GPU Passthrough to Multiple VMs: Advanced Techniques for Proxmox Enthusiasts • Self-Host Nerd

Introduction

Virtualization enthusiasts and Proxmox users often seek to maximize their hardware’s capabilities by leveraging advanced techniques such as GPU passthrough. This guide will walk you through the process of setting up GPU passthrough to multiple VMs using an Intel Arc A770 on Proxmox VE 8.2.2. We’ll cover the necessary configurations, potential challenges, and advanced techniques to achieve seamless GPU sharing between a Jellyfin VM and an oLlama VM.

Proxmox VE is a powerful open-source virtualization platform that provides enterprise-class features. However, GPU passthrough to multiple VMs can be particularly challenging due to hardware and software limitations. This guide aims to provide a step-by-step solution, ensuring both beginners and advanced users can follow along.

Installation Instructions

Prerequisites

Proxmox VE 8.2.2 installed on your server
An additional server to join the Proxmox cluster
Intel Arc A770 GPU installed and recognized by the system
Basic knowledge of Proxmox VE and Linux command line

Step 1: Preparing Your Proxmox Environment

Before we configure GPU passthrough, ensure your Proxmox environment is updated and your new server is added to the Proxmox cluster.

Update your Proxmox VE packages:
```
apt update && apt full-upgrade -y
```
Add the new server to the Proxmox cluster:
1. On the new server, install Proxmox VE if not already installed.
2. Join the cluster from the new server:
```
pvecm add <IP_of_existing_cluster_node>
```

Step 2: Configuring GPU Passthrough

Enable IOMMU

Ensure IOMMU is enabled in your BIOS/UEFI settings. This setting might be listed under advanced, CPU configuration, or similar sections.

Modify GRUB Configuration

Edit the GRUB configuration file:
```
nano /etc/default/grub
```
Find the line starting with GRUB_CMDLINE_LINUX_DEFAULT and add the following parameters:
```
GRUB_CMDLINE_LINUX_DEFAULT="quiet intel_iommu=on iommu=pt"
```
Update GRUB:
```
update-grub
```
Reboot the server:
```
reboot
```

Verify IOMMU Groups

After rebooting, verify the IOMMU groups to ensure the GPU is properly isolated:

find /sys/kernel/iommu_groups/ -type l

Look for the IOMMU group associated with your Intel Arc A770.

Bind GPU to VFIO Driver

Identify the GPU and its associated devices:

lspci -nn | grep -i vga

lspci -nn | grep -i audio

Note the PCI IDs (e.g., 0000:01:00.0 and 0000:01:00.1).
Create a VFIO configuration file:
```
nano /etc/modprobe.d/vfio.conf
```
Add the following lines, replacing with your PCI IDs:
```
options vfio-pci ids=8086:56a0,8086:56a1
```
Update initramfs:
```
update-initramfs -u
```
Reboot the server again:
```
reboot
```

Step 3: Configuring VMs for GPU Passthrough

Create the Jellyfin VM

Create a new VM from the Proxmox web interface.
Assign CPU, memory, and storage as per Jellyfin requirements.
In the Hardware tab, add the GPU:
1. Click “Add” -> “PCI Device”.
2. Select the GPU and check “All Functions” if applicable.
Install the operating system and Jellyfin on the VM.

Create the oLlama VM

Repeat the steps above to create a new VM for oLlama.
Assign the same GPU to this VM.
Install the operating system and oLlama on the VM.

Main Content Sections

Advanced GPU Sharing Techniques

Sharing a single GPU across multiple VMs can be complex due to hardware limitations. However, certain techniques can help:

MDEV (Mediated Device): Some GPUs support mediated device capabilities, allowing them to be shared across VMs. Unfortunately, as of now, Intel Arc A770 does not support MDEV.
Software-Based Solutions: Consider using software-based transcoding for Jellyfin or leveraging CPU resources for oLlama if GPU sharing is not feasible.

Troubleshooting Common Issues

Here are some common issues you might encounter and their solutions:

VM Fails to Start: Ensure the GPU is not being used by the host. Check for conflicts with other hardware devices.
GPU Not Detected by VM: Verify that the GPU is bound to the VFIO driver and not in use by the host OS.
Performance Issues: Ensure that the VM has sufficient CPU and memory resources. GPU passthrough can be resource-intensive.

Practical Examples or Case Studies

Case Study: Jellyfin and oLlama on a Single GPU

In this case study, we successfully configured GPU passthrough for both Jellyfin and oLlama VMs on a single Intel Arc A770:

Configured Jellyfin VM with GPU passthrough for hardware-accelerated transcoding.
Configured oLlama VM with GPU passthrough for efficient processing.
Achieved stable performance by balancing CPU and memory resources between VMs.

Tips, Warnings, and Best Practices

Security: Ensure proper network segmentation and firewall rules to protect your VMs.
Backup: Regularly backup your VMs to avoid data loss in case of hardware failure.
Monitoring: Use monitoring tools to keep track of GPU usage and performance metrics.
Update Firmware: Keep your server’s BIOS/UEFI and GPU firmware updated for optimal compatibility and performance.

Conclusion

While GPU passthrough to multiple VMs on Proxmox can be challenging, it is achievable with the right configuration and techniques. By following this guide, you can set up a Proxmox environment that efficiently utilizes your Intel Arc A770 GPU across Jellyfin and oLlama VMs. Experiment with advanced configurations and share your experiences to help the community grow.

Additional Resources

Proxmox PCI Passthrough Documentation: Official Proxmox documentation for PCI passthrough.
Jellyfin Hardware Acceleration: Guide to enabling hardware acceleration in Jellyfin.
Intel Device Plugins for Kubernetes: Useful for understanding Intel GPU capabilities.

Frequently Asked Questions (FAQs)

Can I share a single GPU between multiple VMs in Proxmox? While it is technically possible, it is not officially supported for all GPUs and may require advanced configurations.
Do all GPUs support passthrough? No, GPU passthrough support depends on the GPU model and its drivers.
What are the alternatives to GPU passthrough? Software-based transcoding and leveraging CPU resources are viable alternatives.

Troubleshooting Guide

VM Fails to Start: Check the Proxmox logs for error messages and ensure no hardware conflicts.
GPU Not Detected: Verify that the GPU is properly bound to the VFIO driver and not in use by the host.
Performance Issues: Monitor resource usage and optimize the VM configurations for better performance.

By following this comprehensive guide, you should be well-equipped to configure GPU passthrough to multiple VMs on Proxmox, leveraging the power of your Intel Arc A770 GPU effectively. Happy virtualizing!