Broadcom is starting to ship out its latest Ethernet switch, the Tomahawk Ultra, designed for high-performance computing (HPC) and AI workloads.
According to Broadcom, the “Tomahawk Ultra redefines what an Ethernet switch can deliver.” It offers ultra-low latency, and the company claims it can achieve 250ns latency at 51.2 Tbps throughput.
It also offers line-rate switching performance at the minimum packet size of 64 bytes, and can support up to 77 billion packets per second.
Header overhead was also optimized in this model, being reduced down to as low as 10 bytes.
Tomahawk Ultra implements Link Layer Retry (LLR) and Credit-Based Flow Control (CBFC) to cut down on packet loss.
The company says this switch is optimized for AI workloads because it executes In-Network Collectives operations directly rather than outsourcing them to XPUs. This reduces job completion time, improves utilization of compute resources, and is endpoint-agnostic, so it can be adopted across a range of system architectures and vendor ecosystems, Broadcom explained.
“Tomahawk Ultra is a testament to innovation, involving a multi-year effort by hundreds of engineers who reimagined every aspect of the Ethernet switch,” said Ram Velaga, senior vice president and general manager of the Core Switching Group at Broadcom. “This highlights Broadcom’s commitment to invest in advancing Ethernet for high-performance networking and AI scale-up.”
