github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-07 00:22:43 -05:00
Code Issues 14.5k Packages Projects Releases 120 Wiki Activity

[GH-ISSUE #2581] Windows Preview 8x slower than Running Through Docker #1518

New Issue
Closed
opened 2026-04-12 11:25:38 -05:00 by GiteaMirror · 4 comments
GiteaMirror commented 2026-04-12 11:25:38 -05:00
Owner
Copy Link

Originally created by @chigkim on GitHub (Feb 18, 2024).
Original GitHub issue: https://github.com/ollama/ollama/issues/2581

Originally assigned to: @dhiltgen on GitHub.

I've been running Ollama through Docker on Windows with cpu only.
Someone running Windows preview for the same time told me Ollama was extremely slow, like no response for 5 minutes.
So I tried Windows preview for myself, and I can confirm that The speed is extremely slow.
Windows preview: Total: 77.38 secs, Load: 4.72 secs, Prompt: 46 tokens (0.80 t/s), Output: 13 tokens (0.86 t/s)
Docker: Total: 9.28 secs, Load: 1.15 secs, Prompt: 26 tokens (4.44 t/s), Output: 11 tokens (4.82 t/s)
Everything is slower on Windows preview overall.
I used the same model, same prompt, same machine.
Hopefully it get sorted out soon!

Originally created by @chigkim on GitHub (Feb 18, 2024). Original GitHub issue: https://github.com/ollama/ollama/issues/2581 Originally assigned to: @dhiltgen on GitHub. I've been running Ollama through Docker on Windows with cpu only. Someone running Windows preview for the same time told me Ollama was extremely slow, like no response for 5 minutes. So I tried Windows preview for myself, and I can confirm that The speed is extremely slow. Windows preview: Total: 77.38 secs, Load: 4.72 secs, Prompt: 46 tokens (0.80 t/s), Output: 13 tokens (0.86 t/s) Docker: Total: 9.28 secs, Load: 1.15 secs, Prompt: 26 tokens (4.44 t/s), Output: 11 tokens (4.82 t/s) Everything is slower on Windows preview overall. I used the same model, same prompt, same machine. Hopefully it get sorted out soon!
GiteaMirror added the bug label 2026-04-12 11:25:38 -05:00
GiteaMirror commented 2026-04-12 11:25:39 -05:00
Author
Owner
Copy Link

@chuangtc commented on GitHub (Feb 19, 2024):

I used the same model (mistral:7b), same prompt, same Win11 machine (Intel Core i5-12400, no NVIDIA GPU). The Windows Preview version is 8~10x slower than Ubuntu 22.04.3 on WSL2. (eval rate is 0.9 tokens/s vs 6 tokens/s)

<!-- gh-comment-id:1951649009 --> @chuangtc commented on GitHub (Feb 19, 2024): I used the same model (mistral:7b), same prompt, same Win11 machine (Intel Core i5-12400, no NVIDIA GPU). The Windows Preview version is 8~10x slower than Ubuntu 22.04.3 on WSL2. (eval rate is 0.9 tokens/s vs 6 tokens/s)
GiteaMirror commented 2026-04-12 11:25:39 -05:00
Author
Owner
Copy Link

@chuangtc commented on GitHub (Feb 19, 2024):

The tool that I used for testing throughput performance of eval time is this one.
https://github.com/aidatatools/ollama-benchmark/
Maybe @jmorganca can think about how to integrate this into the original ollama project.

<!-- gh-comment-id:1952076104 --> @chuangtc commented on GitHub (Feb 19, 2024): The tool that I used for testing throughput performance of eval time is this one. https://github.com/aidatatools/ollama-benchmark/ Maybe @jmorganca can think about how to integrate this into the original ollama project.
GiteaMirror commented 2026-04-12 11:25:39 -05:00
Author
Owner
Copy Link

@dhiltgen commented on GitHub (Feb 26, 2024):

We fixed CPU performance on windows in 0.1.27. Can you try again and see if that resolves your performance problem? If not, could you share logs from the container and native windows server runs for comparison so we can see why there's still a difference?

<!-- gh-comment-id:1964978429 --> @dhiltgen commented on GitHub (Feb 26, 2024): We fixed CPU performance on windows in 0.1.27. Can you try again and see if that resolves your performance problem? If not, could you share logs from the container and native windows server runs for comparison so we can see why there's still a difference?
GiteaMirror commented 2026-04-12 11:25:39 -05:00
Author
Owner
Copy Link

@chuangtc commented on GitHub (Feb 26, 2024):

I tried it in 0.1.27 and it seems working fine now in native Windows application. The details had been published here. https://medium.com/aidatatools/ollama-v0-1-27-ai-benchmark-23dc598bb8dc
Screenshot Ollama v0 1 27 AI benchmark

<!-- gh-comment-id:1965251057 --> @chuangtc commented on GitHub (Feb 26, 2024): I tried it in 0.1.27 and it seems working fine now in native Windows application. The details had been published here. https://medium.com/aidatatools/ollama-v0-1-27-ai-benchmark-23dc598bb8dc <img width="700" alt="Screenshot Ollama v0 1 27 AI benchmark" src="https://github.com/ollama/ollama/assets/2288469/7fd395d1-2e97-4a98-8347-c39783246a82">
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#1518
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?