github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-07-15 20:54:54 -05:00
Code Issues 16.9k Packages Projects Releases 122 Wiki Activity

[GH-ISSUE #11450] ROCm GPU not detected on OpenSUSE Leap 15.6 with Ryzen 7 5700G (Vega) #7560

New Issue
Open
opened 2026-04-12 19:39:30 -05:00 by GiteaMirror · 2 comments
GiteaMirror commented 2026-04-12 19:39:30 -05:00
Owner
Copy Link

Originally created by @ratermir on GitHub (Jul 16, 2025).
Original GitHub issue: https://github.com/ollama/ollama/issues/11450

What is the issue?

System Information

  • OS: OpenSuSE Leap 15.6
  • CPU/APU: AMD Ryzen 7 5700G (gfx90c)
  • ROCm Version: Installed via amdgpu-install script (ROCm 6.x stack)
  • Kernel: 6.4.0-150600.23.53-default

Problem Description

Ollama consistently fails to detect the integrated Vega GPU, falling back to CPU mode. The log shows amdgpu detected, but no compatible rocm library found.

This happens despite the host system being correctly configured.

Key Evidence

  • Host system's rocminfo shows the new kernel module is loaded correctly.

ricminfo.txt

  • Host system's rocm-smi works correctly.
    (Zde vložte výpis z /opt/rocm/bin/rocm-smi)

  • Crucially, other official AMD ROCm containers do work with GPU acceleration on this same host. This proves the host configuration is correct.

  • The error persists across all tested Ollama versions, including :latest and :0.1.32.

  • The error persists even when building a custom container with Ollama installed on top of a rocm/rocm-runtime base image.

Attempts to fix

We have exhaustively tried the following without success:

  • Reinstalling ROCm stack from official AMD repositories.
  • Ensuring dkms and kernel-devel packages are correct.
  • Verifying Secure Boot is disabled.
  • Manually cleaning up stray files from old installations.
  • Using the HSA_OVERRIDE_GFX_VERSION environment variable.
  • Downgrading the Ollama image to community-recommended versions.
  • Building a custom container on a known-good AMD base image.

The issue seems to be inherent to how the Ollama binary is built or how it discovers ROCm libraries in a SUSE environment.

Logs

Final Ollama Log:

ollama.log

rocm-smi output:

rocm-smi-out.txt

Relevant log output

OS

Linux

GPU

AMD

CPU

AMD

Ollama version

0.1.30, 0.1.32, latest (2025-07-15)

Originally created by @ratermir on GitHub (Jul 16, 2025). Original GitHub issue: https://github.com/ollama/ollama/issues/11450 ### What is the issue? ### System Information - **OS**: OpenSuSE Leap 15.6 - **CPU/APU**: AMD Ryzen 7 5700G (gfx90c) - **ROCm Version**: Installed via `amdgpu-install` script (ROCm 6.x stack) - **Kernel**: 6.4.0-150600.23.53-default ### Problem Description Ollama consistently fails to detect the integrated Vega GPU, falling back to CPU mode. The log shows `amdgpu detected, but no compatible rocm library found`. This happens despite the host system being correctly configured. ### Key Evidence - **Host system's `rocminfo` shows the new kernel module is loaded correctly.** [ricminfo.txt](https://github.com/user-attachments/files/21266911/ricminfo.txt) - **Host system's `rocm-smi` works correctly.** (Zde vložte výpis z `/opt/rocm/bin/rocm-smi`) - **Crucially, other official AMD ROCm containers *do* work with GPU acceleration on this same host.** This proves the host configuration is correct. - The error persists across all tested Ollama versions, including `:latest` and `:0.1.32`. - The error persists even when building a custom container with Ollama installed on top of a `rocm/rocm-runtime` base image. ### Attempts to fix We have exhaustively tried the following without success: - Reinstalling ROCm stack from official AMD repositories. - Ensuring `dkms` and `kernel-devel` packages are correct. - Verifying Secure Boot is disabled. - Manually cleaning up stray files from old installations. - Using the `HSA_OVERRIDE_GFX_VERSION` environment variable. - Downgrading the Ollama image to community-recommended versions. - Building a custom container on a known-good AMD base image. The issue seems to be inherent to how the Ollama binary is built or how it discovers ROCm libraries in a SUSE environment. ### Logs **Final Ollama Log:** [ollama.log](https://github.com/user-attachments/files/21266706/ollama.log) **rocm-smi output:** [rocm-smi-out.txt](https://github.com/user-attachments/files/21266918/rocm-smi-out.txt) ### Relevant log output ```shell ``` ### OS Linux ### GPU AMD ### CPU AMD ### Ollama version 0.1.30, 0.1.32, latest (2025-07-15)
GiteaMirror added the bug label 2026-04-12 19:39:30 -05:00
GiteaMirror commented 2026-04-12 19:39:31 -05:00
Author
Owner
Copy Link

@fighter3005 commented on GitHub (Jul 17, 2025):

Are you sure this is a bug, and not expected behavior? As far as I know, ollama (and rocm for that matter) do not support gfx90c officially. Even if you overwrite it, ollama will not detect the APU, since only the 8050S and 8060S are supported (to my knowledge). Why I am not sure. Might be memory (<=1024mb) or a dedicated check. Maybe I am mistaken.

you can try this though: https://github.com/rjmalagon/ollama-linux-amd-apu/issues/13

Also, do you have the GGT memory working (shared memory)? I believe you need kernel 6.10 or above, but don't quote me on that.

<!-- gh-comment-id:3085279999 --> @fighter3005 commented on GitHub (Jul 17, 2025): Are you sure this is a bug, and not expected behavior? As far as I know, ollama (and rocm for that matter) do not support gfx90c officially. Even if you overwrite it, ollama will not detect the APU, since only the 8050S and 8060S are supported (to my knowledge). Why I am not sure. Might be memory (<=1024mb) or a dedicated check. Maybe I am mistaken. you can try this though: https://github.com/rjmalagon/ollama-linux-amd-apu/issues/13 Also, do you have the GGT memory working (shared memory)? I believe you need kernel 6.10 or above, but don't quote me on that.
GiteaMirror commented 2026-04-12 19:39:31 -05:00
Author
Owner
Copy Link

@abcbarryn commented on GitHub (Jul 21, 2025):

Only 1Gb of VRAM? Yikes, that's a tiny GPU. In my case Ollama refused to use my Nvidia Tesla K40C because it only had a compute score of 3.5. I commented out the compute score check and it seems to be happily using my Nvidia Tesla K40C now.

# ollama ps
NAME          ID              SIZE     PROCESSOR          CONTEXT    UNTIL      
gemma3:12b    f4031aab637d    11 GB    23%/77% CPU/GPU    4096       2 minutes from now

Why was this check put in and the value for the minimum compute score set so high? Some of us don't have the resources of Elon Musk or Bill Gates and need to try to make do with used/second hand equipment. That said, I doubt a less than 1GB GPU is going to doing anything useful other than flop over and die in an entertaining way... My Nvidia Tesla K40C though, while being an older model GPU, still managed to boost the performance of my system substantially from what it was with CPU only.

<!-- gh-comment-id:3099620731 --> @abcbarryn commented on GitHub (Jul 21, 2025): Only 1Gb of VRAM? Yikes, that's a tiny GPU. In my case Ollama refused to use my Nvidia Tesla K40C because it only had a compute score of 3.5. I commented out the compute score check and it seems to be happily using my Nvidia Tesla K40C now. ``` # ollama ps NAME ID SIZE PROCESSOR CONTEXT UNTIL gemma3:12b f4031aab637d 11 GB 23%/77% CPU/GPU 4096 2 minutes from now ``` Why was this check put in and the value for the minimum compute score set so high? Some of us don't have the resources of Elon Musk or Bill Gates and need to try to make do with used/second hand equipment. That said, I doubt a less than 1GB GPU is going to doing anything useful other than flop over and die in an entertaining way... My Nvidia Tesla K40C though, while being an older model GPU, still managed to boost the performance of my system substantially from what it was with CPU only.
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
Clear labels
amd
api
app
bug
build
cli
cloud
compatibility
context-length
create
docker
documentation
embeddings
feature request
feedback wanted
good first issue
gpt-oss
gpu
harmony
help wanted
image
install
intel
js
launch
linux
macos
memory
mlx
model
needs more info
networking
nvidia
ollama.com
performance
pull-request
Mirrored from GitHub Pull Request
 python
question
registry
rendering
thinking
tools
top
vulkan
windows
wsl
No Label bug
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
GiteaMirror ninjasurge
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: github-starred/ollama#7560
Delete Branch "%!s()"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?