github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-07-17 13:47:08 -05:00
Code Issues 16.9k Packages Projects Releases 122 Wiki Activity

[GH-ISSUE #11734] New ollama not using gpu #69831

New Issue
Closed
opened 2026-05-04 19:31:11 -05:00 by GiteaMirror · 11 comments
GiteaMirror commented 2026-05-04 19:31:11 -05:00
Owner
Copy Link

Originally created by @arpondark on GitHub (Aug 6, 2025).
Original GitHub issue: https://github.com/ollama/ollama/issues/11734

What is the issue?

Image not using nvdia gpu . using cpu only for every model . without qwen2 model .

Relevant log output

time=2025-08-06T15:57:08.463+06:00 level=INFO source=routes.go:1297 msg="server config" env="map[CUDA_VISIBLE_DEVICES: GPU_DEVICE_ORDINAL: HIP_VISIBLE_DEVICES: HSA_OVERRIDE_GFX_VERSION: HTTPS_PROXY: HTTP_PROXY: NO_PROXY: OLLAMA_CONTEXT_LENGTH:4096 OLLAMA_DEBUG:INFO OLLAMA_FLASH_ATTENTION:false OLLAMA_GPU_OVERHEAD:0 OLLAMA_HOST:http://127.0.0.1:11434 OLLAMA_INTEL_GPU:false OLLAMA_KEEP_ALIVE:5m0s OLLAMA_KV_CACHE_TYPE: OLLAMA_LLM_LIBRARY: OLLAMA_LOAD_TIMEOUT:5m0s OLLAMA_MAX_LOADED_MODELS:0 OLLAMA_MAX_QUEUE:512 OLLAMA_MODELS:C:\\Users\\Administrator\\.ollama\\models OLLAMA_MULTIUSER_CACHE:false OLLAMA_NEW_ENGINE:false OLLAMA_NOHISTORY:false OLLAMA_NOPRUNE:false OLLAMA_NUM_PARALLEL:1 OLLAMA_ORIGINS:[http://localhost https://localhost http://localhost:* https://localhost:* http://127.0.0.1 https://127.0.0.1 http://127.0.0.1:* https://127.0.0.1:* http://0.0.0.0 https://0.0.0.0 http://0.0.0.0:* https://0.0.0.0:* app://* file://* tauri://* vscode-webview://* vscode-file://*] OLLAMA_SCHED_SPREAD:false ROCR_VISIBLE_DEVICES:]"
time=2025-08-06T15:57:08.463+06:00 level=INFO source=images.go:477 msg="total blobs: 0"
time=2025-08-06T15:57:08.463+06:00 level=INFO source=images.go:484 msg="total unused blobs removed: 0"
time=2025-08-06T15:57:08.464+06:00 level=INFO source=routes.go:1350 msg="Listening on 127.0.0.1:11434 (version 0.11.2)"
time=2025-08-06T15:57:08.464+06:00 level=INFO source=gpu.go:217 msg="looking for compatible GPUs"
time=2025-08-06T15:57:08.464+06:00 level=INFO source=gpu_windows.go:167 msg=packages count=1
time=2025-08-06T15:57:08.464+06:00 level=INFO source=gpu_windows.go:183 msg="efficiency cores detected" maxEfficiencyClass=1
time=2025-08-06T15:57:08.464+06:00 level=INFO source=gpu_windows.go:214 msg="" package=0 cores=8 efficiency=4 threads=12
time=2025-08-06T15:57:08.600+06:00 level=INFO source=gpu.go:319 msg="detected OS VRAM overhead" id=GPU-7f06c888-ceb6-9893-d38d-a08e0ffd1954 library=cuda compute=8.9 driver=13.0 name="NVIDIA GeForce RTX 4050 Laptop GPU" overhead="582.2 MiB"
time=2025-08-06T15:57:08.602+06:00 level=INFO source=types.go:130 msg="inference compute" id=GPU-7f06c888-ceb6-9893-d38d-a08e0ffd1954 library=cuda variant=v12 compute=8.9 driver=13.0 name="NVIDIA GeForce RTX 4050 Laptop GPU" total="6.0 GiB" available="5.0 GiB"

OS

No response

GPU

No response

CPU

No response

Ollama version

No response

Originally created by @arpondark on GitHub (Aug 6, 2025). Original GitHub issue: https://github.com/ollama/ollama/issues/11734 ### What is the issue? <img width="1627" height="1018" alt="Image" src="https://github.com/user-attachments/assets/6ed1f22e-4f39-4210-ae19-70f203f5121d" /> not using nvdia gpu . using cpu only for every model . without qwen2 model . ### Relevant log output ```shell time=2025-08-06T15:57:08.463+06:00 level=INFO source=routes.go:1297 msg="server config" env="map[CUDA_VISIBLE_DEVICES: GPU_DEVICE_ORDINAL: HIP_VISIBLE_DEVICES: HSA_OVERRIDE_GFX_VERSION: HTTPS_PROXY: HTTP_PROXY: NO_PROXY: OLLAMA_CONTEXT_LENGTH:4096 OLLAMA_DEBUG:INFO OLLAMA_FLASH_ATTENTION:false OLLAMA_GPU_OVERHEAD:0 OLLAMA_HOST:http://127.0.0.1:11434 OLLAMA_INTEL_GPU:false OLLAMA_KEEP_ALIVE:5m0s OLLAMA_KV_CACHE_TYPE: OLLAMA_LLM_LIBRARY: OLLAMA_LOAD_TIMEOUT:5m0s OLLAMA_MAX_LOADED_MODELS:0 OLLAMA_MAX_QUEUE:512 OLLAMA_MODELS:C:\\Users\\Administrator\\.ollama\\models OLLAMA_MULTIUSER_CACHE:false OLLAMA_NEW_ENGINE:false OLLAMA_NOHISTORY:false OLLAMA_NOPRUNE:false OLLAMA_NUM_PARALLEL:1 OLLAMA_ORIGINS:[http://localhost https://localhost http://localhost:* https://localhost:* http://127.0.0.1 https://127.0.0.1 http://127.0.0.1:* https://127.0.0.1:* http://0.0.0.0 https://0.0.0.0 http://0.0.0.0:* https://0.0.0.0:* app://* file://* tauri://* vscode-webview://* vscode-file://*] OLLAMA_SCHED_SPREAD:false ROCR_VISIBLE_DEVICES:]" time=2025-08-06T15:57:08.463+06:00 level=INFO source=images.go:477 msg="total blobs: 0" time=2025-08-06T15:57:08.463+06:00 level=INFO source=images.go:484 msg="total unused blobs removed: 0" time=2025-08-06T15:57:08.464+06:00 level=INFO source=routes.go:1350 msg="Listening on 127.0.0.1:11434 (version 0.11.2)" time=2025-08-06T15:57:08.464+06:00 level=INFO source=gpu.go:217 msg="looking for compatible GPUs" time=2025-08-06T15:57:08.464+06:00 level=INFO source=gpu_windows.go:167 msg=packages count=1 time=2025-08-06T15:57:08.464+06:00 level=INFO source=gpu_windows.go:183 msg="efficiency cores detected" maxEfficiencyClass=1 time=2025-08-06T15:57:08.464+06:00 level=INFO source=gpu_windows.go:214 msg="" package=0 cores=8 efficiency=4 threads=12 time=2025-08-06T15:57:08.600+06:00 level=INFO source=gpu.go:319 msg="detected OS VRAM overhead" id=GPU-7f06c888-ceb6-9893-d38d-a08e0ffd1954 library=cuda compute=8.9 driver=13.0 name="NVIDIA GeForce RTX 4050 Laptop GPU" overhead="582.2 MiB" time=2025-08-06T15:57:08.602+06:00 level=INFO source=types.go:130 msg="inference compute" id=GPU-7f06c888-ceb6-9893-d38d-a08e0ffd1954 library=cuda variant=v12 compute=8.9 driver=13.0 name="NVIDIA GeForce RTX 4050 Laptop GPU" total="6.0 GiB" available="5.0 GiB" ``` ### OS _No response_ ### GPU _No response_ ### CPU _No response_ ### Ollama version _No response_
GiteaMirror added the bug label 2026-05-04 19:31:11 -05:00
GiteaMirror commented 2026-05-04 19:31:15 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Aug 6, 2025):

Add the full server log.

<!-- gh-comment-id:3159948950 --> @rick-github commented on GitHub (Aug 6, 2025): Add the full server log.
GiteaMirror commented 2026-05-04 19:31:16 -05:00
Author
Owner
Copy Link

@arpondark commented on GitHub (Aug 6, 2025):

app-2.log
app-3.log
app-4.log
app-5.log
server.log
server-1.log
server-2.log
server-3.log
server-4.log
server-5.log
app.log
app-1.log

<!-- gh-comment-id:3159971830 --> @arpondark commented on GitHub (Aug 6, 2025): [app-2.log](https://github.com/user-attachments/files/21618032/app-2.log) [app-3.log](https://github.com/user-attachments/files/21618033/app-3.log) [app-4.log](https://github.com/user-attachments/files/21618034/app-4.log) [app-5.log](https://github.com/user-attachments/files/21618037/app-5.log) [server.log](https://github.com/user-attachments/files/21618038/server.log) [server-1.log](https://github.com/user-attachments/files/21618043/server-1.log) [server-2.log](https://github.com/user-attachments/files/21618035/server-2.log) [server-3.log](https://github.com/user-attachments/files/21618040/server-3.log) [server-4.log](https://github.com/user-attachments/files/21618042/server-4.log) [server-5.log](https://github.com/user-attachments/files/21618039/server-5.log) [app.log](https://github.com/user-attachments/files/21618036/app.log) [app-1.log](https://github.com/user-attachments/files/21618041/app-1.log)
GiteaMirror commented 2026-05-04 19:31:18 -05:00
Author
Owner
Copy Link

@davelindevelt commented on GitHub (Aug 6, 2025):

My experience with the latest version: Only gpt-oss model (20b) not offloading to GPU, runs on CPU. But other models use gpu without any issue.

<!-- gh-comment-id:3160387829 --> @davelindevelt commented on GitHub (Aug 6, 2025): My experience with the latest version: Only gpt-oss model (20b) not offloading to GPU, runs on CPU. But other models use gpu without any issue.
GiteaMirror commented 2026-05-04 19:31:19 -05:00
Author
Owner
Copy Link

@y-tor commented on GitHub (Aug 6, 2025):

Yes, the same behaviour for me

My experience with the latest version: Only gpt-oss model (20b) not offloading to GPU, runs on CPU. But other models use gpu without any issue.

<!-- gh-comment-id:3160456685 --> @y-tor commented on GitHub (Aug 6, 2025): Yes, the same behaviour for me > My experience with the latest version: Only gpt-oss model (20b) not offloading to GPU, runs on CPU. But other models use gpu without any issue.
GiteaMirror commented 2026-05-04 19:31:20 -05:00
Author
Owner
Copy Link

@arpondark commented on GitHub (Aug 6, 2025):

but in my case only qwen is runnng on gpu llam3 gemma and other not running

<!-- gh-comment-id:3160639138 --> @arpondark commented on GitHub (Aug 6, 2025): but in my case only qwen is runnng on gpu llam3 gemma and other not running
GiteaMirror commented 2026-05-04 19:31:22 -05:00
Author
Owner
Copy Link

@rnagabhyrava commented on GitHub (Aug 6, 2025):

My experience with the latest version: Only gpt-oss model (20b) not offloading to GPU, runs on CPU. But other models use gpu without any issue.

Same here. Gemma 3 works fine with GPU but gpt-oss is not using GPU at all (6900xt) on v0.11.0/2/3

<!-- gh-comment-id:3161094843 --> @rnagabhyrava commented on GitHub (Aug 6, 2025): > My experience with the latest version: Only gpt-oss model (20b) not offloading to GPU, runs on CPU. But other models use gpu without any issue. Same here. Gemma 3 works fine with GPU but gpt-oss is not using GPU at all (6900xt) on v0.11.0/2/3
GiteaMirror commented 2026-05-04 19:31:24 -05:00
Author
Owner
Copy Link

@AncientMystic commented on GitHub (Aug 6, 2025):

I am having the same issue, except with all models, i tried quite a few and ollama is having issues allocating vram, saying 0b available, saying my gpu doesn't support flash attention, etc

I had to downgrade to 0.10.1, 11+ just drops my gpu basically.

(Pascal gen Compute compatibility 6.1 based card)

<!-- gh-comment-id:3161100521 --> @AncientMystic commented on GitHub (Aug 6, 2025): I am having the same issue, except with all models, i tried quite a few and ollama is having issues allocating vram, saying 0b available, saying my gpu doesn't support flash attention, etc I had to downgrade to 0.10.1, 11+ just drops my gpu basically. (Pascal gen Compute compatibility 6.1 based card)
GiteaMirror commented 2026-05-04 19:31:26 -05:00
Author
Owner
Copy Link

@botanegg commented on GitHub (Aug 6, 2025):

https://github.com/ollama/ollama/pull/11700

<!-- gh-comment-id:3161356761 --> @botanegg commented on GitHub (Aug 6, 2025): https://github.com/ollama/ollama/pull/11700
GiteaMirror commented 2026-05-04 19:31:27 -05:00
Author
Owner
Copy Link

@JanJohansen commented on GitHub (Aug 6, 2025):

Same issue here (on RTX4090). Former version of Ollama worked fine!
Both "Game ready driver" and "Studio Driver" tested. Same issue... So should be Ollama that has the problem.

<!-- gh-comment-id:3161498095 --> @JanJohansen commented on GitHub (Aug 6, 2025): Same issue here (on RTX4090). Former version of Ollama worked fine! Both "Game ready driver" and "Studio Driver" tested. Same issue... So should be Ollama that has the problem.
GiteaMirror commented 2026-05-04 19:31:33 -05:00
Author
Owner
Copy Link

@jessegross commented on GitHub (Aug 6, 2025):

There was a bug in 0.11.2 and below where the memory estimation would become too high for gpt-oss if the model needed to be split across GPU and CPU or multiple GPUs. This would often cause 100% on CPU usage once the model overflows a single GPU.

This is fixed in 0.11.3.

<!-- gh-comment-id:3161755396 --> @jessegross commented on GitHub (Aug 6, 2025): There was a bug in 0.11.2 and below where the memory estimation would become too high for gpt-oss if the model needed to be split across GPU and CPU or multiple GPUs. This would often cause 100% on CPU usage once the model overflows a single GPU. This is fixed in 0.11.3.
GiteaMirror commented 2026-05-04 19:31:38 -05:00
Author
Owner
Copy Link

@cwiggi01 commented on GitHub (Aug 6, 2025):

I still have problems, GPT-OSS is pretty bad on my machine compared to LM studio, either runs slowly or not at all, and in LM Studio I get like 150 tokens per second with max context window.

<!-- gh-comment-id:3161926895 --> @cwiggi01 commented on GitHub (Aug 6, 2025): I still have problems, GPT-OSS is pretty bad on my machine compared to LM studio, either runs slowly or not at all, and in LM Studio I get like 150 tokens per second with max context window.
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#69831
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?