Ethernet switch has on-chip neural network
Trident 5-X12, a software-programmable Ethernet switch from Broadcom, packs an on-chip neural-network inference engine—a first according to the company. NetGNT, which stands for networking general-purpose neural network traffic analyzer, improves network efficiency and performance.
NetGNT works in parallel to augment the standard packet-processing pipeline. The standard pipeline is one-packet/one-path. In other words, it looks at one packet as it takes a specific path through the chip’s ports and buffers. NetGNT, in contrast, is an ML inference engine and can be trained to look for different types of traffic patterns that span the entire chip. The enhanced telemetry capabilities of Trident 5 allow deeper real-time insights into bandwidth, which can then be used to train NetGNT.
The Trident 5-X12 provides 16 terabits/s of bandwidth, double that of its predecessor, the Trident 4-X9. Built using a 5-nm manufacturing process, the chip consumes 25% less power per 400G port than the Trident 4-X9. It also adds support for 800G ports, allowing direct connection to Broadcom’s Tomahawk 5, which is used as the spine/fabric in compute and AI/ML data centers.
The Trident 5-X12 BCM78800 Ethernet switch is now shipping to qualified customers.
Find more datasheets on products like this one at Datasheets.com, searchable by category, part #, description, manufacturer, and more.
googletag.cmd.push(function() { googletag.display(‘div-gpt-ad-native’); });
–>
The post Ethernet switch has on-chip neural network appeared first on EDN.
