T620 gpu issue

Question

Hi everyone

I have a weird issue with my T620:

For a couple of months now, it's been running with 2 Tesla K80 just fine (mostly machine learning stuff)

A couple of days ago, I got a Quadro RTX6000 and replaced one of the teslas and here's what's happening:

1st boot, new pcie found, reconfiguring and rebooting

on 2nd launch, it's telling me that: A bus fatal error was detected on a component at slot 7.

switched the card to another slot, same error after 2nd restart, just a different slot.

I also went in a little deeper and replaced the quadro with a 2070 super just to see if it's working, same fatal error even for the 2070.

Weird thing about the Teslas is that, it uses that weird cpu connector which splits into 2 pcie power connectors, that works fine. However, the other cards have a 8pin pcie and a 6 pin pcie, use less power than the teslas, and it doesn't like it for some reason.

it's running a xcp hypervisor, the teslas do show up with lspci | grep 3D, quadro or 2070 not at all

System Information
Description PowerEdge T620
BIOS Version 2.9.0

Node Id GC8CZ42

Host Name xcp1
OS Name XCP-ng
OS Version release 8.2.0 (xenenterprise) Kernel 4.19.0+1 (x86_64)
System Revision I
Lifecycle Controller Firmware 2.65.65.65
IDSDM Firmware Version N/A

Any thoughts what should I check/test?

DELL-Young E · Answer

Hi, thanks for choosing Dell.The GPU you wanted to use is a recent model but T620 isn't so we can't find it in the table. Please refer to page 20 https://dell.to/3pmT42w

hhoria · Answer

Hi, I'm more interested in actually what's happening. In table 7 on page 20, tesla K80 isn't listed either and it's been working just fine for almost 1 year now.

Also, yesterday night I tested it with a Vega64 and it worked just fine on 1st try so, ... maybe that table is incomplete and there are many more cards that should work but haven't been tested? Any logs I should check for a more specific issue description in idrac or somewhere?

Dell-Martin S · Answer

Hi, we did not have any further lists than this. All other GPUs are not tested and not supported.   Regards Martin

vorlket · Answer

Hi @hhoria, how did you get the bios detect K80? I'm using BIOS 2.9.0 and two cpus but the bios doesn't detect the K80 at all. Appreciate if you could please share info on this. Thanks.

ahakobyan · Answer

I have a T620 and I am trying to install 2 TESLA K80s, and I don't think I have the power connector for it. How did you solve this problem?

vorlket · Answer

@ahakobyan​ Tesla K80's didn't work on mine. You need GPU enablement kit.

PowerEdge Hardware General

T620 gpu issue

Was this post helpful?