> What does 'sudo lspci -v' print for the NPU on the host and in the guest?

 

 

Here's one of the host GPUs:

1b:00.0 3D controller: NVIDIA Corporation GH100 [H100 SXM5 80GB] (rev a1)
    DeviceName: GPU0
    Subsystem: NVIDIA Corporation GH100 [H100 SXM5 80GB]
    Flags: bus master, fast devsel, latency 0, IRQ 18, NUMA node 0, IOMMU group 67
    Memory at 21e042000000 (64-bit, prefetchable) [size=16M]
    Memory at 21a000000000 (64-bit, prefetchable) [size=128G]
    Memory at 21e040000000 (64-bit, prefetchable) [size=32M]
    Capabilities: <access denied>
    Kernel driver in use: vfio-pci
    Kernel modules: nvidiafb

 

And the same GPU as it appears on the guest:

06:00.0 3D controller: NVIDIA Corporation Device 2330 (rev a1)
    Subsystem: NVIDIA Corporation Device 16c1
    Physical Slot: 0-6
    Flags: bus master, fast devsel, latency 0, IRQ 10
    Memory at 382002000000 (64-bit, prefetchable) [size=16M]
    Memory at 380000000000 (64-bit, prefetchable) [size=128G]
    Memory at 382000000000 (64-bit, prefetchable) [size=32M]
    Capabilities: <access denied>
    Kernel modules: nvidiafb

 

Also, regarding the root cause of the slowdown, we're making some good progress in this thread. (it is increasingly seeming like more of a kernel issue than an OVMF issue).

 

- Mitchell Augustin

_._,_._,_

_._,_._,_