Supercomputers Are For Everyone

Our SerDes world tour continues. This past month, we demonstrated our 7nm 56G long-reach SerDes in Dallas and Israel. In Dallas, our demonstration included error-free operation in 56G PAM4 over a 30dB channel without forward error correction through an eye-popping five-meter cable. Many thanks to our partner Samtec for providing that cable, allowing backplane designers to now “reach beyond the rack.” The show was a great success with record numbers of attendees coming to the Samtec booth to see our demo. There is a short video of that SerDes demo on our website if you couldn’t make it to the show. You’ll find it under “High-Performance Networking & Computing ASICs and IP.”

The show in Dallas is dedicated to supercomputing. In their words, “an international conference of high-performance computing, networking, storage and analysis.” The event was simply referred to as SC18.  Attending this show was quite an experience for me. Until this event, I believe supercomputing was reserved for the ultra-high end of the market – the rare air if you will. What I found at SC18 changed that perspective. First of all, this is a huge show with over 12,000 attendees and approximately 400 exhibitors across 44 aisles of exhibit space.

Of course, the usual suspects were at the show, but a large array of newcomers were also there with a large presence and impressive exhibits. Everything from processing blades, communications gear, exotic cooling, and workload management, with a strong unifying voice from AI across most of the show.

Many universities were there as well, covering both hardware and software research.

Many exhibits showcased new technologies for liquid cooling. I recall the early days of mainframes when liquid cooling was used. Then came minicomputers and air cooling was the standard. It appears we’re moving back to liquid to deal with the high heat loads of advanced processing power.

After walking around the exhibit floor for a few hours (and a few miles), it occurred to me that supercomputing was no longer a “rare air” technology. Thanks to advances in processing technology, artificial intelligence and the unstoppable move to the cloud, “utility” model supercomputing is now available to everyone. It is indeed a massive trend, perhaps a revolution. At eSilicon, we’re excited to be part of the revolution. In the coming months we’ll be at more conferences demonstrating new capabilities to make supercomputing and AI more accessible.  Keep an eye on our events page to see where we’ll be. And if you can make it to SC19 next year in Denver, I would definitely go.