DIY LLM server build: EPYC 9575F, 4x RTX 3090, 768GB RAM

View on Reddit

Reddit user C0smo777 shares their new LLM server built with a 64-core AMD EPYC 9575F, 4× RTX 3090 (96GB VRAM total), and 768GB DDR5 ECC RAM. Planned use includes running vLLM for high-throughput inference.