Show HN: Llama 3.1 70B on a single RTX 3090 via NVMe-to-GPU bypassing the CPU

📅 2026-02-21    ⚓ Hacker News    🌐 Source    🖼️ Load Image