← sources

GPU Recommendations for running local LLMs with ollama on proxmox

source: reddit (reddit:ollama)published 2026-10-14 10:43ingested 2026-10-15 02:42

retained text

Hello everyone,

I'm trying local models with ollama on my pc desktop used as a home server.

The server runs proxmox, and i created a container lxc running ollama.  
I currently allocated all the cpu of the machine and 18GB of ram to the container.  
I've tried models such as `qwen2.5-coder:7b` and `qwen3:8b` using the Continue extension in VS Code.

Here are my server's specifications:

|**Cpu**|intel core i5-12400 (6 core, 3.0 GHz )|
|:-|:-|
|**MotherBoard**|ASUS Prime H610M-K D4|
|**RAM**|2x 16GB DDR4 3200Mhz (KVR32N22S8/8)|
|**Scheda Video**|Intel UHD Graphics 730 (integrated into cpu)|
|**Alimentatore**|Corsair CX550|
|**Disks**|SSD Nvme 512GB|

Since I don't have a dedicated GPU, I'm currently running the models entirely on the CPU, achieving around 7–8 tokens per second.

I'm considering buying a gpu for my machine (crying with actual costs) that will be used with proxmox and the container lxc for run local models for programming.

Which gpus would you recommend for my setup?  
  
Thank you very much!

document fetched 2026-10-15 00:45 · canonical https://reddit.com/r/ollama/comments/1x5nxtg/gpu_recommendations_for_running_local_llms_with