CHANNELS

Subscribe to the InfoTech eNewsletter

infoTECH Feature

November 12, 2012

Intel Smashes Chip Speed Limits With 60-Core Xeon Phi

After two years of waiting between the initial announcement and the final release, Intel (News - Alert) is now shipping its line of 60-core Xeon Phi processors to its customers. The new processors represent both a significant advancement over their previous versions, as well as a significant step toward one of computing's great milestones.

The Xeon Phi – also known as "Knights Corner" processors – are specifically geared to work with server processors as well as supercomputers as a way to speed up tasks commonly found in science or math functions. The Phi series itself offers processors with at least 60 cores, while other versions more than 60, but it's the 60-core variety that is currently getting shipped.

The fastest models in the Phi variety can bring nearly a teraflop per second in performance to the table, but they also represent a major move toward a milestone that Intel has had as a goal for some time now; specifically, the achievement of an exaflop, or about 1,000 petaflops, computer by 2018.

The very first Phi chips are set to be installed in a 10-petaflop supercomputer, dubbed Stampede, at the University of Texas' Texas Advanced Computing Center. Stampede is set to be made active early in 2013, but even Stampede will have a long way to go to reach the goal, as it's well behind the current front-runner, a Cray XK7 powered monster named Titan running on behalf of the U.S. Department of Energy's Oak Ridge National Laboratory, which reportedly topped out at 17.59 petaflops.

The first Phi / Knights Corner chip to ship is known as the Xeon Phi 5110P, packing that 60-core loadout and clocking in at 1.05 GHz. It also boasts a 30 gigabyte cache and six gigabyte memory capacity, and can be loaded into a PCI (News - Alert)-Express 2.0 slot and cooled by a standard system fan. Better still for those planning to build the top end of supercomputers, it's priced competitively to the Tesla from Nvidia and the FireStream from AMD (News - Alert).

Others in the line, which won't be released commercially, include the Phi SE10X and the SE10P, packing in 61 cores and 1,073 gigaflops of power. They're specifically built for the upcoming Stampede, so for now, it's looking like the 5110P or nothing.

Given that right now, the top of the heap is just 17.59 petaflops, it seems like a long jump to get from there to the approximately 1,000 petaflops required to make an exaflop computer happen. The Phi line itself is comprised of design elements from several other chips, like a 48-core chip that was geared for cloud computing and a chip with a whopping 80 cores, and was said to be a response to keeping up with Moore's Law. The issue here, of course, is that Moore's Law has something of a shelf life, and while companies are doing their level best to make Moore's Law a self-fulfilling prophecy, there are certain limits of physics that are coming into play.

Sure, IBM (News - Alert) is doing some very impressive things with carbon nanotubes right now that may well represent a shift in the landscape once more, but getting to an exaflop is not going to be easy.

Still, with applications in science and national security riding on having the best in computing hardware, it's a safe bet that Intel – and the rest of the market – will carry on and build impressively high-powered hardware for some time to come.

Want to learn more about the latest in communications and technology? Then be sure to attend ITEXPO Miami 2013, Jan 29- Feb. 1 in Miami, Florida.  Stay in touch with everything happening at ITEXPO (News - Alert). Follow us on Twitter.




Edited by Allison Boccamazzo
FOLLOW US

Subscribe to InfoTECH Spotlight eNews

InfoTECH Spotlight eNews delivers the latest news impacting technology in the IT industry each week. Sign up to receive FREE breaking news today!
FREE eNewsletter