Coursera Flash Sale
40% Off Coursera Plus for 3 Months!
Grab it
Learn about the technical challenges and solutions for scaling Open Compute Project (OCP) Network Interface Cards (NICs) to 1.6T speeds and beyond specifically for artificial intelligence applications in this conference talk. Explore how AI training applications demand massive computing resources across large server clusters and understand the mechanical, thermal, and system management challenges that limit server scaling within racks. Discover the current state of NIC form factors, including their maximum speeds, size constraints, and manageability issues, while examining the critical technical hurdles that must be overcome to achieve higher speeds of 1.6T, 3.2T, and beyond for AI networks. Gain insights into efficient space and power utilization requirements for scale-out bandwidth beyond single rack deployments, presented by industry experts from Broadcom and Meta who share their expertise in hardware architecture and engineering for next-generation AI infrastructure.
