← sources

Local Ai Pc 7663 Dual Epyc / Dual 9709

source: reddit (reddit:LocalLLaMA)published 2026-10-03 10:17ingested 2026-10-04 02:38

retained text

CPU Information  
Name  
AMD EPYC 7663  
Topology  
2 Processors, 112 Cores, 224 Threads

Memory Information  
RDiMM 2933Mhz  
Size 1007.61 GB

System Information  
Operating System  
Ubuntu 24.04.5 LTS

Motherboard  
Giga Computing MZ72-HB2-00

GPUs  
2x ASUS Turbo R9700 AI Pro 32 GB newest BIOS low Fan Profile (throttled  to 210w currently)  
ROCm version: 7.2.3

Beside that baby I have a Strix Halo M5 128Gb 

Now wanted to setup that big boy for local llm

For myself want to use it for coding . But   
I am also looking for something where I can also easily switch Model within the UI . Can Openwebui reload the model and what I read is that vllm is best for tensor split formte two cards . Also want to use  it for family for image creation within the ui and also Image checking kind of Allrounder as chatgpt 

Is that possible with vLLM?

I am also looking for docker setups so i can keep my host clean

Thank you for you suggestions

document fetched 2026-10-04 00:36 · canonical https://reddit.com/r/LocalLLaMA/comments/1wwjrmw/local_ai_pc_7663_dual_epyc_dual_9709