Cisco's latest chip connects AI data centers located far apart
What's the story
Cisco Systems has unveiled a new networking chip, the P200, to connect artificial intelligence (AI) data centers. The cloud computing divisions of Microsoft and Alibaba have already signed up as customers for this innovative technology. The P200 will compete with similar products from Broadcom and will be integrated into a new routing device launched by Cisco.
Technological advancement
The P200 chip's purpose
The P200 chip is designed to connect large-scale data centers that are located far apart. These centers are used by companies like NVIDIA, which link tens of thousands of powerful computing chips into a single unit for handling AI tasks. The main goal of the new Cisco chip and router is to connect multiple data centers into one giant computer.
Market need
Firm's insights on the technology
Martin Lund, Cisco's EVP of Common Hardware Group, told Reuters that the new technology is a response to the growing demand for connecting multiple data centers. He said, "Now we're saying, 'the training job is so large, I need multiple data centers to connect together.'" These can be up to 1,609km apart due to huge electricity consumption by data centers.
Energy efficiency
The benefits of the new chip and router
The P200 chip replaces what used to take 92 separate chips with just one. The resulting router consumes 65% less power than comparable ones. One of the major challenges in this field is keeping data in sync across multiple data centers without loss, a problem that Cisco has been tackling for decades with its buffering technology.