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 #13844] GPU not used. Wrong folders. Ubuntu #34825

New Issue
Closed
opened 2026-04-22 18:44:09 -05:00 by GiteaMirror · 10 comments
GiteaMirror commented 2026-04-22 18:44:09 -05:00
Owner
Copy Link

Originally created by @Alex-CodeLab on GitHub (Jan 22, 2026).
Original GitHub issue: https://github.com/ollama/ollama/issues/13844

What is the issue?

Issue:
After the update, Ollama does not use/detect the GPU

after lots of attempts to re-install it turns out the it is looking in the wrong folder for the ROCm libs.
even if the folder is added to PATH or OLLAMA_LIBRARY_PATH , it is not detected.

Workaround:
cp -r /usr/lib/ollama/rocm/* /usr/local/lib/ollama/rocm/

Ubuntu 25.10
Kernel: 6.14.0-37
Max+ 395 / 8060S

Relevant log output

OS

Linux

GPU

AMD

CPU

AMD

Ollama version

0.14.3

Originally created by @Alex-CodeLab on GitHub (Jan 22, 2026). Original GitHub issue: https://github.com/ollama/ollama/issues/13844 ### What is the issue? Issue: After the update, Ollama does not use/detect the GPU after lots of attempts to re-install it turns out the it is looking in the wrong folder for the ROCm libs. even if the folder is added to PATH or OLLAMA_LIBRARY_PATH , it is not detected. Workaround: cp -r /usr/lib/ollama/rocm/* /usr/local/lib/ollama/rocm/ Ubuntu 25.10 Kernel: 6.14.0-37 Max+ 395 / 8060S ### Relevant log output ```shell ``` ### OS Linux ### GPU AMD ### CPU AMD ### Ollama version 0.14.3
GiteaMirror added the bug label 2026-04-22 18:44:09 -05:00
GiteaMirror commented 2026-04-22 18:44:11 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Jan 22, 2026):

How did you install ollama? How have you installed previous versions of ollama?

<!-- gh-comment-id:3784993574 --> @rick-github commented on GitHub (Jan 22, 2026): How did you install ollama? How have you installed previous versions of ollama?
GiteaMirror commented 2026-04-22 18:44:12 -05:00
Author
Owner
Copy Link

@Alex-CodeLab commented on GitHub (Jan 22, 2026):

How did you install ollama? How have you installed previous versions of ollama?

I removed the old version manually, deleting the files and folders.
Then installed ollama-linux-amd64-rocm.tar.zst manually , and used the installer script.

<!-- gh-comment-id:3785349888 --> @Alex-CodeLab commented on GitHub (Jan 22, 2026): > How did you install ollama? How have you installed previous versions of ollama? I removed the old version manually, deleting the files and folders. Then installed ollama-linux-amd64-rocm.tar.zst manually , and used the installer script.
GiteaMirror commented 2026-04-22 18:44:13 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Jan 22, 2026):

The installer script should install the ROCm libraries automatically, manual download should not be required. If you combined both installation methods, be aware that the install script installs in /usr/local and the manual install instruction use /usr. In future if you continue to use both methods, use /usr/local as the target in the manual install method.

<!-- gh-comment-id:3785442523 --> @rick-github commented on GitHub (Jan 22, 2026): The installer script should install the ROCm libraries automatically, manual download should not be required. If you combined both installation methods, be aware that the install script installs in `/usr/local` and the manual install instruction use `/usr`. In future if you continue to use both methods, use `/usr/local` as the target in the manual install method.
GiteaMirror commented 2026-04-22 18:44:14 -05:00
Author
Owner
Copy Link

@Alex-CodeLab commented on GitHub (Jan 22, 2026):

The installer script should install the ROCm libraries automatically, manual download should not be required. If you combined both installation methods, be aware that the install script installs in /usr/local and the manual install instruction use /usr. In future if you continue to use both methods, use /usr/local as the target in the manual install method.

It should, but it doesn't.
Using the installer was the first thing I tried.

<!-- gh-comment-id:3785676172 --> @Alex-CodeLab commented on GitHub (Jan 22, 2026): > The installer script should install the ROCm libraries automatically, manual download should not be required. If you combined both installation methods, be aware that the install script installs in `/usr/local` and the manual install instruction use `/usr`. In future if you continue to use both methods, use `/usr/local` as the target in the manual install method. It should, but it doesn't. Using the installer was the first thing I tried.
GiteaMirror commented 2026-04-22 18:44:14 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Jan 22, 2026):

$ lscpu | grep Model.name
Model name:                           AMD RYZEN AI MAX+ 395 w/ Radeon 8060S
$ curl -fsSL https://ollama.com/install.sh | sh
>>> Cleaning up old version at /usr/local/lib/ollama
>>> Installing ollama to /usr/local
>>> Downloading ollama-linux-amd64.tar.zst
######################################################################## 100.0%
>>> Adding ollama user to render group...
>>> Adding ollama user to video group...
>>> Adding current user to ollama group...
>>> Creating ollama systemd service...
>>> Enabling and starting ollama service...
Created symlink /etc/systemd/system/default.target.wants/ollama.service → /etc/systemd/system/ollama.service.
>>> Downloading ollama-linux-amd64-rocm.tar.zst
######################################################################## 100.0%
>>> The Ollama API is now available at 127.0.0.1:11434.
>>> Install complete. Run "ollama" from the command line.
>>> AMD GPU ready.
$ ollama -v
ollama version is 0.14.3

The installer downloads and installs the base server, and then downloads and install the ROCm libraries.

<!-- gh-comment-id:3785687586 --> @rick-github commented on GitHub (Jan 22, 2026): ```console $ lscpu | grep Model.name Model name: AMD RYZEN AI MAX+ 395 w/ Radeon 8060S $ curl -fsSL https://ollama.com/install.sh | sh >>> Cleaning up old version at /usr/local/lib/ollama >>> Installing ollama to /usr/local >>> Downloading ollama-linux-amd64.tar.zst ######################################################################## 100.0% >>> Adding ollama user to render group... >>> Adding ollama user to video group... >>> Adding current user to ollama group... >>> Creating ollama systemd service... >>> Enabling and starting ollama service... Created symlink /etc/systemd/system/default.target.wants/ollama.service → /etc/systemd/system/ollama.service. >>> Downloading ollama-linux-amd64-rocm.tar.zst ######################################################################## 100.0% >>> The Ollama API is now available at 127.0.0.1:11434. >>> Install complete. Run "ollama" from the command line. >>> AMD GPU ready. $ ollama -v ollama version is 0.14.3 ``` The installer downloads and installs the base server, and then downloads and install the ROCm libraries.
GiteaMirror commented 2026-04-22 18:44:14 -05:00
Author
Owner
Copy Link

@Alex-CodeLab commented on GitHub (Jan 22, 2026):

$ curl -fsSL https://ollama.com/install.sh | sh
>>> Cleaning up old version at /usr/local/lib/ollama
[sudo: authenticate] Password: 
>>> Installing ollama to /usr/local
>>> Downloading ollama-linux-amd64.tar.zst
######################################################################## 100.0%
>>> Adding ollama user to render group...
>>> Adding ollama user to video group...
>>> Adding current user to ollama group...
>>> Creating ollama systemd service...
>>> Enabling and starting ollama service...
>>> NVIDIA GPU installed.

...
log:

$ ollama serve 
source=routes.go:1629 msg="server config" env="map[CUDA_VISIBLE_DEVICES: GGML_VK_VISIBLE_DEVICES: GPU_DEVICE_ORDINAL: HIP_VISIBLE_DEVICES: HSA_OVERRIDE_GFX_VERSION: HTTPS_PROXY: HTTP_PROXY: NO_PROXY: OLLAMA_CONTEXT_LENGTH:4096 OLLAMA_DEBUG:INFO OLLAMA_FLASH_ATTENTION:false OLLAMA_GPU_OVERHEAD:0 OLLAMA_HOST:http://0.0.0.0:11434 OLLAMA_KEEP_ALIVE:5m0s OLLAMA_KV_CACHE_TYPE: OLLAMA_LLM_LIBRARY: OLLAMA_LOAD_TIMEOUT:5m0s OLLAMA_MAX_LOADED_MODELS:0 OLLAMA_MAX_QUEUE:512 OLLAMA_MODELS:/home/user/.ollama/models OLLAMA_MULTIUSER_CACHE:false OLLAMA_NEW_ENGINE:false OLLAMA_NOHISTORY:false OLLAMA_NOPRUNE:false OLLAMA_NUM_PARALLEL:1 OLLAMA_ORIGINS:[http://localhost https://localhost http://localhost:* https://localhost:* http://127.0.0.1 https://127.0.0.1 http://127.0.0.1:* https://127.0.0.1:* http://0.0.0.0 https://0.0.0.0 http://0.0.0.0:* https://0.0.0.0:* app://* file://* tauri://* vscode-webview://* vscode-file://*] OLLAMA_REMOTES:[ollama.com] OLLAMA_SCHED_SPREAD:false OLLAMA_VULKAN:false ROCR_VISIBLE_DEVICES: http_proxy: https_proxy: no_proxy:]"
source=images.go:501 msg="total blobs: 62"
source=images.go:508 msg="total unused blobs removed: 0"
source=routes.go:1682 msg="Listening on [::]:11434 (version 0.14.3)"
source=runner.go:67 msg="discovering available GPUs..."
source=server.go:429 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 40911"
source=server.go:429 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 36943"
source=runner.go:106 msg="experimental Vulkan support disabled.  To enable, set OLLAMA_VULKAN=1"
source=types.go:60 msg="inference compute" id=cpu library=cpu compute="" name=cpu description=cpu libdirs=ollama driver="" pci_id="" type="" total="14.8 GiB" available="5.8 GiB"
source=routes.go:1723 msg="entering low vram mode" "total vram"="0 B" threshold="20.0 GiB"
<!-- gh-comment-id:3786017278 --> @Alex-CodeLab commented on GitHub (Jan 22, 2026): ``` $ curl -fsSL https://ollama.com/install.sh | sh >>> Cleaning up old version at /usr/local/lib/ollama [sudo: authenticate] Password: >>> Installing ollama to /usr/local >>> Downloading ollama-linux-amd64.tar.zst ######################################################################## 100.0% >>> Adding ollama user to render group... >>> Adding ollama user to video group... >>> Adding current user to ollama group... >>> Creating ollama systemd service... >>> Enabling and starting ollama service... >>> NVIDIA GPU installed. ``` ... log: ``` $ ollama serve source=routes.go:1629 msg="server config" env="map[CUDA_VISIBLE_DEVICES: GGML_VK_VISIBLE_DEVICES: GPU_DEVICE_ORDINAL: HIP_VISIBLE_DEVICES: HSA_OVERRIDE_GFX_VERSION: HTTPS_PROXY: HTTP_PROXY: NO_PROXY: OLLAMA_CONTEXT_LENGTH:4096 OLLAMA_DEBUG:INFO OLLAMA_FLASH_ATTENTION:false OLLAMA_GPU_OVERHEAD:0 OLLAMA_HOST:http://0.0.0.0:11434 OLLAMA_KEEP_ALIVE:5m0s OLLAMA_KV_CACHE_TYPE: OLLAMA_LLM_LIBRARY: OLLAMA_LOAD_TIMEOUT:5m0s OLLAMA_MAX_LOADED_MODELS:0 OLLAMA_MAX_QUEUE:512 OLLAMA_MODELS:/home/user/.ollama/models OLLAMA_MULTIUSER_CACHE:false OLLAMA_NEW_ENGINE:false OLLAMA_NOHISTORY:false OLLAMA_NOPRUNE:false OLLAMA_NUM_PARALLEL:1 OLLAMA_ORIGINS:[http://localhost https://localhost http://localhost:* https://localhost:* http://127.0.0.1 https://127.0.0.1 http://127.0.0.1:* https://127.0.0.1:* http://0.0.0.0 https://0.0.0.0 http://0.0.0.0:* https://0.0.0.0:* app://* file://* tauri://* vscode-webview://* vscode-file://*] OLLAMA_REMOTES:[ollama.com] OLLAMA_SCHED_SPREAD:false OLLAMA_VULKAN:false ROCR_VISIBLE_DEVICES: http_proxy: https_proxy: no_proxy:]" source=images.go:501 msg="total blobs: 62" source=images.go:508 msg="total unused blobs removed: 0" source=routes.go:1682 msg="Listening on [::]:11434 (version 0.14.3)" source=runner.go:67 msg="discovering available GPUs..." source=server.go:429 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 40911" source=server.go:429 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 36943" source=runner.go:106 msg="experimental Vulkan support disabled. To enable, set OLLAMA_VULKAN=1" source=types.go:60 msg="inference compute" id=cpu library=cpu compute="" name=cpu description=cpu libdirs=ollama driver="" pci_id="" type="" total="14.8 GiB" available="5.8 GiB" source=routes.go:1723 msg="entering low vram mode" "total vram"="0 B" threshold="20.0 GiB" ```
GiteaMirror commented 2026-04-22 18:44:15 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Jan 22, 2026):

What's the output of lspci and sudo lshw -c display -disable network?

<!-- gh-comment-id:3786049433 --> @rick-github commented on GitHub (Jan 22, 2026): What's the output of `lspci` and `sudo lshw -c display -disable network`?
GiteaMirror commented 2026-04-22 18:44:15 -05:00
Author
Owner
Copy Link

@Alex-CodeLab commented on GitHub (Jan 22, 2026):

$ sudo lshw -c display -disable network
  *-display                 
       description: Display controller
       product: Strix Halo [Radeon Graphics / Radeon 8050S Graphics / Radeon 8060S Graphics]
       vendor: Advanced Micro Devices, Inc. [AMD/ATI]
       physical id: 0
       bus info: pci@0000:c5:00.0
       logical name: /dev/fb0
       version: c1
       width: 64 bits
       clock: 33MHz
       capabilities: pm pciexpress msi msix bus_master cap_list fb
       configuration: depth=32 driver=amdgpu latency=0 mode=3440x1440 resolution=3440,1440 visual=truecolor xres=3440 yres=1440
       resources: iomemory:800-7ff irq:102 memory:8000000000-87ffffffff memory:90000000-9fffffff ioport:e000(size=256) memory:a0200000-a02fffff


$ lspci 
...
c5:00.0 Display controller: Advanced Micro Devices, Inc. [AMD/ATI] Strix Halo [Radeon Graphics / Radeon 8050S Graphics / Radeon 8060S Graphics] (rev c1)
...

Not sure why I'm posting this, because I've been through that 20 times already.

After the installer, this folder should have been created:
/usr/local/lib/ollama/rocm/

manually copied ROCm folder in there, and GPU is detected again.

<!-- gh-comment-id:3786081452 --> @Alex-CodeLab commented on GitHub (Jan 22, 2026): ``` $ sudo lshw -c display -disable network *-display description: Display controller product: Strix Halo [Radeon Graphics / Radeon 8050S Graphics / Radeon 8060S Graphics] vendor: Advanced Micro Devices, Inc. [AMD/ATI] physical id: 0 bus info: pci@0000:c5:00.0 logical name: /dev/fb0 version: c1 width: 64 bits clock: 33MHz capabilities: pm pciexpress msi msix bus_master cap_list fb configuration: depth=32 driver=amdgpu latency=0 mode=3440x1440 resolution=3440,1440 visual=truecolor xres=3440 yres=1440 resources: iomemory:800-7ff irq:102 memory:8000000000-87ffffffff memory:90000000-9fffffff ioport:e000(size=256) memory:a0200000-a02fffff ``` ``` $ lspci ... c5:00.0 Display controller: Advanced Micro Devices, Inc. [AMD/ATI] Strix Halo [Radeon Graphics / Radeon 8050S Graphics / Radeon 8060S Graphics] (rev c1) ... ``` Not sure why I'm posting this, because I've been through that 20 times already. After the installer, this folder should have been created: /usr/local/lib/ollama/rocm/ manually copied ROCm folder in there, and GPU is detected again.
GiteaMirror commented 2026-04-22 18:44:16 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Jan 22, 2026):

Not sure why I'm posting this, because I've been through that 20 times already.

Because you want a solution?

Remove nvidia-smi.

<!-- gh-comment-id:3786100463 --> @rick-github commented on GitHub (Jan 22, 2026): > Not sure why I'm posting this, because I've been through that 20 times already. Because you want a solution? Remove nvidia-smi.
GiteaMirror commented 2026-04-22 18:44:16 -05:00
Author
Owner
Copy Link

@Alex-CodeLab commented on GitHub (Jan 22, 2026):

Remove nvidia-smi.

yes, that seems to help.

Now the installer now installs ROCm

<!-- gh-comment-id:3786209396 --> @Alex-CodeLab commented on GitHub (Jan 22, 2026): > Remove nvidia-smi. yes, that seems to help. Now the installer now installs ROCm
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#34825
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?