github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-07-16 13:19:40 -05:00
Code Issues 16.9k Packages Projects Releases 122 Wiki Activity

[GH-ISSUE #10691] ollama not using AMD GPU and using CPU instead #7028

New Issue
Closed
opened 2026-04-12 18:56:02 -05:00 by GiteaMirror · 5 comments
GiteaMirror commented 2026-04-12 18:56:02 -05:00
Owner
Copy Link

Originally created by @WhyNotHugo on GitHub (May 13, 2025).
Original GitHub issue: https://github.com/ollama/ollama/issues/10691

What is the issue?

On my host, ollama isn't using the gpu. It also logs that the version parameter is not exposed via sysfs:

level=WARN source=amd_linux.go:61 msg="ollama recommends running the https://www.amd.com/en/support/linux-drivers" error="amdgpu version file missing: /sys/module/amdgpu/version stat /sys/module/amdgpu/version: no such file or directory"
level=INFO source=amd_linux.go:386 msg="amdgpu is supported" gpu=GPU-8668c3be2237aa5f gpu_type=gfx1100
level=INFO source=types.go:130 msg="inference compute" id=GPU-8668c3be2237aa5f library=rocm variant="" compute=gfx1100 driver=0.0 name=1002:744c total="20.0 GiB" available="14.6 GiB"

The amdgpu module is loaded:

> lsmod | grep ^amdgpu 
amdgpu              11591680  66

It also exposes quite a few other parameters, but no version:

> ls -1F /sys/module/amdgpu/ 
coresize
drivers/
holders/
initsize
initstate
notes/
parameters/
refcnt
sections/
taint
uevent

When executing a model, I also see:

ggml_cuda_init: failed to initialize ROCm: no ROCm-capable device is detected

GPU usage remains at 0%, only CPU is used. I'm running ollama via docker:

docker run --rm --name ollama --device /dev/kfd --device /dev/dri -v ollama:/root/.ollama -p 11434:11434 ollama/ollama:rocm

Relevant log output

OS

Linux

GPU

AMD

CPU

Intel

Ollama version

ollama version is 0.6.8

Originally created by @WhyNotHugo on GitHub (May 13, 2025). Original GitHub issue: https://github.com/ollama/ollama/issues/10691 ### What is the issue? On my host, ollama isn't using the gpu. It also logs that the version parameter is not exposed via `sysfs`: ``` level=WARN source=amd_linux.go:61 msg="ollama recommends running the https://www.amd.com/en/support/linux-drivers" error="amdgpu version file missing: /sys/module/amdgpu/version stat /sys/module/amdgpu/version: no such file or directory" level=INFO source=amd_linux.go:386 msg="amdgpu is supported" gpu=GPU-8668c3be2237aa5f gpu_type=gfx1100 level=INFO source=types.go:130 msg="inference compute" id=GPU-8668c3be2237aa5f library=rocm variant="" compute=gfx1100 driver=0.0 name=1002:744c total="20.0 GiB" available="14.6 GiB" ``` The amdgpu module is loaded: ```console > lsmod | grep ^amdgpu amdgpu 11591680 66 ``` It also exposes quite a few other parameters, but no `version`: ```con > ls -1F /sys/module/amdgpu/ coresize drivers/ holders/ initsize initstate notes/ parameters/ refcnt sections/ taint uevent ``` When executing a model, I also see: ``` ggml_cuda_init: failed to initialize ROCm: no ROCm-capable device is detected ``` GPU usage remains at 0%, only CPU is used. I'm running ollama via docker: ```console docker run --rm --name ollama --device /dev/kfd --device /dev/dri -v ollama:/root/.ollama -p 11434:11434 ollama/ollama:rocm ``` ### Relevant log output ```shell ``` ### OS Linux ### GPU AMD ### CPU Intel ### Ollama version ollama version is 0.6.8
GiteaMirror added the bug label 2026-04-12 18:56:02 -05:00
GiteaMirror commented 2026-04-12 18:56:04 -05:00
Author
Owner
Copy Link

@WhyNotHugo commented on GitHub (May 13, 2025):

Figured it out!

The user running ollama needs access to /dev/dri/renderD128. Changing file ownership works as a quick fix. Usages should be safe; this device is used only for rendering, and should not provide access to, e.g.: the current display's buffer.

<!-- gh-comment-id:2877682721 --> @WhyNotHugo commented on GitHub (May 13, 2025): Figured it out! The user running ollama needs access to `/dev/dri/renderD128`. Changing file ownership works as a quick fix. Usages should be safe; this device is used only for rendering, and should not provide access to, e.g.: the current display's buffer.
GiteaMirror commented 2026-04-12 18:56:05 -05:00
Author
Owner
Copy Link

@WhyNotHugo commented on GitHub (May 13, 2025):

Maybe ollama could explicitly check for access to this file and warn if permissions need to be tweaked?

<!-- gh-comment-id:2877683324 --> @WhyNotHugo commented on GitHub (May 13, 2025): Maybe ollama could explicitly check for access to this file and warn if permissions need to be tweaked?
GiteaMirror commented 2026-04-12 18:56:05 -05:00
Author
Owner
Copy Link

@Nyx1197 commented on GitHub (May 14, 2025):

maybe add /dev/fb0 ?

<!-- gh-comment-id:2878897980 --> @Nyx1197 commented on GitHub (May 14, 2025): maybe add `/dev/fb0` ?
GiteaMirror commented 2026-04-12 18:56:06 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (May 14, 2025):

The usual way to do this is to add the user running ollama to the render group.

https://github.com/ollama/ollama/blob/main/docs/troubleshooting.md#amd-gpu-discovery

<!-- gh-comment-id:2879712306 --> @rick-github commented on GitHub (May 14, 2025): The usual way to do this is to add the user running ollama to the `render` group. https://github.com/ollama/ollama/blob/main/docs/troubleshooting.md#amd-gpu-discovery
GiteaMirror commented 2026-04-12 18:56:07 -05:00
Author
Owner
Copy Link

@WhyNotHugo commented on GitHub (May 14, 2025):

On Alpine, the node is owned by root:video by default, so I configured my host to use root:gpu and added the user to that group.

At this point my issue is solved. Not sure if you want to leave the issue open to track "show a clear error message when permissions are wrong" or just close it.

<!-- gh-comment-id:2879731523 --> @WhyNotHugo commented on GitHub (May 14, 2025): On Alpine, the node is owned by `root:video` by default, so I configured my host to use `root:gpu` and added the user to that group. At this point my issue is solved. Not sure if you want to leave the issue open to track "show a clear error message when permissions are wrong" or just close it.
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#7028
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?