Follow
Show HN: Llama 3.1 70B on a single RTX 3090 via NVMe-to-GPU bypassing the CPU
https://news.ycombinator.com/item?id=47104667
#hackernews #tech
