Arch Linux Setup with VFIO GPU Passthrough

Using AMD iGPU for Host + NVIDIA RTX 4090 for VM (VFIO)

linuxarchvfiohyprland

Archived setup note (October 2026): This is a machine-specific record from 2025. Package names, device IDs, boot tooling, and third-party installers can change. Back up your data, use the current upstream documentation, and review scripts locally before running them.

My notes and reference for using AMD iGPU for Host + NVIDIA RTX 4090 for VM (VFIO) on Arch Linux.

System Specs

  • Motherboard: Gigabyte B650 GAMING X AX (rev. 1.3)
  • CPU: AMD Ryzen 9 7950X3D (32 Threads)
  • GPU 1 (Host): AMD iGPU (Raphael, RDNA2)
  • GPU 2 (Passthrough/Optional Host): NVIDIA GeForce RTX 4090
  • Memory: Corsair Vengeance DDR5 64GB (2×32GB) 5600 MHz CL40 (XMP, CMK64GX5M2B5600C40)
  • OS: Arch Linux (Hyprland) + Windows 11 Pro (separate disk)

BIOS Configuration

Before installing Arch, reboot into BIOS and configure:

  • Enable: IOMMU (sometimes called SVM or AMD-Vi)
  • Enable: ErP (to reduce power when off)
  • Disable: Secure Boot
  • Disable: XMP (if facing instability)
  • Set iGPU as primary display output

Prepare Arch Installation

Boot from Arch ISO

Check your disks:

lsblk
ls -l /dev/disk/by-id/

Optional: Chris Titus Arch Install

Do not pipe a remote installer directly into a shell. Download a pinned release or clone the Chris Titus Tech utility, inspect it locally, and follow its current installation instructions.

Choose:

  • System Setup -> Arch Linux -> Arch Server Setup
  • Follow partitioning and installation prompts

Step by Step with archinstall

1. Disk Layout

  • Manual partitioning if Windows 11 is already installed.
  • Example Arch disk:
    • EFI: 512 MB, FAT32 → /boot
    • Root: rest (ext4 for stability or btrfs for snapshots)
    • Swap: optional (or use zram).

2. Bootloader

  • Choose systemd-boot.

3. Kernel

  • Install linux-zen (main) + linux as fallback.

4. Networking

  • Select NetworkManager.

5. Audio

  • Use PipeWire.

6. User Accounts

  • Set root password.
  • Create your main user and add to wheel.

7. Profile

  • Choose Minimal Environment.

8. Additional Packages

base-devel git curl wget nano htop openssh pipewire pipewire-alsa pipewire-pulse pipewire-jack wireplumber

9. Optional Repositories

  • Enable multilib.

First Boot Tasks

Example setup with Hyprland + ML4W dotfiles:

sudo pacman -S hyprland vim nvim kitty flatpak timeshift firefox

Managing systemd-boot

List all boot entries:

bootctl list

Create a Windows entry:

# /boot/loader/entries/windows.conf
title Windows 11
efi /EFI/Microsoft/Boot/bootmgfw.efi

Increase boot menu timeout (default is often too fast):

# /boot/loader/loader.conf
default arch-zen.conf
timeout 10

Remove an old/duplicate entry:

sudo bootctl unlink ENTRY_ID

Installing ML4W Dotfiles with Flatpak

As part of my Hyprland setup, I used the ML4W Dotfiles Installer to quickly bring in a pre-configured environment.
Below are the commands and the steps I followed.

flatpak install flathub com.ml4w.dotfilesinstaller

Installation Steps

  1. Open Firefox and visit:
    https://ml4w.com/os/

  2. Run the Flatpak installer:

    flatpak run com.ml4w.dotfilesinstaller

  3. If the website generates a shell script, save it to a local file instead of piping or pasting it directly into a shell. Pin the referenced release where possible and inspect the script and every download it invokes.

  4. Run the reviewed local copy only after you understand the changes it will make.

  5. Once the script finishes, return to the installer and click Next.

  6. Adjust your display resolution.

    In my case, I use 2K QHD (2560×1440).

  7. Complete the process by selecting Install and Activate.

  8. Close the installer and reboot your system.

Remove NVIDIA Host Drivers

Check installed NVIDIA packages:

pacman -Q | grep -i nvidia

Remove them only after confirming the NVIDIA GPU does not drive the host display. This can break the graphical boot, so keep a fallback kernel and console or live-media access ready:

sudo pacman -Rns nvidia-utils nvidia-lts linux-firmware-nvidia

VFIO Setup

Identify GPU IDs

lspci -nnk | grep -A3 NVIDIA

Blacklist NVIDIA

# /etc/modprobe.d/blacklist-nvidia.conf
blacklist nouveau
blacklist nvidia
blacklist nvidia_drm
blacklist nvidia_modeset

Bind to VFIO

Replace these placeholders with every function of the passthrough GPU. Never bind the GPU that drives the host display, and keep a fallback kernel or live environment available.

# /etc/modprobe.d/vfio.conf
options vfio-pci ids=VENDOR_ID:DEVICE_ID,AUDIO_VENDOR_ID:AUDIO_DEVICE_ID

Add vfio_pci vfio vfio_iommu_type1 to the existing MODULES=(...) array in /etc/mkinitcpio.conf; preserve any modules already listed. This loads VFIO early enough to claim the device before a host graphics driver.

Rebuild the initramfs, then verify the boot entries:

sudo mkinitcpio -P
bootctl list

Debugging Commands

lsmod | grep amdgpu
ls -l /dev/dri/
lspci -nnk | grep -A3 NVIDIA

Desktop Setup

For a managed setup, use the current upstream instructions for Arch-Hyprland or the ML4W dotfiles. Clone or download a pinned release and inspect the scripts locally rather than piping remote content into a shell.

Conclusion

With this setup:

  • Host uses the AMD iGPU (stable, efficient).
  • NVIDIA RTX 4090 is isolated for passthrough into a VM.
  • Dual boot with Windows is cleanly integrated via systemd-boot.