ROCm / ROCm

AMD ROCm™ Software - GitHub Home

Home Page:https://rocm.docs.amd.com

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

[Feature]: RX 7600 XT support

jonasgf opened this issue · comments

Suggestion Description

I recently bought an RX 7600 XT for primarily GPGPU purposes. I did unfortunately however not check the compatibility list before purchase and was sad to learn that my graphics card is currently not working with ROCm.


I have the latest ROCm packages installed (6.0.0).

rocminfo gives me the following:

$ /opt/rocm/bin/rocminfo --support
ROCk module is loaded
hsa api call failure at: /usr/src/debug/rocminfo/rocminfo-rocm-6.0.0/rocminfo.cc:1219
Call returned HSA_STATUS_ERROR_OUT_OF_RESOURCES: The runtime failed to allocate the necessary resources. This error may also occur when the core runtime library needs to spawn threads or create internal OS-specific events.

rocm-smi lists the GPU as such:

0       [0x0518 : 0xc0]       N/A     N/A    N/A, N/A        None  None  0%   unknown  Unsupported    0%   0%    
        Navi 33 [Radeon RX 7                                                                                     

I first found the page listing supported GPUs on Linux. There is very few GPUs listed here.

The supported GPUs on Windows page does list quite a few more GPUs, also the ones that AFAIK work well on Linux. I do not however see the RX 7600 XT listed here at all, so this gives me some hope that the GPU might be supported at some point.

I hope you can consider adding support for the RX 7600 XT, since it seems like a fantastic choice for ML and GPGPU in this budget range.

Operating System

Arch Linux (Linux 6.7.4-arch1-1 x86_64)

GPU

AMD Radeon RX 7600 XT

ROCm Component

No response

Or am I mistaken, and this is supported as gfx1102? There might just be something wrong with my setup if others have gotten this to work. I am trying to get this to work on a headless server I just set up today.

FWIW, I get the following in the kernel log when attempting to run rocminfo:

amdgpu 0000:09:00.0: amdgpu: Timeout waiting for VM flush ACK!
amdgpu 0000:09:00.0: amdgpu: Timeout waiting for VM flush ACK!
[drm] PCIE GART of 512M enabled (table at 0x0000008000000000).
[drm] PSP is resuming...
[drm:psp_hw_start [amdgpu]] *ERROR* PSP load kdb failed!
[drm:psp_resume [amdgpu]] *ERROR* PSP resume failed
[drm:amdgpu_device_fw_loading [amdgpu]] *ERROR* resume of IP block <psp> failed -62
amdgpu 0000:09:00.0: amdgpu: amdgpu_device_ip_resume failed (-62).
[drm:mes_v11_0_submit_pkt_and_poll_completion.constprop.0 [amdgpu]] *ERROR* MES failed to response msg=14
[drm:amdgpu_mes_flush_shader_debugger [amdgpu]] *ERROR* failed to set_shader_debugger

So it seems like the problem was something with my setup. It works fine if I have a monitor attached to the graphics card.

I do however have to run my pytorch programs with HSA_OVERRIDE_GFX_VERSION=11.0.0.

And when it comes to the supported GPUs list, if the support is implied with the RX 7600 (non-XT), then my bad for misunderstanding. If there is a relevant difference then I think it should be added to the list if there is support for it.