Coursera Spring Sale
40% Off Coursera Plus Annual!
Grab it
Explore the critical network architecture challenges facing modern AI clusters through this 14-minute conference talk that examines how exponential growth in GPU interconnectivity demands fundamental shifts from traditional scale-out designs. Learn about the high-bandwidth, low-latency requirements of collective communication in AI supercomputers and discover how 100T scale-up switching addresses these technical challenges. Analyze cutting-edge switch ASIC technology that leverages high-speed 100G-200G SERDES to create single-stage, non-blocking network fabrics that dramatically increase intra-cluster bandwidth while reducing latency by eliminating intermediate network hops. Gain technical insights into this advanced architecture and understand its specific benefits for memory-semantic communication and coordinated computing in AI workloads. Understand how adopting scale-up fabric designs can unlock the full potential of AI supercomputers and ensure future-proof performance for the most demanding computational requirements.
