OpenAI Opens MRC Protocol for Ultra-Large-Scale AI Training Networks
Published · May 7 · Thu Source · 量子位 (CN)

OpenAI Opens MRC Protocol for Ultra-Large-Scale AI Training Networks

OpenAI, in collaboration with five major giants—AMD, Broadcom, Intel, Microsoft, and NVIDIA—has launched the Multi-Path Reliable Connectivity (MRC) protocol, open-sourcing it to the entire industry via OCP. MRC is an extension based on RoCE, utilizing multi-plane network design and adaptive packet spraying technology. It disperses data packets across hundreds of paths for parallel transmission, avoiding network congestion and hotspots, and reducing failure recovery time from seconds to microseconds.

KeywordsOpenAIMicrosoftNVIDIAOpensMRCProtocolUltra-Large-ScaleAI

OpenAI, in collaboration with five major giants—AMD, Broadcom, Intel, Microsoft, and NVIDIA—has launched the Multi-Path Reliable Connectivity (MRC) protocol, open-sourcing it to the entire industry via OCP. MRC is an extension based on RoCE, utilizing multi-plane network design and adaptive packet spraying technology. It disperses data packets across hundreds of paths for parallel transmission, avoiding network congestion and hotspots, and reducing failure recovery time from seconds to microseconds.

This page provides an editorial summary based on publicly available information. It is not a republished article. Use the source link below for the original report.