github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-06 16:11:34 -05:00
Code Issues 14.5k Packages Projects Releases 120 Wiki Activity

[GH-ISSUE #12573] Rocm with 9070 xt windows #8341

New Issue
Open
opened 2026-04-12 20:55:42 -05:00 by GiteaMirror · 5 comments
GiteaMirror commented 2026-04-12 20:55:42 -05:00
Owner
Copy Link

Originally created by @Nfams on GitHub (Oct 11, 2025).
Original GitHub issue: https://github.com/ollama/ollama/issues/12573

What is the issue?

I had used this up until i guess 12.3 or so with the manual rocm rocblas files copied over. Sometimes i think i had to manually copy a rocblas file too. Any help on this later version? can't seem to get it fully working? In task manager it gets full vram usage with a 8B up to 21b model and some of the ways i've tried gets compute usage but in the end of the output it uses cpu and the tokens are around 8-16 vs the 45 i was getting

Relevant log output

OS

Windows

GPU

AMD

CPU

AMD

Ollama version

12.3-12.5 same issues

Originally created by @Nfams on GitHub (Oct 11, 2025). Original GitHub issue: https://github.com/ollama/ollama/issues/12573 ### What is the issue? I had used this up until i guess 12.3 or so with the manual rocm rocblas files copied over. Sometimes i think i had to manually copy a rocblas file too. Any help on this later version? can't seem to get it fully working? In task manager it gets full vram usage with a 8B up to 21b model and some of the ways i've tried gets compute usage but in the end of the output it uses cpu and the tokens are around 8-16 vs the 45 i was getting ### Relevant log output ```shell ``` ### OS Windows ### GPU AMD ### CPU AMD ### Ollama version 12.3-12.5 same issues
GiteaMirror added the bug label 2026-04-12 20:55:42 -05:00
GiteaMirror commented 2026-04-12 20:55:44 -05:00
Author
Owner
Copy Link

@Nfams commented on GitHub (Oct 11, 2025):

server.log

<!-- gh-comment-id:3393212848 --> @Nfams commented on GitHub (Oct 11, 2025): [server.log](https://github.com/user-attachments/files/22863613/server.log)
GiteaMirror commented 2026-04-12 20:55:45 -05:00
Author
Owner
Copy Link

@Nfams commented on GitHub (Oct 12, 2025):

okay i changed back to 12.3 and using the oss model (not my favorite but i know some work has been done) it gets 80t/s so it's clearly working. Hmm. After quite a bit of testing it seems 12.4 onwards broke this for me. with rocm 6.4. Different models load and i can tell what has been optimised. Some still use vram but cpu does the work and it takes a while to process.

<!-- gh-comment-id:3393801238 --> @Nfams commented on GitHub (Oct 12, 2025): okay i changed back to 12.3 and using the oss model (not my favorite but i know some work has been done) it gets 80t/s so it's clearly working. Hmm. After quite a bit of testing it seems 12.4 onwards broke this for me. with rocm 6.4. Different models load and i can tell what has been optimised. Some still use vram but cpu does the work and it takes a while to process.
GiteaMirror commented 2026-04-12 20:55:45 -05:00
Author
Owner
Copy Link

@Nfams commented on GitHub (Oct 12, 2025):

this is the rocblas that works for 12.3 and earlier for me

rocblas.zip

<!-- gh-comment-id:3393901485 --> @Nfams commented on GitHub (Oct 12, 2025): this is the rocblas that works for 12.3 and earlier for me [rocblas.zip](https://github.com/user-attachments/files/22868891/rocblas.zip)
GiteaMirror commented 2026-04-12 20:55:47 -05:00
Author
Owner
Copy Link

@dhiltgen commented on GitHub (Nov 14, 2025):

In 0.12.11 Vulkan is now included in the official binaries, but still experimental. To enable, set OLLAMA_VULKAN=1 for the server. https://github.com/ollama/ollama/blob/main/docs/faq.mdx#how-do-i-configure-ollama-server

<!-- gh-comment-id:3530391172 --> @dhiltgen commented on GitHub (Nov 14, 2025): In 0.12.11 Vulkan is now included in the official binaries, but still experimental. To enable, set OLLAMA_VULKAN=1 for the server. https://github.com/ollama/ollama/blob/main/docs/faq.mdx#how-do-i-configure-ollama-server
GiteaMirror commented 2026-04-12 20:55:49 -05:00
Author
Owner
Copy Link

@rafrafek commented on GitHub (Jan 22, 2026):

Fun fact, Ollama doesn't support the RX 9070 XT, but AMD doesn’t seem to know that. Their latest GPU driver installs an "AI Bundle" that includes Ollama:

https://www.amd.com/en/blogs/2026/amd-software-adrenalin-edition-ai-bundle-ai-made-si.html

On top of that, they're removing ROCm support on Windows for many GPUs - like RX 7800 XT and RX 7700 XT - without prior notice:

https://rocm.docs.amd.com/projects/radeon-ryzen/en/latest/docs/compatibility/compatibilityrad/windows/windows_compatibility.html

This makes me rethink my decision to buy the RX 7800 XT and RX 9070 XT. I probably should have bought Nvidia instead.

<!-- gh-comment-id:3784027487 --> @rafrafek commented on GitHub (Jan 22, 2026): Fun fact, Ollama doesn't support the RX 9070 XT, but AMD doesn’t seem to know that. Their latest GPU driver installs an "AI Bundle" that includes Ollama: https://www.amd.com/en/blogs/2026/amd-software-adrenalin-edition-ai-bundle-ai-made-si.html On top of that, they're removing ROCm support on Windows for many GPUs - like RX 7800 XT and RX 7700 XT - without prior notice: https://rocm.docs.amd.com/projects/radeon-ryzen/en/latest/docs/compatibility/compatibilityrad/windows/windows_compatibility.html This makes me rethink my decision to buy the RX 7800 XT and RX 9070 XT. I probably should have bought Nvidia instead.
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#8341
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?