1|# 03 - AMD GPU Passthrough (Strix Halo) — Home
     2|**Purpose:** Detailed AMD APU passthrough configuration for host LLM access.
     3|
     4|## Overview
     5|
     6|This guide covers the complete setup for AMD Strix Halo GPU passthrough, enabling native LLM inference on AMD hardware.
     7|
     8|## What You'll Learn
     9|
    10|- **Hardware preparation:** Verify Strix Halo compatibility
    11|- **BIOS configuration:** Enable IOMMU and PCIe passthrough
    12|- **Kernel configuration:** Load VFIO modules properly
    13|- **Proxmox setup:** Configure GPU for VM passthrough
    14|- **Driver installation:** Vulkan, ROCm, HIP SDK integration
    15|- **Qwen3.5-MoE setup:** Configure for AMD GPU acceleration
    16|
    17|## Prerequisites
    18|
    19|- AMD Strix Halo APU with integrated Radeon graphics
    20|- Linux system kernel 5.4+ with IOMMU support
    21|- Proxmox VE 7.x or later
    22|- 16GB+ system RAM (8GB for APU graphics, 8GB for OS)
    23|- SSD storage for VM images
    24|
    25|## Quick Start
    26|
    27|```bash
    28|# Enable IOMMU in GRUB
    29|# Edit /etc/default/grub, add:
    30|# GRUB_CMDLINE_LINUX="amd_iommu=on iommu=pt"
    31|
    32|# Update GRUB
    33|update-grub
    34|
    35|# Load VFIO modules
    36|echo "vfio
    37|vfio_iommu_type1
    38|vfio_pci
    39|vfio_virqfd" | sudo tee /etc/modules-load.d/vfio.conf
    40|
    41|# Blacklist Radeon in VM
    42|modprobe -r radeon amdgpu
    43|
    44|# Reboot to apply changes
    45|reboot
    46|```
    47|
    48|## Related Guides
    49|
    50|- **[01 - Proxmox & VM Passthrough](../01\ -\ Proxmox\ \&\ VM\ Passthrough/)** — General passthrough setup
    51|- **[04 - Qwen3.5 Model Setup](../04\ -\ Qwen3.5\ Model\ Setup/)** — Model deployment
    52|
    53|---
    54|
    55|*Essential for running Qwen3.5-MoE locally on AMD hardware with Vulkan backend acceleration.*
    56|