2026-10-04 · America/Los_Angeles · 社区动态 · #8

Qwen3.5 arch implementation in FPGA fabric for 9B/27B INT4 models on relatively cheap eBay mining hardware

I've been wanting to test out LLM inference on FPGAs for a while now, but didn't have a big reason too since there were no frontier class models at 9B-27B scale (3.6 27B was great still but it didn't motivate me enough). When 3.8 was released I was pretty impressed by what it could at that size. So I began looking for cheap FPGAs that could hold the model. I found SQRL FK33 (280$) (8GB HBM2 ~400GB/s BW) and thought I could possibly run it with multiple cards, I'd been working on llama.c inference on a smaller AXU3EG FPGA dev board before that and primarily used Opus 4.8 (CC) for the…

热度 47.0 / 100；排名与评分保留该期记录。