In a recent announcement by Broadcom, the Jericho4 system boasts impressive scalability, with each HyperPort running at 3.2 Tbps and supporting deep buffering, line-rate MACsec encryption, and RoCE transport over long distances exceeding 100 kilometers.
The integration of HBM in Jericho4 represents a significant advancement in memory capacity and power efficiency, as noted by Lian Jie Su, chief analyst at Omdia. This innovation allows for faster data processing compared to traditional buffering methods, paving the way for improved AI workload distribution and enhanced data center utilization across multiple locations.
Experts highlight the unique features of Jericho4, built on TSMC’s 3-nanometer process, such as increased transistor density to accommodate more ports, integrated memory, and improved power efficiency. These capabilities are essential for handling large AI workloads and coordinating distributed AI processing efficiently, making Jericho4 an ideal solution for enterprises and cloud service providers seeking stable interconnects for their evolving AI data centers.
