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 #9663] Please document when AMD iGPU support is planned #6304

New Issue
Open
opened 2026-04-12 17:45:43 -05:00 by GiteaMirror · 11 comments
GiteaMirror commented 2026-04-12 17:45:43 -05:00
Owner
Copy Link

Originally created by @justincranford on GitHub (Mar 11, 2025).
Original GitHub issue: https://github.com/ollama/ollama/issues/9663

Originally assigned to: @dhiltgen on GitHub.

Request

Could you add AMD iGPU acceleration to the roadmap?

It would be great to get clarification about if and when AMD iGPU support might be coming.

I found these comments in the source code that imply AMD iGPU support has been considered.

aee28501b5/discover/amd_windows.go (L124)

aee28501b5/discover/amd_linux.go (L293)

I can't find any documentation under github.com/ollama/ollama to know if and when AMD iGPUs will be supported in ollama rocm.

My Use case

I am running ollama latest bundled in openweb-ui latest Docker container.

docker run -d -p 3000:8080 -v ollama:/root/.ollama -v open-webui:/app/backend/data --name open-webui --restart always ghcr.io/open-webui/open-webui:ollama

My desktop for software development:

  • CPU: AMD 7600X CPU
  • RAM 2x16GB DDR5-6000 CL30
  • SSD: 4TB Samsung Pro 990 PCIe 4.0
  • dGPU: n/a
  • IDE: VScode w/ Copilot
  • OS: Windows 11

Ideally, I would love to test and compare how ollama performs with AMD CPU vs AMD iGPU.

I am not a gamer. I don't ever plan to buy a discrete GPU. The iGPU more than fast enough. When time comes to invest in a hardware upgrade, I will likely look at a CPU., not a dGPU.

Thank you.

Originally created by @justincranford on GitHub (Mar 11, 2025). Original GitHub issue: https://github.com/ollama/ollama/issues/9663 Originally assigned to: @dhiltgen on GitHub. # Request Could you add AMD iGPU acceleration to the roadmap? It would be great to get clarification about if and when AMD iGPU support might be coming. I found these comments in the source code that imply AMD iGPU support has been considered. https://github.com/ollama/ollama/blob/aee28501b592e2fe98863212913ffa8fb22e1ca0/discover/amd_windows.go#L124 https://github.com/ollama/ollama/blob/aee28501b592e2fe98863212913ffa8fb22e1ca0/discover/amd_linux.go#L293 I can't find any documentation under github.com/ollama/ollama to know if and when AMD iGPUs will be supported in ollama rocm. # My Use case I am running ollama latest bundled in openweb-ui latest Docker container. - https://github.com/open-webui/open-webui?tab=readme-ov-file#installing-open-webui-with-bundled-ollama-support > `docker run -d -p 3000:8080 -v ollama:/root/.ollama -v open-webui:/app/backend/data --name open-webui --restart always ghcr.io/open-webui/open-webui:ollama` My desktop for software development: - CPU: AMD 7600X CPU - RAM 2x16GB DDR5-6000 CL30 - SSD: 4TB Samsung Pro 990 PCIe 4.0 - dGPU: n/a - IDE: VScode w/ Copilot - OS: Windows 11 Ideally, I would love to test and compare how ollama performs with AMD CPU vs AMD iGPU. I am not a gamer. I don't ever plan to buy a discrete GPU. The iGPU more than fast enough. When time comes to invest in a hardware upgrade, I will likely look at a CPU., not a dGPU. Thank you.
GiteaMirror added the feature requestamd labels 2026-04-12 17:45:43 -05:00
GiteaMirror commented 2026-04-12 17:45:45 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Mar 12, 2025):

You have to pass through the GPU devices for ollama to make use of the GPU, see https://github.com/ollama/ollama/blob/main/docs/docker.md#amd-gpu.

<!-- gh-comment-id:2716013967 --> @rick-github commented on GitHub (Mar 12, 2025): You have to pass through the GPU devices for ollama to make use of the GPU, see https://github.com/ollama/ollama/blob/main/docs/docker.md#amd-gpu.
GiteaMirror commented 2026-04-12 17:45:46 -05:00
Author
Owner
Copy Link

@jhemmond commented on GitHub (Mar 12, 2025):

@justincranford This already is implemented in the AMD ROCM Fork, and my favorite AMD Vulkan Fork. Links to these, and also to the latest Vulkan binaries are here:

OllamaSetup.zip - Linux
ollama-windows-amd64.zip

Files above are for Vulkan from this:
https://github.com/whyvl/ollama-vulkan/issues/7

Ollama for AMD: https://github.com/likelovewant/ollama-for-amd/releases

That said, I would Love for the Main Ollama (this repo) to natively support iGPUs.

<!-- gh-comment-id:2716365078 --> @jhemmond commented on GitHub (Mar 12, 2025): @justincranford This already is implemented in the AMD ROCM Fork, and my favorite AMD Vulkan Fork. Links to these, and also to the latest Vulkan binaries are here: [OllamaSetup.zip - Linux](https://github.com/user-attachments/files/19150979/OllamaSetup.zip) [ollama-windows-amd64.zip](https://github.com/user-attachments/files/19150980/ollama-windows-amd64.zip) Files above are for Vulkan from this: https://github.com/whyvl/ollama-vulkan/issues/7 Ollama for AMD: https://github.com/likelovewant/ollama-for-amd/releases That said, I would Love for the Main Ollama (this repo) to natively support iGPUs.
GiteaMirror commented 2026-04-12 17:45:48 -05:00
Author
Owner
Copy Link

@zztop007 commented on GitHub (Mar 15, 2025):

@jhemmond Are you planning to support gfx1151? The latest Ollama 0.6.0 includes ROCm support for gfx1151, but so far it still gives an error. Your link to Ollama Vulkan worked perfectly! Where can one turn for a potentially updated version in the future?

<!-- gh-comment-id:2726562938 --> @zztop007 commented on GitHub (Mar 15, 2025): @jhemmond Are you planning to support gfx1151? The latest Ollama 0.6.0 includes ROCm support for gfx1151, but so far it still gives an error. Your link to Ollama Vulkan worked perfectly! Where can one turn for a potentially updated version in the future?
GiteaMirror commented 2026-04-12 17:45:50 -05:00
Author
Owner
Copy Link

@jhemmond commented on GitHub (Mar 15, 2025):

Please thank the repo owner @McBane87 who is awesome! Since Vulkan is generic APIs the gfx1151 should work automatically once AMD updates their drivers for the new iGPU. I'd check in their release notes:

https://www.amd.com/en/resources/support-articles/release-notes/RN-RAD-WIN-VULKAN.html

<!-- gh-comment-id:2726757527 --> @jhemmond commented on GitHub (Mar 15, 2025): Please thank the repo owner @McBane87 who is awesome! Since Vulkan is generic APIs the gfx1151 should work automatically once AMD updates their drivers for the new iGPU. I'd check in their release notes: https://www.amd.com/en/resources/support-articles/release-notes/RN-RAD-WIN-VULKAN.html
GiteaMirror commented 2026-04-12 17:45:52 -05:00
Author
Owner
Copy Link

@McBane87 commented on GitHub (Mar 15, 2025):

Please thank the repo owner @McBane87 who is awesome! Since Vulkan is generic APIs the gfx1151 should work automatically once AMD updates their drivers for the new iGPU. I'd check in their release notes:

https://www.amd.com/en/resources/support-articles/release-notes/RN-RAD-WIN-VULKAN.html

I'm not the repo owner, but thanks ^^
I'm just a guy like you, who found the repo, created by whyvl, and was playing around with things there :-D

<!-- gh-comment-id:2726762135 --> @McBane87 commented on GitHub (Mar 15, 2025): > Please thank the repo owner [@McBane87](https://github.com/McBane87) who is awesome! Since Vulkan is generic APIs the gfx1151 should work automatically once AMD updates their drivers for the new iGPU. I'd check in their release notes: > > https://www.amd.com/en/resources/support-articles/release-notes/RN-RAD-WIN-VULKAN.html I'm not the repo owner, but thanks ^^ I'm just a guy like you, who found the repo, created by [whyvl](https://github.com/whyvl), and was playing around with things there :-D
GiteaMirror commented 2026-04-12 17:45:53 -05:00
Author
Owner
Copy Link

@jhemmond commented on GitHub (Mar 15, 2025):

My bad, but I'm still grateful :)
We need to get that repo constantly updated with latest builds. Hope I can carve some time to start building the Windows binaries this weekend since Ollama Main is pumping releases consistently. 0.6.1 is pre-release and hope it addresses the gemma load issues.

<!-- gh-comment-id:2726764354 --> @jhemmond commented on GitHub (Mar 15, 2025): My bad, but I'm still grateful :) We need to get that repo constantly updated with latest builds. Hope I can carve some time to start building the Windows binaries this weekend since Ollama Main is pumping releases consistently. 0.6.1 is pre-release and hope it addresses the gemma load issues.
GiteaMirror commented 2026-04-12 17:45:54 -05:00
Author
Owner
Copy Link

@zztop007 commented on GitHub (Mar 17, 2025):

I would love to have a later build if and when it pops up (for gemma).

<!-- gh-comment-id:2730628260 --> @zztop007 commented on GitHub (Mar 17, 2025): I would love to have a later build if and when it pops up (for gemma).
GiteaMirror commented 2026-04-12 17:45:55 -05:00
Author
Owner
Copy Link

@McBane87 commented on GitHub (Mar 18, 2025):

I would love to have a later build if and when it pops up (for gemma).

https://github.com/whyvl/ollama-vulkan/issues/7#issue-2828064858

<!-- gh-comment-id:2731746023 --> @McBane87 commented on GitHub (Mar 18, 2025): > I would love to have a later build if and when it pops up (for gemma). https://github.com/whyvl/ollama-vulkan/issues/7#issue-2828064858
GiteaMirror commented 2026-04-12 17:45:56 -05:00
Author
Owner
Copy Link

@zztop007 commented on GitHub (Mar 18, 2025):

I would love to have a later build if and when it pops up (for gemma).

https://github.com/whyvl/ollama-vulkan/issues/7#issue-2828064858

You made my week

<!-- gh-comment-id:2731979694 --> @zztop007 commented on GitHub (Mar 18, 2025): > > I would love to have a later build if and when it pops up (for gemma). > > https://github.com/whyvl/ollama-vulkan/issues/7#issue-2828064858 You made my week
GiteaMirror commented 2026-04-12 17:45:56 -05:00
Author
Owner
Copy Link

@YuliaS11 commented on GitHub (Apr 14, 2025):

I was able to build Kobold Cpp successfully simply following the instructions and specifying the architecture to be GFX1151 using the windows HIP SDK 6.2.4.

<!-- gh-comment-id:2800596367 --> @YuliaS11 commented on GitHub (Apr 14, 2025): I was able to build Kobold Cpp successfully simply following the instructions and specifying the architecture to be GFX1151 using the windows HIP SDK 6.2.4.
GiteaMirror commented 2026-04-12 17:45:58 -05:00
Author
Owner
Copy Link

@RangerMauve commented on GitHub (Jul 3, 2025):

I was under the impression that ROCm explicitly doesn't support iGPUs. Was there some sort of fix to make this not an issue? I found that forcing iGPU usage would crash my computer after a while which isn't ideal.

<!-- gh-comment-id:3032832762 --> @RangerMauve commented on GitHub (Jul 3, 2025): I was under the impression that ROCm explicitly doesn't support iGPUs. Was there some sort of fix to make this not an issue? I found that forcing iGPU usage would crash my computer after a while which isn't ideal.
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 amd feature request
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#6304
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?