<p>➀ AMD is facing challenges in catching up with NCCL and needs exclusive access to a persistent cluster of at least 1,024 MI300 class GPUs.</p><p>➁ AMD's RCCL library is a fork of Nvidia's NCCL and requires significant engineering hours to sync with Nvidia's major refactor.</p><p>➂ AMD is planning to rewrite RCCL from scratch to stop being a fork of NCCL.</p><p>➃ NVIDIA's NCCL continues to advance with new features and performance improvements.</p><p>➄ AMD has made progress in software infrastructure but is falling behind in ML libraries.</p><p>➅ AMD lacks support for features like disaggregated prefill and NVMe KV Cache Tiering.</p><p>➆ Recommendations are made to both AMD and NVIDIA for improving their competitive positions.</p>