PROMO Today’s AI and HPC systems can encompass 1,000s of CPUs and GPUs, all drawing on vast amounts of memory and storage. And, if we’re honest, all over woefully inadequate interconnects.
Tomorrow’s AI models and HPC workloads will only get bigger and more demanding – NLP models increased three orders of magnitude from 2019 to 2020 alone.
It’s increasingly clear that the most logical way to scale up systems to handle these workloads is disaggregation, allowing CPUs and GPUs to access bigger pools of resources, such as shared DRAM.
But this highlights the shortcomings of current interconnect technologies. While Compute Express Link builds on PCIe to provide a unified protocol for IO within disaggregated systems, it still has limitations regarding reach and fan out.
At a fundamental level, it’s clear that electrical-based signaling over copper wires will always act as a brake for systems makers.
However, there is already talk of combining CXL and optical, offering the potential of massively increased bandwidth over greater distances, with lower power and latency.
And to get an insight into progress in this area, you should join this Ayar Labs’ Livestream Event on Enabling HPC/AI Disaggregated System Architectures with Optical I/O, on June 23, between 9.30 and 11am, EDT.
Proceedings will be overseen by The Next Platform’s Timothy Prickett Morgan, who will be welcoming a stellar panel featuring key researchers from the world’s top supercomputing labs and senior technologists working on solutions to the interconnect problem, including:
- Timothy Prickett Morgan, (Moderator) Co-Editor, Co-Founder, The Next Platform
- Mike Ignatowski, Senior Fellow, AMD
- Hugo Saleh, VP of Marketing & Business Development, Ayar Labs
- Dr. Ian Karlin, Principal HPC Strategist at Lawrence Livermore National Laboratory
- Craig Prunty, VP of Marketing, SiPearl
- Mark Parsons, Director, EPCC, The University of Edinburgh
They’ll talk you through the state of the art when it comes to disaggregated systems, where CXL gets us, what the industry needs to work through on protocols and standards to deliver optical IO, and what this potentially enables.
Anyone with a stake in the future of HPC and AI system architectures, from researchers, engineers, and system architects to analysts and investors, will want to hear what this IO brain trust has to say. All you need to do is save the date.
Sponsored by Ayar Labs
Sign up to our Newsletter
Featuring highlights, analysis, and stories from the week directly from us to your inbox with nothing in between.
Be the first to comment