Cisco and Nvidia Unveil Rack-Scale AI Data Center Architecture
Cisco Systems and Nvidia have strengthened their partnership to develop a rack-scale AI data center architecture. The collaboration aims to meet the growing demand for generative artificial intelligence, leveraging Cisco's Ethernet networking infrastructure and server management with Nvidia's Tensor Core GPUs.
The new architecture integrates Cisco's N9100 series data center switches powered by Nvidia's Spectrum-X Ethernet silicon and Cisco's Silicon One ASICs. This hybrid approach ensures ultra-low latency and advanced congestion control for training generative AI models across thousands of GPUs simultaneously.
Cisco emphasizes that the full-stack solution guarantees data privacy and operational security, a vital selling point for corporations concerned about leaking proprietary data into public AI models.