智通财经网·2025-07-16 04:24Group 1 - Broadcom has begun mass production of its groundbreaking network processor "Tomahawk Ultra," designed for high-performance computing (HPC) and artificial intelligence applications [1] - The Tomahawk Ultra series is optimized for tightly coupled, low-latency communication patterns in HPC systems and AI clusters, providing predictable and efficient performance for large-scale simulations and AI model training [1] - The chip aims to compete with NVIDIA's NVLink switch chip, offering four times the chip count and utilizing an enhanced Ethernet version instead of a proprietary data transfer protocol [1] Group 2 - As part of its Ethernet advancement strategy for AI scaling, Broadcom has launched SUE-Lite, an optimized version of the SUE specification for power and area-sensitive accelerator applications [2] - The Tomahawk Ultra series is fully compatible with the Tomahawk 5 series, ensuring rapid product deployment [2] - The series has already started shipping and will be used in rack-based AI training clusters and supercomputing environments [2]