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 #10357] Why is the CPU and GPU used in a mix instead of 100% GPU? #32564

New Issue
Closed
opened 2026-04-22 13:57:49 -05:00 by GiteaMirror · 6 comments
GiteaMirror commented 2026-04-22 13:57:49 -05:00
Owner
Copy Link

Originally created by @kirito201711 on GitHub (Apr 21, 2025).
Original GitHub issue: https://github.com/ollama/ollama/issues/10357

 PROCESSOR              
 33%/67%   CPU/GPU
Originally created by @kirito201711 on GitHub (Apr 21, 2025). Original GitHub issue: https://github.com/ollama/ollama/issues/10357 PROCESSOR 33%/67% CPU/GPU
GiteaMirror commented 2026-04-22 13:57:52 -05:00
Author
Owner
Copy Link

@kth8 commented on GitHub (Apr 21, 2025):

the model size exceeds your GPU VRAM capacity so it has to load part of it on to the CPU/system RAM

<!-- gh-comment-id:2818033465 --> @kth8 commented on GitHub (Apr 21, 2025): the model size exceeds your GPU VRAM capacity so it has to load part of it on to the CPU/system RAM
GiteaMirror commented 2026-04-22 13:57:53 -05:00
Author
Owner
Copy Link

@pdevine commented on GitHub (Apr 21, 2025):

Going to go ahead and close this as answered. @kirito201711 unfortunately your GPU doesn't have enough VRAM to hold the model you're trying to run, so it starts using some of your system memory. You can try using a different model, or one which is more quantized.

<!-- gh-comment-id:2819571606 --> @pdevine commented on GitHub (Apr 21, 2025): Going to go ahead and close this as answered. @kirito201711 unfortunately your GPU doesn't have enough VRAM to hold the model you're trying to run, so it starts using some of your system memory. You can try using a different model, or one which is more quantized.
GiteaMirror commented 2026-04-22 13:57:54 -05:00
Author
Owner
Copy Link

@kirito201711 commented on GitHub (Apr 22, 2025):

My VRAM is only being used at less than half

the model size exceeds your GPU VRAM capacity so it has to load part of it on to the CPU/system RAM

<!-- gh-comment-id:2821359523 --> @kirito201711 commented on GitHub (Apr 22, 2025): My VRAM is only being used at less than half > the model size exceeds your GPU VRAM capacity so it has to load part of it on to the CPU/system RAM
GiteaMirror commented 2026-04-22 13:57:55 -05:00
Author
Owner
Copy Link

@kirito201711 commented on GitHub (Apr 22, 2025):

Going to go ahead and close this as answered. @kirito201711 unfortunately your GPU doesn't have enough VRAM to hold the model you're trying to run, so it starts using some of your system memory. You can try using a different model, or one which is more quantized.

My VRAM is only being used at less than half

<!-- gh-comment-id:2821360189 --> @kirito201711 commented on GitHub (Apr 22, 2025): > Going to go ahead and close this as answered. [@kirito201711](https://github.com/kirito201711) unfortunately your GPU doesn't have enough VRAM to hold the model you're trying to run, so it starts using some of your system memory. You can try using a different model, or one which is more quantized. My VRAM is only being used at less than half
GiteaMirror commented 2026-04-22 13:57:56 -05:00
Author
Owner
Copy Link

@pdevine commented on GitHub (Apr 22, 2025):

@kirito201711 each layer (which includes multiple tensors) has to be offloaded into system memory, so if you don't have a lot of VRAM, it's going to look like it's not using enough of your GPU. You didn't mention which model you were trying to load or the GPU you're using though.

<!-- gh-comment-id:2821793871 --> @pdevine commented on GitHub (Apr 22, 2025): @kirito201711 each layer (which includes multiple tensors) has to be offloaded into system memory, so if you don't have a lot of VRAM, it's going to look like it's not using enough of your GPU. You didn't mention which model you were trying to load or the GPU you're using though.
GiteaMirror commented 2026-04-22 13:57:57 -05:00
Author
Owner
Copy Link

@kirito201711 commented on GitHub (Apr 24, 2025):

@kirito201711 each layer (which includes multiple tensors) has to be offloaded into system memory, so if you don't have a lot of VRAM, it's going to look like it's not using enough of your GPU. You didn't mention which model you were trying to load or the GPU you're using though.

I found the problem. The gemma3:12b model I was running was 8.1GB, and my video memory was 8GB.

<!-- gh-comment-id:2827426832 --> @kirito201711 commented on GitHub (Apr 24, 2025): > [@kirito201711](https://github.com/kirito201711) each layer (which includes multiple tensors) has to be offloaded into system memory, so if you don't have a lot of VRAM, it's going to look like it's not using enough of your GPU. You didn't mention which model you were trying to load or the GPU you're using though. I found the problem. The gemma3:12b model I was running was 8.1GB, and my video memory was 8GB.
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
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#32564
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?