GPU passthrough issues

kramer

New Member
Sep 5, 2025
1
0
1
My PVE host has 2 NV cards and 1 onboard Intel. I've followed many guides on setting up my PVE host and I believe I've got those steps completed. However, after creating a VM and manually adding 1 of the NV devices to the VM, rebooting, lspci will not show a NV device.

Should I just undo everything and go the NVIDIA vGPU route? I don't have subscriptions to either. This is just my homelab. Showing all relevant info for help. TIA


Code:
root@proxmox:~# pvesh get /nodes/`hostname`/hardware/pci --pci-class-blacklist "" | grep GeForce
│ 0x030000 │ 0x1b82 │ 0000:01:00.0 │         11 │ 0x10de │ GP104 [GeForce GTX 1070 Ti]                       │      │ 0x6678           │                       │ 0x3842           │ eVga.com. Corp.            │ NVIDIA Corporation         │
│ 0x030000 │ 0x1c02 │ 0000:04:00.0 │         13 │ 0x10de │ GP106 [GeForce GTX 1060 3GB]                      │      │ 0x6160           │                       │ 0x3842           │ eVga.com. Corp.            │ NVIDIA Corporation         │

root@proxmox:~# cat /proc/cmdline
BOOT_IMAGE=/boot/vmlinuz-6.14.8-2-pve root=/dev/mapper/pve-root ro quiet intel_iommu=on iommu=pt

root@proxmox:~# cat /etc/modprobe.d/
blacklist.conf      kvm.conf            mdadm.conf          pve-blacklist.conf  vfio.conf           

root@proxmox:~# cat /etc/modprobe.d/vfio.conf
options vfio-pci ids=10de:1b82,10de:1c02 disable_vga=1
options vfio_iommu_type1 allow_unsafe_interrupts=1

root@proxmox:~# dmesg | grep 'remapping'
[    0.409889] DMAR-IR: Queued invalidation will be enabled to support x2apic and Intr-remapping.
[    0.411538] DMAR-IR: Enabled IRQ remapping in x2apic mode

root@proxmox:~# lsmod | grep -e nvidia -e nouveau -e vfio
vfio_pci               16384  0
vfio_pci_core          86016  1 vfio_pci
vfio_iommu_type1       49152  0
vfio                   65536  4 vfio_pci_core,vfio_iommu_type1,vfio_pci
iommufd               110592  1 vfio
irqbypass              12288  2 vfio_pci_core,kvm

root@proxmox:~# dmesg | grep -e DMAR -e IOMMU
[    0.010725] ACPI: DMAR 0x000000005B6A9000 000088 (v02 INTEL  EDK2     00000002      01000013)
[    0.010753] ACPI: Reserving DMAR table memory at [mem 0x5b6a9000-0x5b6a9087]
[    0.160343] DMAR: IOMMU enabled
[    0.409865] DMAR: Host address width 39
[    0.409866] DMAR: DRHD base: 0x000000fed90000 flags: 0x0
[    0.409876] DMAR: dmar0: reg_base_addr fed90000 ver 4:0 cap 1c0000c40660462 ecap 29a00f0505e
[    0.409878] DMAR: DRHD base: 0x000000fed91000 flags: 0x1
[    0.409883] DMAR: dmar1: reg_base_addr fed91000 ver 1:0 cap d2008c40660462 ecap f050da
[    0.409885] DMAR: RMRR base: 0x00000060000000 end: 0x000000707fffff
[    0.409887] DMAR-IR: IOAPIC id 2 under DRHD base  0xfed91000 IOMMU 1
[    0.409888] DMAR-IR: HPET id 0 under DRHD base 0xfed91000
[    0.409889] DMAR-IR: Queued invalidation will be enabled to support x2apic and Intr-remapping.
[    0.411538] DMAR-IR: Enabled IRQ remapping in x2apic mode
[    0.661556] pci 0000:00:02.0: DMAR: Skip IOMMU disabling for graphics
[    0.731427] DMAR: No ATSR found
[    0.731428] DMAR: No SATC found
[    0.731429] DMAR: dmar0: Using Queued invalidation
[    0.731433] DMAR: dmar1: Using Queued invalidation
[    0.732188] DMAR: Intel(R) Virtualization Technology for Directed I/O

root@proxmox:~# qm config 2000 | grep hostpci0
hostpci0: 0000:01:00,pcie=1


deploy@nv-1070:~$ lspci -vvv | grep -i nvidia
deploy@nv-1070:~$