Broadcom Enhances Trident Networking Chip with Silicon AI Features


Tech giant Broadcom (NASDAQ:AVGO) announced on Thursday the integration of artificial intelligence features into a new iteration of its flagship networking chip, aiming to enhance the efficiency of data transfer within data centers. The addition of AI capabilities to a segment of Broadcom’s Trident networking processors not only improves the chip’s overall efficiency but also bolsters its capacity to handle diverse tasks, including network security. The Trident 5-X12 chip introduces typical performance enhancements, such as reduced power consumption and increased network bandwidth.

Addressing a significant challenge in constructing extensive AI clusters—efficient data movement within them—the Trident 5-X12 chip is designed to mitigate network traffic congestion. According to Robin Grindley, an executive in Broadcom’s Core Switching Group, the chip’s AI component plays a crucial role in tasks like AI, where traditional approaches prove too sluggish. Grindley explained that the neural network aspect of the chip examines all packets and traffic patterns, identifying elements that conventional methods might miss.

The AI functionality of the chip can be activated once a customer develops an AI model based on the traffic within their data center. Data center operators can train the model to detect specific types of undesired traffic, such as denial-of-service attacks or network congestion. Once created, a version of the model is input into the chip, aiding in efficient traffic routing.

The decision to incorporate AI features into the chip was made approximately two years ago, following the introduction of programmability in a previous chip version, as per Grindley. The Trident’s latest iteration utilizes 5-nanometer manufacturing technology and is currently being shipped to “qualified” customers.

Featured Image: Megapixl

Please See Disclaimer