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 #1656] Try to use only 1 GPU if possible #26689

New Issue
Closed
opened 2026-04-22 03:07:17 -05:00 by GiteaMirror · 1 comment
GiteaMirror commented 2026-04-22 03:07:17 -05:00
Owner
Copy Link

Originally created by @m0wer on GitHub (Dec 21, 2023).
Original GitHub issue: https://github.com/ollama/ollama/issues/1656

Originally assigned to: @dhiltgen on GitHub.

I have two GPUs: an RTX 3090 and a 2060. ollama tries to split the load among both, but due to communication speed between the cards being significantly slower than within a GPU, this is counterproductive. Would be great to prioritize using the first GPU only, if it alone has enough VRAM to fit the model.

Here are the results of a small experiment:

Model: solar:q4_0

  • RTX 3090 alone:
    • 350 W: 81 t/s
    • 280W: 79 t/s
  • RTX 3090 (280/350 W) and RTX 2060 (128/160 W):
    • 31.5 t/s

To do the test I just changed the value of CUDA_VISIBLE_DEVICES from 0,1 to 0.

Originally created by @m0wer on GitHub (Dec 21, 2023). Original GitHub issue: https://github.com/ollama/ollama/issues/1656 Originally assigned to: @dhiltgen on GitHub. I have two GPUs: an RTX 3090 and a 2060. `ollama` tries to split the load among both, but due to communication speed between the cards being significantly slower than within a GPU, this is counterproductive. Would be great to prioritize using the first GPU only, if it alone has enough VRAM to fit the model. Here are the results of a small experiment: Model: `solar:q4_0` - RTX 3090 alone: - 350 W: 81 t/s - 280W: 79 t/s - RTX 3090 (280/350 W) and RTX 2060 (128/160 W): - 31.5 t/s To do the test I just changed the value of `CUDA_VISIBLE_DEVICES` from `0,1` to `0`.
GiteaMirror added the gpu label 2026-04-22 03:07:17 -05:00
GiteaMirror commented 2026-04-22 03:07:18 -05:00
Author
Owner
Copy Link

@divinity76 commented on GitHub (Feb 11, 2025):

i think that's what

OLLAMA_SCHED_SPREAD=0 ollama serve

is for - i'm guessing OLLAMA_SCHED_SPREAD didn't exist yet when you created this issue

<!-- gh-comment-id:2652158389 --> @divinity76 commented on GitHub (Feb 11, 2025): i think that's what ``` OLLAMA_SCHED_SPREAD=0 ollama serve ``` is for - i'm guessing OLLAMA_SCHED_SPREAD didn't exist yet when you created this issue
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 gpu
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#26689
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?