← sources

Suggestion for running Qwen3.8 FN local

source: reddit (reddit:LocalLLaMA)published 2026-10-26 08:13ingested 2026-10-27 02:40

retained text

Hi, currently i got these hardware for use .

  
2x 3090 24GB "unlocked", 1x 4070 TI Super 16GB "unlocked", 18 Core, 130GB DDR4 Ram, 1TB NVME



When i Run Qwen3.8FN only from  4070TI +NVME i get like 8-9 T/s When i run it only form NVME i get 2 T/s and Running From all 3 GPU and RAM i get 18 T/s.

  
what can i do to get more speed without buying new hardware ?

document fetched 2026-10-27 00:43 · canonical https://reddit.com/r/LocalLLaMA/comments/1xfqvpj/suggestion_for_running_qwen38_fn_local