address. This list uses a different format for the PCI addresses id, 01:00.0 is listed as 0000:01:00.0.

Add GPU to VM

Select All Functions, ROM-Bar, Primary GPU, PCI-Express and then click Add.

Windows VM GPU PCI Settings

The Windows Virtual Machine Proxmox Setting should look like this:

Windows VM GPU Hardware Settings

Power on the Windows Virtual Machine.

Connect to the VM via Remote Desktop (RDP) or any other remote access protocol you prefer. Install the latest version of GPU Driver for your GPU.

If all when well you should see the following output in Device Manager and GPU-Z:

GPU-Z and Device Manager GPU

That's it!

Linux Virtual Machine GPU Passthrough Configuration

We will be using Ubuntu Server 20.04 LTS. for this guide.

From Proxmox Terminal find the PCI address of the GPU.

lspci -nnv | grep VGA

This should result in output similar to this:

01:00.0 VGA compatible controller [0300]: NVIDIA Corporation TU104 [GeForce RTX 2080 SUPER] [10de:1e81] (rev a1) (prog-if 00 [VGA controller])

If you have multiple VGA, look for the one that has the Intel in the name. Here, the PCI address of the GPU is 01:00.0.

lspci-nnv-vga

For best performance the VM should be configured the Machine type to q35.
This will allow the VM to utilize PCI-Express passthrough.

Ubuntu VM Add PCI Device

Open the Device dropdown and select the GPU, which you can find using it’s PCI address. This list uses a different format for the PCI addresses id, 01:00.0 is listed as 0000:01:00.0.

Add GPU to VM

Select All Functions, ROM-Bar, PCI-Epress and then click Add.

Ubuntu VM GPU PCI Settings

The Ubuntu Virtual Machine Proxmox Setting should look like this:

Ubuntu VM GPU Hardware Settings

Boot the VM. To test the GPU passthrough was successful, you can use the following command in the VM:

 sudo lspci -nnv | grep VGA

The output should incliude the GPU:

01:00.0 VGA compatible controller [0300]: NVIDIA Corporation TU104 [GeForce RTX 2080 SUPER] [10de:1e81] (rev a1) (prog-if 00 [VGA controller])

Now we need to install the GPU Driver. I'll be covering the installation of Nvidia Drivers in the next example.

Search for the latest Nvidia Driver for your GPU.

sudo apt search nvidia-driver

In the next step we will install the Nvidia Driver v510.

Note

--no-install-recommends is important for Headless Server. nvidia-driver-510 will install xorg (GUI) --no-install-recommends flag will prevent the GUI from being installed.

sudo apt install --no-install-recommends -y build-essential nvidia-driver-510 nvidia-headless-510 nvidia-utils-510 nvidia-cuda-toolkit

This will take a while to install. After the installation is complete, you should reboot the VM.

Now let's test the Driver initalization. Run the following command in the VM:

nvidia-smi && nvidia-smi -L

If all went well you should see the following output:

Ubuntu VM GPU Nvidia-smi

That's it! You should now be able to use the GPU for hardware acceleration inside the VM.

Debug

Dbug Messages - Shows Hardware initialization and errors

dmesg -w

Display PCI devices information

lspci

Display Driver in use for PCI devices

lspci -k

Display IOMMU Groups the PCI devices are assigned to

#!/bin/bash
shopt -s nullglob
for g in $(find /sys/kernel/iommu_groups/* -maxdepth 0 -type d | sort -V); do
    echo "IOMMU Group ${g##*/}:"
    for d in $g/devices/*; do
        echo -e "\t$(lspci -nns ${d##*/})"
    done;
done;

Reboot Proxmox to apply the changes

Comments