← sources

No option for Strata Qwen 3.8 FN for dual 3090s and 32GB RAM?

source: reddit (reddit:LocalLLaMA)published 2026-10-30 20:33ingested 2026-10-31 02:40

retained text

I'm trying to run Qwen 3.8 FN on a dual 3090 machine with 32GB DDR5 RAM using this repo: https://github.com/q8atnight/strata-dual-3090

I'm referring to a few posts on here that demonstrate that it's possible with a single 3090 and 64GB+ RAM and this one that did it with dual 3090 and 48GB RAM: [post](https://old.reddit.com/r/LocalLLaMA/comments/1xc7n0n/i_have_an_answer_qwen_flash_next_iq2_is_better/)

I built this machine after the RAM apocalypse and paid $300 for 32GB, thinking I could lean on the dual 3090s for the most part.

But it seems like even the "low RAM, 1 bit coder quant" pathways in the linked repo and the strata repo won't run on my machine. Just wanted to check that this is the case before I drop $479 (gulp) on another 32GB RAM.

document fetched 2026-10-31 00:44 · canonical https://reddit.com/r/LocalLLaMA/comments/1xjqkzg/no_option_for_strata_qwen_38_fn_for_dual_3090s