I tried very hard to start a card I purchased from online shop who said that they're AMD authorized dealer (after watching all shiny advertisements how it will run PyTorch flawlessly).
I installed AMDGpu installer on Linux Ubuntu 22.04.1 LTS, installed one of the predefined modes (DKMS got installed), but the amdgpu driver fails to initialize. When I uninstall it, the stock version of amdgpu driver tells me (in dmesg):
amdgpu 0000:03:00.0: amdgpu: Fatal error during GPU init
amdgpu: probe of 0000:03:00.0 failed with error -12
Does this mean that the card is broken ? I'm going to RMA the card.
lspci -vnn
tells me:
03:00.0 Display controller [0380]: Advanced Micro Devices, Inc. [AMD/ATI] Navi 21 [Radeon Pro V620] [1002:73a1]
Subsystem: Advanced Micro Devices, Inc. [AMD/ATI] Device [1002:0e34]
Flags: fast devsel, IRQ 30, IOMMU group 9
Memory at <unassigned> (64-bit, prefetchable) [disabled]
Memory at <unassigned> (64-bit, prefetchable) [disabled]
Capabilities: <access denied>
Kernel modules: amdgpu
I didn't give up, plugged the thing into an HP DL585 G7 server, running MS Server2016 and installed the Pro software from AMD. The software told me that there is an unsupported graphics hardware and bailed out, leaving me without options - neither Linux, nor Windows wanted to run the card.
Tried to assign it to a guest under virt-manager (Linux). Failed as well, despite I had everything set up - iommu in pt mode, vfio-pci bound to the card, card in its own mmu group.
