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 #499] Dedicated hardware for 16b/70b models #229

New Issue
Closed
opened 2026-04-12 09:44:59 -05:00 by GiteaMirror · 3 comments
GiteaMirror commented 2026-04-12 09:44:59 -05:00
Owner
Copy Link

Originally created by @zdeneksvarc on GitHub (Sep 8, 2023).
Original GitHub issue: https://github.com/ollama/ollama/issues/499

Hey guys, let's say I want to get a dedicated home server that would run ollama serve 13b/70b in Docker. Is there any chance to get such hardware (CPU) to achieve speed at least 5 tok/s? Since Ollama doesn't use GPU acceleration.

Originally created by @zdeneksvarc on GitHub (Sep 8, 2023). Original GitHub issue: https://github.com/ollama/ollama/issues/499 Hey guys, let's say I want to get a dedicated home server that would run `ollama serve` 13b/70b in Docker. Is there any chance to get such hardware (CPU) to achieve speed at least 5 tok/s? Since Ollama doesn't use GPU acceleration.
GiteaMirror commented 2026-04-12 09:45:00 -05:00
Author
Owner
Copy Link

@mxyng commented on GitHub (Sep 8, 2023):

We're actively working on GPU acceleration on Linux and Windows. You can follow the progress in #454 which includes GPU accelerated Docker images

<!-- gh-comment-id:1712308027 --> @mxyng commented on GitHub (Sep 8, 2023): We're actively working on GPU acceleration on Linux and Windows. You can follow the progress in #454 which includes GPU accelerated Docker images
GiteaMirror commented 2026-04-12 09:45:01 -05:00
Author
Owner
Copy Link

@zdeneksvarc commented on GitHub (Sep 8, 2023):

What a great news. This Ollama thing is awesome. Thank you.

<!-- gh-comment-id:1712310529 --> @zdeneksvarc commented on GitHub (Sep 8, 2023): What a great news. This Ollama thing is awesome. Thank you.
GiteaMirror commented 2026-04-12 09:45:02 -05:00
Author
Owner
Copy Link

@zdeneksvarc commented on GitHub (Sep 9, 2023):

Although the answer to the OP was encouraging, I was still looking for hard data to tell me what hardware is needed to get at least 5 tok/s. For completeness here is a partial answer from the link below:

The M1/M2 Pro supports up to 200 GB/s unified memory bandwidth, while the M1/M2 Max supports up to 400 GB/s. For example MacBook M2 Max using Llama.cpp can run 7B model with 38 t/s, 13B model with 22 t/s, and 65B model with 5 t/s.However in terms of inference speed dual setup of RTX 3090/4090 GPUs is faster compared to the Mac M2 Pro/Max/Ultra. Two RTX 4090s can run 65b models at a speed of 20 tokens per second, while two affordable secondhand RTX 3090s achieve 15 tokens per second with Exllama. Additionally, the Mac evaluates prompts slower, making the dual GPU setup more appealing.

And here is awesome comprehensive answer: https://www.hardware-corner.net/guides/computer-to-run-llama-ai-model/

<!-- gh-comment-id:1712444329 --> @zdeneksvarc commented on GitHub (Sep 9, 2023): Although the answer to the OP was encouraging, I was still looking for hard data to tell me what hardware is needed to get at least 5 tok/s. For completeness here is a partial answer from the link below: >The M1/M2 Pro supports up to 200 GB/s unified memory bandwidth, while the M1/M2 Max supports up to 400 GB/s. For example MacBook M2 Max using Llama.cpp can run 7B model with 38 t/s, 13B model with 22 t/s, and [65B model with 5 t/s](https://twitter.com/natfriedman/status/1665408927431884800).However in terms of inference speed dual setup of RTX 3090/4090 GPUs is faster compared to the Mac M2 Pro/Max/Ultra. Two RTX 4090s [can run 65b models at a speed of 20 tokens per second](https://github.com/turboderp/exllama#dual-gpu-results), while two affordable secondhand RTX 3090s achieve 15 tokens per second with Exllama. Additionally, the Mac evaluates prompts slower, making the dual GPU setup more appealing. And here is awesome comprehensive answer: https://www.hardware-corner.net/guides/computer-to-run-llama-ai-model/
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#229
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?