SAN FRANCISCO—Thanks to parallelism and technology scaling, exascale computing will become a reality before the decade is out, but it won't live up to its full potential unless fundamental power consumption barriers are overcome, according to Intel Fellow Shekhar Borkar.
Delivering a keynote address at the Semicon West fab tool vendor tradeshow here Tuesday (July 10), Borkar noted that exascale computing is expected to become a reality by the end of the decade. By about 2018, engineers are expected to create an exascale supercomputer—capable of a 1,000-fold performance improvement compared with today's state-of-the-art petaflop systems.
If history is any guide, about 10 years after the existence of exascale supercomputers, the technology will find its way into PCs and then, eventually, into mobile systems, Borkar said.
But if current trends hold true, an exascale computer will consume vast amounts of power, according to Borkar. The formidable challenge, he said, is to create an exascale computing system that consumes only 20 megawatts (MW) of power.
If engineers can use new technology to create an exascale system that consumes only 20 MW of power, the same technology can also be used to dramatically lower the power consumption of lower performance systems, to the point where giga-scale systems consuming only 20 milliwatts of power can be used in small toys and mega-scale systems that consume only 20 microwatts could be used in heart monitors.
"A mega-scale machine was a supercomputer back in the 60s," Borkar added.
Intel Fellow Shekhar Borkar.speaks at the Semicon West tradeshow Tuesday.
Borkar said the way forward is to improve both energy per transistor and energy per compute operation. Conventional CMOS scaling improves both, he said, but not to a large enough degree. And indication are that energy per transistor at the circuit level will not decline as much as it has in the past, he said. "Clearly we need to do something more than just scaling of technology," Borkar said.
Scaling down supply voltage increases energy efficiency, Borkar said. But doing so has a side effect—leakage power does not reduce as much as total power consumption, meaning that leakage power becomes a higher percentage of total power consumption, he said.
Borkar said near threshold voltage circuit design both reduces total power consumption and improves energy efficiency. "Clearly this is very promising technology," Borkar said. "But as you start solving the problem of energy efficiency, leakage power dominates."
During Borkar's 40-minute address Tuesday, he made several other observations. One was to emphasize the importance of "local computing" at a time when everyone is talking up the virtues of cloud computing. He noted that communications technologies used for moving data, including Bluetooth, Ethernet and Wi-Fi, use far more power than those used for local computing within a chip or system. "Clearly, data movement energy will dominate the future," Borkar said.
When Shekhar Borkar talks, the world listens. It is evident that issues related to ultra low submicron CMOS technologies will keep the power burning. But I am confident we can get this sorted out in the next few years.
Local computing is the original Wintel party line, he's reflecting Intel's wish to go back to glory days.
Better cloud argument can be made from a made up Big Brother horror story, like Amazon or Apple or Google or Microsoft monitoring your internet preferences.