github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-08 00:51:34 -05:00
Code Issues 14.5k Packages Projects Releases 120 Wiki Activity

[GH-ISSUE #13126] Ollama cannot properly use the GPU on a dual-GPU laptop after enabling Vulkan #70746

New Issue
Open
opened 2026-05-04 22:49:14 -05:00 by GiteaMirror · 5 comments
GiteaMirror commented 2026-05-04 22:49:14 -05:00
Owner
Copy Link

Originally created by @Arondight on GitHub (Nov 18, 2025).
Original GitHub issue: https://github.com/ollama/ollama/issues/13126

Originally assigned to: @jessegross on GitHub.

What is the issue?

As shown in the picture, my laptop’s GPU0 is the Intel integrated graphics, and GPU1 is the NVIDIA MX330 graphics card, but Vulkan can only use the integrated graphics.

C:\Users\user>ollama ps
NAME        ID              SIZE      PROCESSOR    CONTEXT    UNTIL
qwen3:4b    359d7dd4bcda    3.5 GB    100% GPU     4096       2 minutes from now
Image Image

Relevant log output

OS

Windows

GPU

Nvidia

CPU

Intel

Ollama version

0.12.11

Originally created by @Arondight on GitHub (Nov 18, 2025). Original GitHub issue: https://github.com/ollama/ollama/issues/13126 Originally assigned to: @jessegross on GitHub. ### What is the issue? As shown in the picture, my laptop’s GPU0 is the Intel integrated graphics, and GPU1 is the NVIDIA MX330 graphics card, but Vulkan can only use the integrated graphics. ``` C:\Users\user>ollama ps NAME ID SIZE PROCESSOR CONTEXT UNTIL qwen3:4b 359d7dd4bcda 3.5 GB 100% GPU 4096 2 minutes from now ``` <img width="1390" height="977" alt="Image" src="https://github.com/user-attachments/assets/c53a6375-cb1e-4815-b840-093c9658a41b" /> <img width="972" height="932" alt="Image" src="https://github.com/user-attachments/assets/947285fe-a529-4076-beb7-43255f8c6cd8" /> ### Relevant log output ```shell ``` ### OS Windows ### GPU Nvidia ### CPU Intel ### Ollama version 0.12.11
GiteaMirror added the vulkanbug labels 2026-05-04 22:49:14 -05:00
GiteaMirror commented 2026-05-04 22:49:17 -05:00
Author
Owner
Copy Link

@dhiltgen commented on GitHub (Nov 18, 2025):

Please start the server with OLLAMA_DEBUG="2" set and share your startup logs so we can see what's going wrong. It should favor the discrete GPU over an iGPU.

https://github.com/ollama/ollama/blob/main/docs/troubleshooting.mdx

<!-- gh-comment-id:3549667207 --> @dhiltgen commented on GitHub (Nov 18, 2025): Please start the server with OLLAMA_DEBUG="2" set and share your startup logs so we can see what's going wrong. It should favor the discrete GPU over an iGPU. https://github.com/ollama/ollama/blob/main/docs/troubleshooting.mdx
GiteaMirror commented 2026-05-04 22:49:18 -05:00
Author
Owner
Copy Link

@Arondight commented on GitHub (Nov 19, 2025):

Please start the server with OLLAMA_DEBUG="2" set and share your startup logs so we can see what's going wrong. It should favor the discrete GPU over an iGPU.

https://github.com/ollama/ollama/blob/main/docs/troubleshooting.mdx

thanks for you reply, this is my ollama log with qwen3:1.7b.

C:\Users\user>ollama ps
NAME          ID              SIZE      PROCESSOR    CONTEXT    UNTIL
qwen3:1.7b    8f68893c685c    1.9 GB    100% GPU     4096       About a minute from now

C:\Users\user>nvidia-smi
Wed Nov 19 18:34:01 2025
+-----------------------------------------------------------------------------------------+
| NVIDIA-SMI 573.57                 Driver Version: 573.57         CUDA Version: 12.8     |
|-----------------------------------------+------------------------+----------------------+
| GPU  Name                  Driver-Model | Bus-Id          Disp.A | Volatile Uncorr. ECC |
| Fan  Temp   Perf          Pwr:Usage/Cap |           Memory-Usage | GPU-Util  Compute M. |
|                                         |                        |               MIG M. |
|=========================================+========================+======================|
|   0  NVIDIA GeForce MX330         WDDM  |   00000000:2D:00.0 Off |                  N/A |
| N/A   59C    P8            N/A  / 5001W |       0MiB /   2048MiB |      0%      Default |
|                                         |                        |                  N/A |
+-----------------------------------------+------------------------+----------------------+

+-----------------------------------------------------------------------------------------+
| Processes:                                                                              |
|  GPU   GI   CI              PID   Type   Process name                        GPU Memory |
|        ID   ID                                                               Usage      |
|=========================================================================================|
|    0   N/A  N/A            2848    C+G   ...ram Files\Tencent\QQNT\QQ.exe      N/A      |
|    0   N/A  N/A           12388    C+G   ...al\Programs\Ollama\ollama.exe      N/A      |
+-----------------------------------------------------------------------------------------+

CWINDOWSsystem32cmd.exe - ollama serve.txt

<!-- gh-comment-id:3551983732 --> @Arondight commented on GitHub (Nov 19, 2025): > Please start the server with OLLAMA_DEBUG="2" set and share your startup logs so we can see what's going wrong. It should favor the discrete GPU over an iGPU. > > https://github.com/ollama/ollama/blob/main/docs/troubleshooting.mdx thanks for you reply, this is my ollama log with `qwen3:1.7b`. ``` C:\Users\user>ollama ps NAME ID SIZE PROCESSOR CONTEXT UNTIL qwen3:1.7b 8f68893c685c 1.9 GB 100% GPU 4096 About a minute from now C:\Users\user>nvidia-smi Wed Nov 19 18:34:01 2025 +-----------------------------------------------------------------------------------------+ | NVIDIA-SMI 573.57 Driver Version: 573.57 CUDA Version: 12.8 | |-----------------------------------------+------------------------+----------------------+ | GPU Name Driver-Model | Bus-Id Disp.A | Volatile Uncorr. ECC | | Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. | | | | MIG M. | |=========================================+========================+======================| | 0 NVIDIA GeForce MX330 WDDM | 00000000:2D:00.0 Off | N/A | | N/A 59C P8 N/A / 5001W | 0MiB / 2048MiB | 0% Default | | | | N/A | +-----------------------------------------+------------------------+----------------------+ +-----------------------------------------------------------------------------------------+ | Processes: | | GPU GI CI PID Type Process name GPU Memory | | ID ID Usage | |=========================================================================================| | 0 N/A N/A 2848 C+G ...ram Files\Tencent\QQNT\QQ.exe N/A | | 0 N/A N/A 12388 C+G ...al\Programs\Ollama\ollama.exe N/A | +-----------------------------------------------------------------------------------------+ ``` [CWINDOWSsystem32cmd.exe - ollama serve.txt](https://github.com/user-attachments/files/23625341/CWINDOWSsystem32cmd.exe.-.ollama.serve.txt)
GiteaMirror commented 2026-05-04 22:49:19 -05:00
Author
Owner
Copy Link

@jessegross commented on GitHub (Nov 19, 2025):

We don't currently allow a single model to be loaded across multiple GPU libraries - CUDA and Vulkan in this case. Even though we try to favor discrete GPUs that is within a single library. In the end, the Vulkan library is favored because it allows offloading more of the model.

It's probably OK to span libraries in the case of iGPUs but we'll need to do more testing to see if that's actually a good idea.

<!-- gh-comment-id:3554792576 --> @jessegross commented on GitHub (Nov 19, 2025): We don't currently allow a single model to be loaded across multiple GPU libraries - CUDA and Vulkan in this case. Even though we try to favor discrete GPUs that is within a single library. In the end, the Vulkan library is favored because it allows offloading more of the model. It's probably OK to span libraries in the case of iGPUs but we'll need to do more testing to see if that's actually a good idea.
GiteaMirror commented 2026-05-04 22:49:20 -05:00
Author
Owner
Copy Link

@d-shehu commented on GitHub (Dec 18, 2025):

I'm also running an issue loading a larger model on 2 identical Radeon GPUs using Vulkan. Is that also not supported with Ollama?

<!-- gh-comment-id:3668491283 --> @d-shehu commented on GitHub (Dec 18, 2025): I'm also running an issue loading a larger model on 2 identical Radeon GPUs using Vulkan. Is that also not supported with Ollama?
GiteaMirror commented 2026-05-04 22:49:22 -05:00
Author
Owner
Copy Link

@getter3 commented on GitHub (Jan 10, 2026):

same here, i am using RTX 5070 8G VRAM and a integerated Radeon GPU 890M

<!-- gh-comment-id:3732135721 --> @getter3 commented on GitHub (Jan 10, 2026): same here, i am using RTX 5070 8G VRAM and a integerated Radeon GPU 890M
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 vulkan
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#70746
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?