github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-08 17:49:24 -05:00
Code Issues 14.5k Packages Projects Releases 122 Wiki Activity

[GH-ISSUE #15055] 0.18.x idle VRAM usage and power consumption #9662

New Issue
Open
opened 2026-04-12 22:33:09 -05:00 by GiteaMirror · 2 comments
GiteaMirror commented 2026-04-12 22:33:09 -05:00
Owner
Copy Link

Originally created by @fluxlinkage on GitHub (Mar 25, 2026).
Original GitHub issue: https://github.com/ollama/ollama/issues/15055

What is the issue?

I was using Ollama 0.17.7 under Windows 11 and everything is fine.
However, after I updated to 0.18.2, my fans become noisy even if idle.
The output of nvidia-smi shows that a ollama process is using 262MB VRAM, even if ollama is idle (Not running any models, only system tray icon).

Wed Mar 25 10:32:19 2026
+-----------------------------------------------------------------------------------------+
| NVIDIA-SMI 591.59                 Driver Version: 591.59         CUDA Version: 13.1     |
+-----------------------------------------+------------------------+----------------------+
| 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 A40                   TCC   |   00000000:01:00.0 Off |                  Off |
|  0%   58C    P0             85W /  300W |     272MiB /  49140MiB |      0%      Default |
|                                         |                        |                  N/A |
+-----------------------------------------+------------------------+----------------------+

+-----------------------------------------------------------------------------------------+
| Processes:                                                                              |
|  GPU   GI   CI              PID   Type   Process name                        GPU Memory |
|        ID   ID                                                               Usage      |
|=========================================================================================|
|    0   N/A  N/A           15276      C   ...al\Programs\Ollama\ollama.exe        262MiB |
+-----------------------------------------------------------------------------------------+

At the same time, ollama ps says no model is running.

Downgrading to 0.18.0, the same problem.

Downgrading to 0.17.7, everything is OK again. The output of nvidia-smi is normal.

Wed Mar 25 11:02:20 2026
+-----------------------------------------------------------------------------------------+
| NVIDIA-SMI 591.59                 Driver Version: 591.59         CUDA Version: 13.1     |
+-----------------------------------------+------------------------+----------------------+
| 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 A40                   TCC   |   00000000:01:00.0 Off |                  Off |
|  0%   45C    P8             14W /  300W |      10MiB /  49140MiB |      0%      Default |
|                                         |                        |                  N/A |
+-----------------------------------------+------------------------+----------------------+

+-----------------------------------------------------------------------------------------+
| Processes:                                                                              |
|  GPU   GI   CI              PID   Type   Process name                        GPU Memory |
|        ID   ID                                                               Usage      |
|=========================================================================================|
|  No running processes found                                                             |
+-----------------------------------------------------------------------------------------+

So why are the VRAM used in 0.18.x when idle? Is this a new feature (if yes, can I manually turn it off?) or just a bug? I can't accept 70 Watts additional idle power!

Relevant log output

I don't know whether it is relevant, but the following error only exists in 0.18.x log (about 3 seconds after server start, reproducible). In 0.17.7, on such error.

Error #01: write tcp 127.0.0.1:11434->127.0.0.1:54305: wsasend: An established connection was aborted by the software in your host machine.

OS

Windows

GPU

Nvidia

CPU

AMD

Ollama version

0.18.2

Originally created by @fluxlinkage on GitHub (Mar 25, 2026). Original GitHub issue: https://github.com/ollama/ollama/issues/15055 ### What is the issue? I was using Ollama 0.17.7 under Windows 11 and everything is fine. However, after I updated to 0.18.2, my fans become noisy even if idle. The output of `nvidia-smi` shows that a ollama process is using 262MB VRAM, even if ollama is idle (Not running any models, only system tray icon). ``` Wed Mar 25 10:32:19 2026 +-----------------------------------------------------------------------------------------+ | NVIDIA-SMI 591.59 Driver Version: 591.59 CUDA Version: 13.1 | +-----------------------------------------+------------------------+----------------------+ | 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 A40 TCC | 00000000:01:00.0 Off | Off | | 0% 58C P0 85W / 300W | 272MiB / 49140MiB | 0% Default | | | | N/A | +-----------------------------------------+------------------------+----------------------+ +-----------------------------------------------------------------------------------------+ | Processes: | | GPU GI CI PID Type Process name GPU Memory | | ID ID Usage | |=========================================================================================| | 0 N/A N/A 15276 C ...al\Programs\Ollama\ollama.exe 262MiB | +-----------------------------------------------------------------------------------------+ ``` At the same time, `ollama ps` says no model is running. Downgrading to 0.18.0, the same problem. Downgrading to 0.17.7, everything is OK again. The output of `nvidia-smi` is normal. ``` Wed Mar 25 11:02:20 2026 +-----------------------------------------------------------------------------------------+ | NVIDIA-SMI 591.59 Driver Version: 591.59 CUDA Version: 13.1 | +-----------------------------------------+------------------------+----------------------+ | 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 A40 TCC | 00000000:01:00.0 Off | Off | | 0% 45C P8 14W / 300W | 10MiB / 49140MiB | 0% Default | | | | N/A | +-----------------------------------------+------------------------+----------------------+ +-----------------------------------------------------------------------------------------+ | Processes: | | GPU GI CI PID Type Process name GPU Memory | | ID ID Usage | |=========================================================================================| | No running processes found | +-----------------------------------------------------------------------------------------+ ``` So why are the VRAM used in 0.18.x when idle? Is this a new feature (if yes, can I manually turn it off?) or just a bug? I can't accept 70 Watts additional idle power! ### Relevant log output ```shell I don't know whether it is relevant, but the following error only exists in 0.18.x log (about 3 seconds after server start, reproducible). In 0.17.7, on such error. Error #01: write tcp 127.0.0.1:11434->127.0.0.1:54305: wsasend: An established connection was aborted by the software in your host machine. ``` ### OS Windows ### GPU Nvidia ### CPU AMD ### Ollama version 0.18.2
GiteaMirror added the bug label 2026-04-12 22:33:09 -05:00
GiteaMirror commented 2026-04-12 22:33:11 -05:00
Author
Owner
Copy Link

@solidaxelproject commented on GitHub (Mar 25, 2026):

Same root cause — the runner process in 0.18.x doesn't release memory properly. In my case it manifests as progressive 404 errors under sustained load, with runner RSS growing ~2 MB per request. See full benchmark data and reproduction script here:
https://github.com/ollama/ollama/issues/15027

<!-- gh-comment-id:4126016442 --> @solidaxelproject commented on GitHub (Mar 25, 2026): Same root cause — the runner process in 0.18.x doesn't release memory properly. In my case it manifests as progressive 404 errors under sustained load, with runner RSS growing ~2 MB per request. See full benchmark data and reproduction script here: [https://github.com/ollama/ollama/issues/15027](https://github.com/ollama/ollama/issues/15027)
GiteaMirror commented 2026-04-12 22:33:11 -05:00
Author
Owner
Copy Link

@fluxlinkage commented on GitHub (Mar 31, 2026):

Updated to 0.19.0, still same issue.
I think it is related to MLX. So I deleted mlx_cuda_v13 folder under %USERPROFILE%\AppData\Local\Programs\Ollama\lib\ollama, and everything is OK (fallback to cuda beckend)!
Why use MLX on Windows?

<!-- gh-comment-id:4159452008 --> @fluxlinkage commented on GitHub (Mar 31, 2026): Updated to 0.19.0, still same issue. I think it is related to MLX. So I deleted `mlx_cuda_v13` folder under `%USERPROFILE%\AppData\Local\Programs\Ollama\lib\ollama`, and everything is OK (fallback to cuda beckend)! Why use MLX on Windows?
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#9662
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?