[GH-ISSUE #1975] Will ollama run dolphin-mixtral on my gtx 1080 Ti? #78816

Closed
opened 2026-05-09 02:08:03 -05:00 by GiteaMirror · 3 comments
Owner

Originally created by @PixelovyLabyrintDev on GitHub (Jan 13, 2024).
Original GitHub issue: https://github.com/ollama/ollama/issues/1975

Im just asking since im about to buy one and im curious if it will see the gpu and use it to generate responses faster? Or does Ollama support all Nvidia gpus?

Originally created by @PixelovyLabyrintDev on GitHub (Jan 13, 2024). Original GitHub issue: https://github.com/ollama/ollama/issues/1975 Im just asking since im about to buy one and im curious if it will see the gpu and use it to generate responses faster? Or does Ollama support all Nvidia gpus?
GiteaMirror added the question label 2026-05-09 02:08:03 -05:00
Author
Owner

@easp commented on GitHub (Jan 13, 2024):

dolphin-mixtral is a fairly large model. Less than 1/2 of the default q4_0 quantization will fit on the card and so text generation speeds are going to be much closer to CPU-only speeds than GPU speeds. I'd guess something less than 2x your CPU-only speeds. That's significant, but no where close to the GPU-only speeds.

<!-- gh-comment-id:1890721080 --> @easp commented on GitHub (Jan 13, 2024): dolphin-mixtral is a fairly large model. Less than 1/2 of the default q4_0 quantization will fit on the card and so text generation speeds are going to be much closer to CPU-only speeds than GPU speeds. I'd guess something less than 2x your CPU-only speeds. That's significant, but no where close to the GPU-only speeds.
Author
Owner

@ghost commented on GitHub (Jan 15, 2024):

I have run Dolphin-Mixtral:v2.7 on 1 - 1080ti and 2 - T4's, it takes over 26 gigs of vram. It will not run on a single 1080ti
2024-01-14_22-09-48

<!-- gh-comment-id:1891313531 --> @ghost commented on GitHub (Jan 15, 2024): I have run Dolphin-Mixtral:v2.7 on 1 - 1080ti and 2 - T4's, it takes over 26 gigs of vram. It will not run on a single 1080ti ![2024-01-14_22-09-48](https://github.com/jmorganca/ollama/assets/9617359/a193b1fc-e9f6-46bf-b9be-753e43577a3b)
Author
Owner

@jmorganca commented on GitHub (Jan 15, 2024):

Hi @PixelovyLabyrintDev! Indeed, as mentioned, it will run, but not much of the model will be offloaded to run on the GPU given how much memory dolphin-mixtral requires (26GB+). Feel free to share any more questions!

<!-- gh-comment-id:1892761662 --> @jmorganca commented on GitHub (Jan 15, 2024): Hi @PixelovyLabyrintDev! Indeed, as mentioned, it will run, but not much of the model will be offloaded to run on the GPU given how much memory `dolphin-mixtral` requires (26GB+). Feel free to share any more questions!
Sign in to join this conversation.
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: github-starred/ollama#78816