Cisco Systems Launches Network Chip Series For Supercomputer AI Competing With Broadcom And Marvell Technology

Cisco Systems launched a network chip for supercomputer AI (photo: Twitter @FranKatsoudas).

JAKARTA - Cisco Systems on Tuesday, June 20 launched a network chip for the AI supercomputer that will compete with products offered by Broadcom and Marvell Technology.

Chips from the SiliconOne series are being tested by five of the six main cloud service providers, Cisco said without naming the companies. Some of the main players in the cloud industry include Amazon Web Services, Microsoft Azure, and Google Cloud, which jointly dominate the cloud computing market, according to Bofa Global Research.

The increasing popularity of AI applications such as ChatGPT, supported by a special chip network called graphics processing units (GPUS), makes the speed of communication between these chips very important.

Cisco is one of the main suppliers of network equipment including the ethernet switch, which connects devices such as computers, laptops, routers, servers, and printers to local area networks.

The company states that the latest generation of their ethernet switches, namely G200 and G202, has double performance compared to the previous generation, and can connect up to 32,000 GPU units.

"G200 & G202 will be the most powerful network chip in the market that will drive AI / ML's workload and enable the most efficient network in terms of power use," said Rakesh Chopra, Follow Cisco, and previously the engineer's principle.

Cisco menyatakan bahwa chip-chip ini dapat membantu dalam menjalankan tugas-tugas AI dan pembelajaran mesin dengan menggunakan 40% lebih sakelar dan memiliki keundatan yang lebih rendah, sambil menjadi lebih efisien dalam penggunaan daya.

In April, Broadcom announced a GUN3-AI chip that could connect up to 32,000 GPU units together.