AMD's MI300X Outperforms Nvidia's H100 for LLM Inference

AMD's MI300X Outperforms Nvidia's H100 for LLM Inference. 264 comments, 280 points on Hacker News.