The machine, called IBM Summit, will be delivered to the US Department of Energy's Oak Ridge National Laboratory in early 2018.
The money for Summit’s and Aurora’s site preparation is not allocated yet, nor is the money for their actual installations, which presumably will be part of the FY18 budget request. So all of this might turn out to be wishful thinking, and given Congress’s recent performance, it surely could. But that’s the plan as it stands today.
The updated spec for Summit (above) provides a higher node count of ~4600, compared to ~3400 for the original 150-petaflop configuration. The total peak power draw increased accordingly, from 10MW to 13.3MW. TaihuLight, the current top system, uses more than 15MW. Also note that the Aurora system, which is planning to use the next-generation Knights Hill Xeon Phi processors and reach 180 petaflops in 2018-2019, will draw nearly the same amount of power as the Power9/Volta GPU-powered Summit, but will do so with about 10 times as many nodes.
Of course, by the time Summit and Aurora are installed, there may be other Chinese systems of similar or even greater performance. Regardless, the DOE’s supercomputing timeline is more methodical than that of its Chinese counterparts, and there does not appear to be any specific effort by US policymakers to catch up to their rivals in the FLOPS department.
That contrasts markedly with what happened in 2002 when Japan’s Earth Simulator opened up a performance gap with the US systems similar to the one we see today with TaihuLight. At that time, the response by the DOE was immediate and spurred the agency to accelerate their supercomputing program significantly.
The US response is likely to be very different if the Chinese systems manage to close the application gap on their way to exascale. That’s a more difficult task than just developing fast processors and adding more of them to a machine, but given China’s trajectory in supercomputing over the last several years, it is certainly conceivable.
China’s latest five-year plan called for building two 100-petaflops-class supercomputers. A U.S export ban prevented the Tainhe-2 system from getting the Intel processors it wanted to upgrade to that level of performance, fueling enthusiasm for domestic efforts like the Sunway system.
Last year, representatives of China’s National University of Defense Technology said they have plans to design custom processors based on ARM cores to upgrade Tianhe-2, said Simon. “There is incredible internal competition in [supercomputing in] China just like there is in U.S.” he said.
China now has three domestic supercomputer vendors with systems among the Top500, server makers Inspur, Sugon and Lenovo which purchased IBM’s x86 server business. In addition, China does not need to support the kinds of legacy applications supercomputers in the U.S. must run.
“All the U.S. systems on the list are getting old,” said Horst Simon, Deputy director of the Lawrence Berkeley National Laboratory, noting several are beyond the three-to-five years of typical upgrades. “There will be no 100-petaflop systems in the U.S. for two years…and we have been complacent for four or five years about pushing for an exaflop system,” he said, blaming the 2008 recession, sequestration and the political climate.
Tomorrow’s exascale class computers “will be an entirely new breed,” said Dongarra in an email exchange. “Not only are they fast, they handle big data in entirely new ways [that] open avenues for techniques in artificial intelligence, data science and simulations that can tease out new insights,” he said, citing applications in modeling everything from cancer cells to climate change.
“Loss of U.S. leadership in supercomputing and simulation will have staggering consequences,” Dongarra said.
“Without deliberate and sustained investment, supercomputers manufactured abroad with ever-improving technologies developed elsewhere will soon dominate, and the simulation techniques invented in the United States will become other nations' innovation advantages,” he added.
SOURCES- The Inquirer UK, Top500, DOE, EE Times Asia