github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-07 00:22:43 -05:00
Code Issues 14.5k Packages Projects Releases 120 Wiki Activity

[GH-ISSUE #8999] Ollama cannot utilize multiple physical CPUs. #52357

New Issue
Open
opened 2026-04-28 23:05:36 -05:00 by GiteaMirror · 6 comments
GiteaMirror commented 2026-04-28 23:05:36 -05:00
Owner
Copy Link

Originally created by @xiangyanggong on GitHub (Feb 11, 2025).
Original GitHub issue: https://github.com/ollama/ollama/issues/8999

Originally assigned to: @mxyng on GitHub.

What is the issue?

There is no GPU in my environment, only two physical CPUs. Each CPU has 48 cores. However, after I ran the model, I noticed the parameter "--threads 48". When I run the model, I find that Ollama only uses one physical CPU. Is this due to the characteristics of Ollama, or does it require some other configurations?

Relevant log output

/usr/local/lib/ollama/runners/cpu_avx2/ollama_llama_server runner --model /usr/share/ollama/.ollama/models/blobs/sha256-4cd576d9aa16961244012223abf01445567b061f1814b57dfef699e4cf8df339 --ctx -size 8192 --batch-size 512 --threads 48 --no-mmap --parallel 4 --port 39029

OS

Linux

GPU

No response

CPU

Intel

Ollama version

0.5.7

Originally created by @xiangyanggong on GitHub (Feb 11, 2025). Original GitHub issue: https://github.com/ollama/ollama/issues/8999 Originally assigned to: @mxyng on GitHub. ### What is the issue? There is no GPU in my environment, only two physical CPUs. Each CPU has 48 cores. However, after I ran the model, I noticed the parameter "--threads 48". When I run the model, I find that Ollama only uses one physical CPU. Is this due to the characteristics of Ollama, or does it require some other configurations? ### Relevant log output ```shell /usr/local/lib/ollama/runners/cpu_avx2/ollama_llama_server runner --model /usr/share/ollama/.ollama/models/blobs/sha256-4cd576d9aa16961244012223abf01445567b061f1814b57dfef699e4cf8df339 --ctx -size 8192 --batch-size 512 --threads 48 --no-mmap --parallel 4 --port 39029 ``` ### OS Linux ### GPU _No response_ ### CPU Intel ### Ollama version 0.5.7
GiteaMirror added the bug label 2026-04-28 23:05:36 -05:00
GiteaMirror commented 2026-04-28 23:05:38 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Feb 11, 2025):

By default, ollama uses only the performance cores in the CPUs. You can override this by setting num_thread.

<!-- gh-comment-id:2650562843 --> @rick-github commented on GitHub (Feb 11, 2025): By default, ollama uses only the performance cores in the CPUs. You can override this by setting [`num_thread`](https://github.com/ollama/ollama/blob/main/docs/api.md#request-8:~:text=use_mlock%22%3A%20false%2C%0A%20%20%20%20%22-,num_thread,-%22%3A%208%0A%20%20%7D%0A%7D).
GiteaMirror commented 2026-04-28 23:05:38 -05:00
Author
Owner
Copy Link

@SmallBlueE commented on GitHub (Feb 11, 2025):

By default, ollama uses only the performance cores in the CPUs. You can override this by setting num_thread.

If I set ollama as linux service, could you know how to set this parameter 'num_thread'? Environment=??

<!-- gh-comment-id:2650625671 --> @SmallBlueE commented on GitHub (Feb 11, 2025): > By default, ollama uses only the performance cores in the CPUs. You can override this by setting [`num_thread`](https://github.com/ollama/ollama/blob/main/docs/api.md#request-8:~:text=use_mlock%22%3A%20false%2C%0A%20%20%20%20%22-,num_thread,-%22%3A%208%0A%20%20%7D%0A%7D). If I set ollama as linux service, could you know how to set this parameter 'num_thread'? Environment=??
GiteaMirror commented 2026-04-28 23:05:38 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Feb 11, 2025):

Either set it in the API call ("options":{"num_thread":96}) or create a new model:

echo FROM deepseek-r1:7b > Modelfile
echo PARAMETER num_thread 96 >> Modelfile
ollama create deepseek:t96

$ ollama run deepseek-r1:7b ''
$ ps who cmd p$(pidof ollama_llama_server) | sed -e 's/.*\(--thread\)/\1/'
--threads 8 --parallel 1 --port 46763
$ ollama run deepseek:t96 ''
$ ps who cmd p$(pidof ollama_llama_server) | sed -e 's/.*\(--thread\)/\1/'
--threads 96 --parallel 1 --port 41649
<!-- gh-comment-id:2650672329 --> @rick-github commented on GitHub (Feb 11, 2025): Either set it in the API call (`"options":{"num_thread":96}`) or create a new model: ```console echo FROM deepseek-r1:7b > Modelfile echo PARAMETER num_thread 96 >> Modelfile ollama create deepseek:t96 ``` ```console $ ollama run deepseek-r1:7b '' $ ps who cmd p$(pidof ollama_llama_server) | sed -e 's/.*\(--thread\)/\1/' --threads 8 --parallel 1 --port 46763 $ ollama run deepseek:t96 '' $ ps who cmd p$(pidof ollama_llama_server) | sed -e 's/.*\(--thread\)/\1/' --threads 96 --parallel 1 --port 41649 ```
GiteaMirror commented 2026-04-28 23:05:38 -05:00
Author
Owner
Copy Link

@SmallBlueE commented on GitHub (Feb 11, 2025):

Either set it in the API call ("options":{"num_thread":96}) or create a new model:

echo FROM deepseek-r1:7b > Modelfile
echo PARAMETER num_thread 96 >> Modelfile
ollama create deepseek:t96
$ ollama run deepseek-r1:7b ''
$ ps who cmd p$(pidof ollama_llama_server) | sed -e 's/.(--thread)/\1/'
--threads 8 --parallel 1 --port 46763
$ ollama run deepseek:t96 ''
$ ps who cmd p$(pidof ollama_llama_server) | sed -e 's/.(--thread)/\1/'
--threads 96 --parallel 1 --port 41649

Can you give us a template of Modefile for Deepseek-r1-671b-Q8_0, this model has 15 gguf file. How to merge it into one gguf file correctly for Ollama? Cat ..15 file > 1 file ?

<!-- gh-comment-id:2650697845 --> @SmallBlueE commented on GitHub (Feb 11, 2025): > Either set it in the API call (`"options":{"num_thread":96}`) or create a new model: > > echo FROM deepseek-r1:7b > Modelfile > echo PARAMETER num_thread 96 >> Modelfile > ollama create deepseek:t96 > $ ollama run deepseek-r1:7b '' > $ ps who cmd p$(pidof ollama_llama_server) | sed -e 's/.*\(--thread\)/\1/' > --threads 8 --parallel 1 --port 46763 > $ ollama run deepseek:t96 '' > $ ps who cmd p$(pidof ollama_llama_server) | sed -e 's/.*\(--thread\)/\1/' > --threads 96 --parallel 1 --port 41649 Can you give us a template of Modefile for Deepseek-r1-671b-Q8_0, this model has 15 gguf file. How to merge it into one gguf file correctly for Ollama? Cat ..15 file > 1 file ?
GiteaMirror commented 2026-04-28 23:05:38 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Feb 11, 2025):

https://unsloth.ai/blog/deepseekr1-dynamic#:~:text=llama%2Dgguf%2Dsplit%20%2D%2D-,merge,-%5C%0A%20%20DeepSeek%2DR1%2DGGUF

<!-- gh-comment-id:2650699828 --> @rick-github commented on GitHub (Feb 11, 2025): https://unsloth.ai/blog/deepseekr1-dynamic#:~:text=llama%2Dgguf%2Dsplit%20%2D%2D-,merge,-%5C%0A%20%20DeepSeek%2DR1%2DGGUF
GiteaMirror commented 2026-04-28 23:05:39 -05:00
Author
Owner
Copy Link

@Diksha06122 commented on GitHub (Apr 28, 2025):

Try this its worked for me:

sudo docker run -d --name ollama
--cpus=16
--cpuset-cpus="0-15"
-p 11434:11434
ollama/ollama

<!-- gh-comment-id:2834810124 --> @Diksha06122 commented on GitHub (Apr 28, 2025): Try this its worked for me: sudo docker run -d --name ollama \ --cpus=16 \ --cpuset-cpus="0-15" \ -p 11434:11434 \ ollama/ollama
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#52357
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?