No GPU, No Problem: Flagship LLMs On A GPU-less Teenaged Server

A YouTuber known as MattMo demonstrates running large flagship LLMs, including GLM 5.3 Flash, Qwen 3.8 Flash, and Qwen 3.8 27B, entirely on CPU using a 14-year-old Dell PowerEdge R720 server with dual Xeon processors and 348 GB of DDR3 RAM, without any GPU. Speeds top out around four tokens per second due to limited threads and older CPU instruction sets, but the massive RAM capacity lets even the largest models fit in memory. The setup costs roughly $600 on the secondhand hardware market and could suit batch-processing workloads or homelab enthusiasts who already own similar gear, though it's far from real-time conversational speed.

Nguồn: https://hackaday.com/2026/09/20/no-gpu-no-problem-flagship-llms-on-a-gpu-less-teenaged-server. 8 Sync News chỉ tóm tắt và dẫn link; bản quyền nội dung thuộc tác giả và nguồn gốc.