AMD Mi100 Not Being Recognized

I recently purchased a used Mi100 for my 8 GPU GIGABYTE G291-Z20 server. I’ve been able to use my Mi25s and Mi60 with ROCm just fine. But my Mi100 is not recognized in either of my gigabyte or Dell servers. I will mention these servers are PCIe-3, but from what I understand, cards that support PCI-e 4 should work in PCIe3 systems.

My gigabyte server technically isn’t on AMD’s list for officially supported hardware, but it seems quite similar to on of Gigabyte’s servers that is on the list, namely the T181-Z70 Server, which supports both Naples and Rome Epyc processors(my server has a Naples processor).

Maybe my Mi100 is just a dud as it doesn’t show up in lspci whilst the other cards do.

Bios latest? Csm disabled, above 4g decoding enabled, rebar enabled

4g enabled, I haven’t been able to figure out how to update the BIOS on my gigabyte machine - have been trying to do that today. The BIOS is American megatrends, and it seems that the option for CSM only shows up into the 2021 and not 2020 BIOS.

The instructions from American Megatrend’s website seem to indicate you need windows to run the BIOS update , so perhaps I will just have to delete Ubuntu and install Windows temporarily.

I didn’t see an option for rebar.

You likely need to use the ipmi to update the bios

Just updated the BIOS with the BMC portal. Still no option for CSM. The latest BIOS is 2020, and I see an option to not use legacy mode for PCI-e, so I’ve disabled that. Also, only have UEFI boot enabled.

Could just be the card is a dud. Replacement arrives next week so we’ll see.

Also will mention the card never gets warm after booting.

Uhhh, did you use eps connectors or pci-e power plugs

This topic was automatically closed 273 days after the last reply. New replies are no longer allowed.