Best use for these two separate systems?
retained text
Just want to confirm I am going about this logically. System 1: Gigabyte x99 mobo Xeon E5-2630V4 w/ 128gb ddr4 2x rtx3090, no nvlink. Pcie 3.0 x16 to each. System 1 is assembled currently and is running open webui with \- qwen 3.8 27b q4 on one rtx3090. \- other 3090 is handling comfyui for image generation and image editing via API from open webui. \- Also running voice to text and text to voice on cpu. \- Another docker container for searxng for web search capability in openwebui. This gives me a lot of features in one box. System 2: Asus x99 mobo Xeon E5-2630V4 w/ 256gb ddr4 4x tesla p100 16gb each at pcie 3.0 x16 Being reassembled from leftover parts soon is system2. I am considering a better quant of qwen 3.8 27b such as q8 with lots of context or trying out the qwen 3.8 flash next to see how well that runs. The idea here is to have a model that is more intelligent when needed than the model running on system 1. I also figure i can let it use the same api for the additional features being hosted on system 2. Is this the best use of the hardware I have or what would you guys be doing different with the same resources? Also i havent toyed with anything agentic yet so open to suggestions that involve moving things around to accomodate experimenting with that. Basically i am looking to have the swiss army knife of functionality running locally.
document fetched 2026-09-14 00:36 · canonical https://reddit.com/r/LocalLLaMA/comments/1wfenvj/best_use_for_these_two_separate_systems