github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-07 08:30:05 -05:00
Code Issues 14.5k Packages Projects Releases 120 Wiki Activity

[GH-ISSUE #10715] iMac not using GPU #7041

New Issue
Closed
opened 2026-04-12 18:57:02 -05:00 by GiteaMirror · 11 comments
GiteaMirror commented 2026-04-12 18:57:02 -05:00
Owner
Copy Link

Originally created by @russell-kitchen on GitHub (May 15, 2025).
Original GitHub issue: https://github.com/ollama/ollama/issues/10715

What is the issue?

Platform:
Mac OSX
Retina 5K, 27-inch, 2019
Sonoma 14.7.2
Radeon Pro Vega 48 8 GB

It doesn't appear that Ollama is using the GPU at all. When running ollama ps is says 100% CPU.

I'd posted an incorrect issue a week or so back without properly getting the right information. It appears my mac GPU does support Metal(as noted below from the About Mac section on my computer) which theoretically means it should work. Any ideas?

Chipset Model: Radeon Pro Vega 48
Type: GPU
Bus: PCIe
PCIe Lane Width: x16
VRAM (Total): 8 GB
Vendor: AMD (0x1002)
Device ID: 0x6869
Revision ID: 0x0000
ROM Revision: 113-D0650E-072
VBIOS Version: 113-D05001A1XG-011
Option ROM Version: 113-D05001A1XG-011
EFI Driver Version: 01.01.072
Metal Support: Metal 3

Thank you!

Relevant log output

OS

macOS

GPU

AMD

CPU

Intel

Ollama version

0.6.8

Originally created by @russell-kitchen on GitHub (May 15, 2025). Original GitHub issue: https://github.com/ollama/ollama/issues/10715 ### What is the issue? Platform: Mac OSX Retina 5K, 27-inch, 2019 Sonoma 14.7.2 Radeon Pro Vega 48 8 GB It doesn't appear that Ollama is using the GPU at all. When running `ollama ps` is says 100% CPU. I'd posted an incorrect issue a week or so back without properly getting the right information. It appears my mac GPU does support Metal(as noted below from the About Mac section on my computer) which theoretically means it should work. Any ideas? Chipset Model: Radeon Pro Vega 48 Type: GPU Bus: PCIe PCIe Lane Width: x16 VRAM (Total): 8 GB Vendor: AMD (0x1002) Device ID: 0x6869 Revision ID: 0x0000 ROM Revision: 113-D0650E-072 VBIOS Version: 113-D05001A1XG-011 Option ROM Version: 113-D05001A1XG-011 EFI Driver Version: 01.01.072 Metal Support: Metal 3 Thank you! ### Relevant log output ```shell ``` ### OS macOS ### GPU AMD ### CPU Intel ### Ollama version 0.6.8
GiteaMirror added the bugmacos labels 2026-04-12 18:57:02 -05:00
GiteaMirror commented 2026-04-12 18:57:04 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (May 15, 2025):

Server logs may aid in debugging.

<!-- gh-comment-id:2884728651 --> @rick-github commented on GitHub (May 15, 2025): [Server logs](https://github.com/ollama/ollama/blob/main/docs/troubleshooting.md#how-to-troubleshoot-issues) may aid in debugging.
GiteaMirror commented 2026-04-12 18:57:05 -05:00
Author
Owner
Copy Link

@russell-kitchen commented on GitHub (May 15, 2025):

Server logs may aid in debugging.

Fab, thank you for the pointer!

My old one:
server.log

A fresh one with two chat prompts ran with 100% CPU:
server.log

<!-- gh-comment-id:2884860056 --> @russell-kitchen commented on GitHub (May 15, 2025): > [Server logs](https://github.com/ollama/ollama/blob/main/docs/troubleshooting.md#how-to-troubleshoot-issues) may aid in debugging. Fab, thank you for the pointer! My old one: [server.log](https://github.com/user-attachments/files/20233415/server.log) A fresh one with two chat prompts ran with 100% CPU: [server.log](https://github.com/user-attachments/files/20233497/server.log)
GiteaMirror commented 2026-04-12 18:57:06 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (May 15, 2025):

time=2025-05-15T20:26:23.964+01:00 level=INFO source=types.go:130 msg="inference compute" id="" library=cpu variant="" compute="" driver=0.0 name="" total="16.0 GiB" available="5.0 GiB"

The GPU is not detected. If you add OLLAMA_DEBUG=1 to the server environment, there will be more details about device detection.

<!-- gh-comment-id:2884990730 --> @rick-github commented on GitHub (May 15, 2025): ``` time=2025-05-15T20:26:23.964+01:00 level=INFO source=types.go:130 msg="inference compute" id="" library=cpu variant="" compute="" driver=0.0 name="" total="16.0 GiB" available="5.0 GiB" ``` The GPU is not detected. If you add `OLLAMA_DEBUG=1` to the server environment, there will be more details about device detection.
GiteaMirror commented 2026-04-12 18:57:06 -05:00
Author
Owner
Copy Link

@russell-kitchen commented on GitHub (May 16, 2025):

Aces, thank you! With debug on I believe:

server.log

<!-- gh-comment-id:2885705065 --> @russell-kitchen commented on GitHub (May 16, 2025): Aces, thank you! With debug on I believe: [server.log](https://github.com/user-attachments/files/20238814/server.log)
GiteaMirror commented 2026-04-12 18:57:07 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (May 16, 2025):

I went back to your previous issue and realized I overlooked something: you are using docker. Metal acceleration is not supported in docker. See also #5652. To use Metal, you need to install ollama in the base system.

<!-- gh-comment-id:2886267838 --> @rick-github commented on GitHub (May 16, 2025): I went back to your previous issue and realized I overlooked something: you are using docker. Metal acceleration is [not supported](https://chariotsolutions.com/blog/post/apple-silicon-gpus-docker-and-ollama-pick-two/) in docker. See also #5652. To use Metal, you need to install ollama in the base system.
GiteaMirror commented 2026-04-12 18:57:07 -05:00
Author
Owner
Copy Link

@russell-kitchen commented on GitHub (May 16, 2025):

Sorry that was a typo with the docker ps. I use docker for work so had a brain fart there which I amended in this post. Not using docker, just the command line interface

<!-- gh-comment-id:2886315009 --> @russell-kitchen commented on GitHub (May 16, 2025): Sorry that was a typo with the `docker ps`. I use docker for work so had a brain fart there which I amended in this post. Not using docker, just the command line interface
GiteaMirror commented 2026-04-12 18:57:08 -05:00
Author
Owner
Copy Link

@russell-kitchen commented on GitHub (May 16, 2025):

Also installed ollama through the site and not homebrew. Saw that homebrew was causing other people issues with GPUs not being registered so installed it directly through main site's download from:
https://ollama.com/download/mac

<!-- gh-comment-id:2886325287 --> @russell-kitchen commented on GitHub (May 16, 2025): Also installed ollama through the site and not homebrew. Saw that homebrew was causing other people issues with GPUs not being registered so installed it directly through main site's download from: https://ollama.com/download/mac
GiteaMirror commented 2026-04-12 18:57:08 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (May 16, 2025):

Hmm, in that case I don't know what's going on. ollama is completely overlooking the GPU. I'm not familiar with the Apple ecosystem, but this page seems to indicate that Metal 3 is supported in Sonoma and this document lists AMD Vega as a supported chipset in Metal 3. Does MacOS have additional restrictions on devices? For example, to use GPU devices on Linux, the ollama user is added to the render and video groups.

<!-- gh-comment-id:2886364621 --> @rick-github commented on GitHub (May 16, 2025): Hmm, in that case I don't know what's going on. ollama is completely overlooking the GPU. I'm not familiar with the Apple ecosystem, but [this page](https://support.apple.com/en-gb/102894) seems to indicate that Metal 3 is supported in Sonoma and [this document](https://developer.apple.com/metal/Metal-Feature-Set-Tables.pdf) lists AMD Vega as a supported chipset in Metal 3. Does MacOS have additional restrictions on devices? For example, to use GPU devices on Linux, the ollama user is added to the render and video groups.
GiteaMirror commented 2026-04-12 18:57:09 -05:00
Author
Owner
Copy Link

@russell-kitchen commented on GitHub (May 16, 2025):

I don't believe so no with regards to GPU stuff

<!-- gh-comment-id:2887032458 --> @russell-kitchen commented on GitHub (May 16, 2025): I don't believe so no with regards to GPU stuff
GiteaMirror commented 2026-04-12 18:57:09 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (May 16, 2025):

https://github.com/ollama/ollama/issues/1016

<!-- gh-comment-id:2887073533 --> @rick-github commented on GitHub (May 16, 2025): https://github.com/ollama/ollama/issues/1016
GiteaMirror commented 2026-04-12 18:57:10 -05:00
Author
Owner
Copy Link

@russell-kitchen commented on GitHub (May 17, 2025):

Ah. If I'm reading that thread right it looks like it's not something that'll get fixed for a bit then, and even if it does it appears GPU lending a hand in these cases often doesn't help much, or even hampers.

<!-- gh-comment-id:2888444782 --> @russell-kitchen commented on GitHub (May 17, 2025): Ah. If I'm reading that thread right it looks like it's not something that'll get fixed for a bit then, and even if it does it appears GPU lending a hand in these cases often doesn't help much, or even hampers.
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 macos
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#7041
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?