Proxmox GPU passthrough for an Ollama local LLM server.
Operational checklist and command guide for binding an NVIDIA GeForce RTX 5060 Ti to vfio-pci, passing it into an Ubuntu Ollama VM, and exposing Ollama to a Hermes/Godzilla VM over LAN/API.
Step 1
Confirm GPU exists
Verify Proxmox sees the RTX 5060 Ti GPU and audio function.
lspci | grep -i -E "nvidia|vga|3d"01:00.0 VGA compatible controller: NVIDIA Corporation GB206 [GeForce RTX 5060 Ti] (rev a1)
01:00.1 Audio device: NVIDIA Corporation GB206 High Definition Audio Controller (rev a1)Step 2
Confirm IOMMU
Verify Intel VT-d / IOMMU is active before passthrough work.
dmesg | grep -e DMAR -e IOMMU -e AMD-ViDMAR: IOMMU enabled
DMAR: Intel(R) Virtualization Technology for Directed I/OStep 3
Check driver state
Confirm current kernel driver and module state.
lspci -nnk -s 01:00.0
lspci -nnk -s 01:00.1
lsmod | grep -E "vfio|nouveau|nvidia" || trueKernel driver in use: vfio-pciStep 4
Host driver rule
Do not install NVIDIA drivers on the Proxmox host.
apt install nvidia-driver nvidia-smiStep 5
Load VFIO modules
Load vfio, vfio_iommu_type1, and vfio_pci at boot.
cat > /etc/modules-load.d/vfio.conf <<'EOF'
vfio
vfio_iommu_type1
vfio_pci
EOFcat /etc/modules-load.d/vfio.confStep 6
Bind GPU to VFIO
Bind PCI IDs 10de:2d04 and 10de:22eb to vfio-pci.
cat > /etc/modprobe.d/vfio.conf <<'EOF'
options vfio-pci ids=10de:2d04,10de:22eb disable_vga=1
EOFcat /etc/modprobe.d/vfio.confStep 7
Blacklist host GPU drivers
Prevent nouveau/nvidia from grabbing the GPU.
cat > /etc/modprobe.d/blacklist-gpu.conf <<'EOF'
blacklist nouveau
blacklist nvidia
blacklist nvidiafb
blacklist nvidia_drm
blacklist nvidia_modeset
blacklist nvidia_uvm
options nouveau modeset=0
EOFcat /etc/modprobe.d/blacklist-gpu.confStep 8
Rebuild and reboot
Rebuild initramfs and reboot into VFIO state.
update-initramfs -u -k all
rebootStep 9
Verify passthrough readiness
Confirm both GPU functions use vfio-pci.
echo "=== GPU ==="
lspci -nnk -s 01:00.0
echo "=== GPU AUDIO ==="
lspci -nnk -s 01:00.1
echo "=== MODULES ==="
lsmod | grep -E "vfio|nouveau|nvidia" || true01:00.0 GPU Kernel driver in use: vfio-pci
01:00.1 GPU Audio Kernel driver in use: vfio-pci
nouveau is not loaded
vfio modules are loadedStep 10
Create Ollama VM
Use q35, OVMF/UEFI, CPU host, VirtIO, 32GB+ RAM.
OS: Ubuntu Server 24.04 LTS
Machine: q35
BIOS: OVMF / UEFI
CPU Type: host
Disk Controller: VirtIO SCSI
Network: VirtIO
RAM: 32GB+ recommended
CPU Cores: 8+ recommendedStep 11
Add PCI device
Add RTX 5060 Ti with All Functions, PCI-Express, ROM-Bar, Primary GPU.
Ollama VM → Hardware → Add → PCI DeviceAll Functions: checked
PCI-Express: checked
ROM-Bar: checked
Primary GPU: checkedStep 12
Install NVIDIA in VM
Install NVIDIA driver inside the Ollama VM only.
sudo apt update
sudo apt install -y ubuntu-drivers-common
ubuntu-drivers devices
sudo ubuntu-drivers autoinstall
sudo rebootnvidia-smiStep 13
Install Ollama
Install Ollama, run a model, verify GPU usage.
curl -fsSL https://ollama.com/install.sh | shollama --version
ollama run llama3.2:3b
nvidia-smiStep 14
Expose Ollama API
Bind Ollama to 0.0.0.0:11434 and test from Hermes VM.
sudo systemctl edit ollama[Service]
Environment="OLLAMA_HOST=0.0.0.0:11434"sudo systemctl daemon-reload
sudo systemctl restart ollama
ss -tulpn | grep 11434
curl http://OLLAMA_VM_IP:11434/api/tagsStep 15
Troubleshooting
Fix nouveau binding, IOMMU, display loss, and host driver mistakes.
cat /etc/modprobe.d/blacklist-gpu.conf
cat /etc/modprobe.d/vfio.conf
cat /etc/modules-load.d/vfio.conf
update-initramfs -u -k all
rebootintel_iommu=on iommu=pt
amd_iommu=on iommu=ptStep 16
Final verified state
Confirm vfio-pci, IOMMU enabled, nouveau unloaded, vfio modules loaded.
01:00.0 RTX 5060 Ti GPU Kernel driver in use: vfio-pci
01:00.1 RTX 5060 Ti Audio Kernel driver in use: vfio-pci
IOMMU enabled
nouveau not loaded
vfio modules loadedNVIDIA GeForce RTX 5060 Ti
GPU PCI ID: 10de:2d04
Audio PCI ID: 10de:22eb
Proxmox kernel: 6.17.2-1-pveOperational warning
Host display and driver boundary
After the GPU is bound to VFIO, Proxmox may lose local monitor output if the RTX 5060 Ti was the only display adapter. Manage Proxmox through Web UI or SSH. Install NVIDIA drivers inside the Ollama VM only.