Broadcom has introduced the Tomahawk Ultra chip designed to enhance AI data flow. It connects more processors than Nvidia’s NVLink Switch, leveraging high-speed Ethernet for communication, aiming to improve AI and HPC workloads.
The Tomahawk Ultra aims to challenge Nvidia’s dominance by providing low latency and high throughput. Developed over three years, this chip is fabricated by Taiwan Semiconductor Manufacturing Co. And focuses on scaling larger data centers.
With the Tomahawk Ultra now available, Broadcom hopes to allow data center operators to easily scale their AI operations, shifting the landscape in the race to optimize AI infrastructure and maintain competitive speeds.
Leave a Reply