github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-07-15 12:48:05 -05:00
Code Issues 16.9k Packages Projects Releases 122 Wiki Activity

[GH-ISSUE #11919] Ollama 0.11.5-RC2: GPT-OSS-120b - FA + KV Cache -poor GPU utilization - very slow #7911

New Issue
Closed
opened 2026-04-12 20:04:39 -05:00 by GiteaMirror · 2 comments
GiteaMirror commented 2026-04-12 20:04:39 -05:00
Owner
Copy Link

Originally created by @maglat on GitHub (Aug 15, 2025).
Original GitHub issue: https://github.com/ollama/ollama/issues/11919

OLLAMA: 0.11.5-RC2

Hardware: 4x RTX 3090

OS: Ubuntu 22.04

Model: GPT-OSS-120b
Context: 28k
Env:

  • Environment="OLLAMA_FLASH_ATTENTION=1"
  • Environment="OLLAMA_KV_CACHE_TYPE=q8_0"

Issue:

As soon FA + KV Cache is activated, the response time dramatically get slower. from arround 75t/s to 19t/s.
Checking nvidia-smi during the process, the gpu utilization is very slow. the power of the GPUs are not used entirely.

Originally created by @maglat on GitHub (Aug 15, 2025). Original GitHub issue: https://github.com/ollama/ollama/issues/11919 OLLAMA: 0.11.5-RC2 Hardware: 4x RTX 3090 OS: Ubuntu 22.04 Model: GPT-OSS-120b Context: 28k Env: - Environment="OLLAMA_FLASH_ATTENTION=1" - Environment="OLLAMA_KV_CACHE_TYPE=q8_0" Issue: As soon FA + KV Cache is activated, the response time dramatically get slower. from arround 75t/s to 19t/s. Checking nvidia-smi during the process, the gpu utilization is very slow. the power of the GPUs are not used entirely.
GiteaMirror commented 2026-04-12 20:04:41 -05:00
Author
Owner
Copy Link

@DXXS commented on GitHub (Aug 20, 2025):

FWIW, For some reason, on M1 MacBook, the latest 0.11.5 was initially running faster for me*, (w/logs claiming as usual 'low vram mode'), but when I lowered the cache from 128k to 8k, it suddenly allocated significantly more memory to ollama, yet appeared to run much slower!?

[Also at that point, trying turning on OLLAMA_NEW_ESTIMATES cut the speed in ~half]

<!-- gh-comment-id:3206451103 --> @DXXS commented on GitHub (Aug 20, 2025): FWIW, For some reason, on M1 MacBook, the latest 0.11.5 was initially running faster for me*, (w/logs claiming as usual 'low vram mode'), but when I lowered the cache from 128k to 8k, it suddenly allocated significantly more memory to ollama, yet appeared to run much slower!? [Also at that point, trying turning on OLLAMA_NEW_ESTIMATES cut the speed in ~half]
GiteaMirror commented 2026-04-12 20:04:42 -05:00
Author
Owner
Copy Link

@jessegross commented on GitHub (Aug 20, 2025):

I'm going to close this as the original bug has been fixed in the released version of 0.11.5. @DXXS, I don't think your problem is related. Please file a new issue and include logs.

<!-- gh-comment-id:3207196036 --> @jessegross commented on GitHub (Aug 20, 2025): I'm going to close this as the original bug has been fixed in the released version of 0.11.5. @DXXS, I don't think your problem is related. Please file a new issue and include logs.
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#7911
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?