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 #1435] Ollama and Xeon 5660 #26527

New Issue
Closed
opened 2026-04-22 02:51:27 -05:00 by GiteaMirror · 9 comments
GiteaMirror commented 2026-04-22 02:51:27 -05:00
Owner
Copy Link

Originally created by @Andreh1982 on GitHub (Dec 8, 2023).
Original GitHub issue: https://github.com/ollama/ollama/issues/1435

Hello! I'm facing an issue running Ollama in a Dell T610 server with 64GB ram and Xeon 5660:

Dec 8 15:52:15 constellation kernel: [ 1529.028302] traps: ollama-runner[2213] trap invalid opcode ip:472d79 sp:7ffdbe33f680 error:0 in ollama-runner[407000+da000] Dec 8 15:52:15 constellation ollama[2161]: 2023/12/08 15:52:15 llama.go:436: signal: illegal instruction (core dumped)

Please, any help in this situation? The processor is too old to run llm models? :(

Originally created by @Andreh1982 on GitHub (Dec 8, 2023). Original GitHub issue: https://github.com/ollama/ollama/issues/1435 Hello! I'm facing an issue running Ollama in a Dell T610 server with 64GB ram and Xeon 5660: `Dec 8 15:52:15 constellation kernel: [ 1529.028302] traps: ollama-runner[2213] trap invalid opcode ip:472d79 sp:7ffdbe33f680 error:0 in ollama-runner[407000+da000] Dec 8 15:52:15 constellation ollama[2161]: 2023/12/08 15:52:15 llama.go:436: signal: illegal instruction (core dumped)` Please, any help in this situation? The processor is too old to run llm models? :(
GiteaMirror commented 2026-04-22 02:51:31 -05:00
Author
Owner
Copy Link

@easp commented on GitHub (Dec 8, 2023):

Ollama is compiled to use the AVX instruction set instructions. That CPU was released well over a decade ago and doesn't support AVX.
https://ark.intel.com/content/www/us/en/ark/products/47921/intel-xeon-processor-x5660-12m-cache-2-80-ghz-6-40-gt-s-intel-qpi.html

There has been some talk about relaxing that requirement, but for now your only option (besides different hardware) is downloading the repo, changing compile options, and compiling your own version. I believe that all the compile options that need to be changed are in this file: /llm/llama.cpp/generate_linux.go

<!-- gh-comment-id:1847853281 --> @easp commented on GitHub (Dec 8, 2023): Ollama is compiled to use the AVX instruction set instructions. That CPU was released well over a decade ago and doesn't support AVX. https://ark.intel.com/content/www/us/en/ark/products/47921/intel-xeon-processor-x5660-12m-cache-2-80-ghz-6-40-gt-s-intel-qpi.html There has been some talk about relaxing that requirement, but for now your only option (besides different hardware) is downloading the repo, changing compile options, and compiling your own version. I believe that all the compile options that need to be changed are in this file: /llm/llama.cpp/generate_linux.go
GiteaMirror commented 2026-04-22 02:51:32 -05:00
Author
Owner
Copy Link

@technovangelist commented on GitHub (Dec 8, 2023):

I think it is a bit old. That was released in 2010 and AVX instructions started showing up in 2011. And since there is no mention on the Intel page about the 5660, I think it may be missing that critical component, and maybe others. We really want Ollama to be usable by everyone who wants to try it and hope to find a way to accommodate this in the future. I have created an issue to see if there is a way to get around this that we can track : #1442

<!-- gh-comment-id:1847853920 --> @technovangelist commented on GitHub (Dec 8, 2023): I think it is a bit old. That was released in 2010 and AVX instructions started showing up in 2011. And since there is no mention on the Intel page about the 5660, I think it may be missing that critical component, and maybe others. We really want Ollama to be usable by everyone who wants to try it and hope to find a way to accommodate this in the future. I have created an issue to see if there is a way to get around this that we can track : #1442
GiteaMirror commented 2026-04-22 02:51:33 -05:00
Author
Owner
Copy Link

@Cybervet commented on GitHub (Dec 9, 2023):

Just recompiled the repo and I can confirm that runs fine in CPUs without AVX. Running it in a proxmox LXC debian container.
Only thing I did was set AVX=off to /llm/llama.cpp/generate_linux.go

<!-- gh-comment-id:1848604011 --> @Cybervet commented on GitHub (Dec 9, 2023): Just recompiled the repo and I can confirm that runs fine in CPUs without AVX. Running it in a proxmox LXC debian container. Only thing I did was set AVX=off to /llm/llama.cpp/generate_linux.go
GiteaMirror commented 2026-04-22 02:51:33 -05:00
Author
Owner
Copy Link

@Andreh1982 commented on GitHub (Dec 9, 2023):

Thank you so much, guys! Got success here, disabling AVX direct on llama.cpp also did the trick. Cheers! :)

<!-- gh-comment-id:1848646738 --> @Andreh1982 commented on GitHub (Dec 9, 2023): Thank you so much, guys! Got success here, disabling AVX direct on llama.cpp also did the trick. Cheers! :)
GiteaMirror commented 2026-04-22 02:51:35 -05:00
Author
Owner
Copy Link

@Andreh1982 commented on GitHub (Dec 9, 2023):

Oh by the way, i made it on llama.cpp based on the messages on this issue, thank you!

Just recompiled the repo and I can confirm that runs fine in CPUs without AVX. Running it in a proxmox LXC debian container. Only thing I did was set AVX=off to /llm/llama.cpp/generate_linux.go

did you get any performance issues?

<!-- gh-comment-id:1848660974 --> @Andreh1982 commented on GitHub (Dec 9, 2023): Oh by the way, i made it on llama.cpp based on the messages on this issue, thank you! > Just recompiled the repo and I can confirm that runs fine in CPUs without AVX. Running it in a proxmox LXC debian container. Only thing I did was set AVX=off to /llm/llama.cpp/generate_linux.go did you get any performance issues?
GiteaMirror commented 2026-04-22 02:51:36 -05:00
Author
Owner
Copy Link

@Cybervet commented on GitHub (Dec 9, 2023):

Its a bit slow, but that is expected , 2.5 -5 (Orca2) tokens/sec with 4 cores assigned to the container. I am trying now to pass through the GPU to see if it will make a difference.

<!-- gh-comment-id:1848677961 --> @Cybervet commented on GitHub (Dec 9, 2023): Its a bit slow, but that is expected , 2.5 -5 (Orca2) tokens/sec with 4 cores assigned to the container. I am trying now to pass through the GPU to see if it will make a difference.
GiteaMirror commented 2026-04-22 02:51:37 -05:00
Author
Owner
Copy Link

@Andreh1982 commented on GitHub (Dec 9, 2023):

Its a bit slow, but that is expected , 2.5 -5 (Orca2) tokens/sec with 4 cores assigned to the container. I am trying now to pass through the GPU to see if it will make a difference.

here the first run is really slow, but after that, goes fine i guess. I will try later to emulate the flag using a custom cpu in proxmox. Not having AVX instructions is a problem, many other services need it like mongodb for example...

<!-- gh-comment-id:1848680911 --> @Andreh1982 commented on GitHub (Dec 9, 2023): > Its a bit slow, but that is expected , 2.5 -5 (Orca2) tokens/sec with 4 cores assigned to the container. I am trying now to pass through the GPU to see if it will make a difference. here the first run is really slow, but after that, goes fine i guess. I will try later to emulate the flag using a custom cpu in proxmox. Not having AVX instructions is a problem, many other services need it like mongodb for example...
GiteaMirror commented 2026-04-22 02:51:38 -05:00
Author
Owner
Copy Link

@Cybervet commented on GitHub (Dec 9, 2023):

Yeap I know , I just have a couple of workstations with dual cpus around and I thought to try them out.

<!-- gh-comment-id:1848684459 --> @Cybervet commented on GitHub (Dec 9, 2023): Yeap I know , I just have a couple of workstations with dual cpus around and I thought to try them out.
GiteaMirror commented 2026-04-22 02:51:38 -05:00
Author
Owner
Copy Link

@Andreh1982 commented on GitHub (Dec 9, 2023):

I just bought one past week... haha, didnt expected this kind of problem :(

<!-- gh-comment-id:1848718847 --> @Andreh1982 commented on GitHub (Dec 9, 2023): I just bought one past week... haha, didnt expected this kind of problem :(
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
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#26527
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?