Closed qtdzz closed 1 month ago
I've diagnosed the issue a bit, and what seems to be the problem is that if you have a CPU with integrated graphics, that gets picked up by rocm-smi
. CPUs have an additional voltage probe for vddnb
, which is not in this enum, which then causes an access into a map to throw. I think the proper fix would be to extract the probe type from in#_label
.
@qtdzz Internal ticket has been created to fix this issue. Thanks!
Hi @Freakness109 , thanks for debugging this issue, I'm referencing your findings for #182.
A current limitation of ROCm is that you need to disable integrated graphics in your BIOS before running ROCm, as it ROCm will attempt to use (or like you've said in your case, read system information) from the IGP and cause misbehaviour. Can you try disabling your integrated graphics processor in BIOS and rerunning your test?
This is documented (https://rocm.docs.amd.com/projects/install-on-linux/en/latest/install/amdgpu-install.html) but I think the warning should be more visible. I'll work with internal docs teams to get this fixed.
Please reopen the ticket if your issue recurs. Thanks!
Problem Description
When I run
/opt/rocm/bin/rocm-smi -g
, it throws "Exception caught: map::at"The version is actually 7.0 which I can't choose from the form. The arch package is from this build https://gitlab.archlinux.org/archlinux/packaging/packages/rocm-smi-lib/-/commit/a6a96dc61bb09fdffc96a82b1f349162f8a66f74)
Please let me know if you need more information for debugging. Thanks!
Operating System
NAME="Manjaro Linux" KERNEL="Linux 6.6.26-1-MANJARO"
CPU
AMD Ryzen 7 7800X3D 8-Core Processor
GPU
AMD Radeon RX 7900 XTX
ROCm Version
ROCm 6.1.0
ROCm Component
rocm_smi_lib
Steps to Reproduce
No response
(Optional for Linux users) Output of /opt/rocm/bin/rocminfo --support
Additional Information
No response