Broadcom has announced the new Tomahawk Ultra processor, designed to enhance data processing efficiency in AI applications, aimed at improving coordination among compute units.
Competition with Nvidia
The new Tomahawk Ultra chip from Broadcom is designed for handling large data volumes and is aimed at competing with Nvidia’s products. According to the company, the new processor can link up to four times more processors in a network compared to Nvidia’s NVLink Switch.
Development of Tomahawk Ultra
The processor was developed by a team of Broadcom engineers over three years and utilizes cutting-edge manufacturing technology with a five-nanometer process from Taiwan Semiconductor Manufacturing Co. It is targeted at high-performance computing and the booming generative AI sector.
The Future of Ethernet in AI and HPC
With the rollout of Tomahawk Ultra, Broadcom aims to use Ethernet as a standard for high-performance networking in AI. As chip performance becomes critical, the company hopes to provide data center operators with an easier path to scaling their AI capabilities.
With the launch of Tomahawk Ultra, competition in the AI chip market is entering a new phase, which could shift dynamics in this rapidly evolving space.
