Broadcom has launched its Jericho4 networking chip, designed to enhance AI computation by connecting data centres separated by distances up to 96.5 kilometres. This chip facilitates the transfer of large data volumes at high speeds, enabling the creation of a large system by linking smaller data centres for AI model development and operation.
The Jericho4 chip improves network traffic flow within and between data centres, crucial for computationally intensive AI tasks requiring the coordination of numerous graphics processors. It uses high-bandwidth memory, similar to that found in Nvidia and AMD AI processors, to manage network congestion effectively. The chip also prioritises security by encrypting data during transfer. A single system can support approximately 4,500 of these chips.
Broadcom's Jericho4 switch family delivers 51.2 Tbps of deep-buffered Ethernet, enabling low-latency networking optimised for distributed AI workloads. The Jericho4 family is built for scale-out across data centres, providing features that meet the specific needs of AI workloads.
Subscribe for Weekly Updates
Stay ahead with our weekly AI and tech briefings, delivered every Tuesday.
