China has built the world’s most powerful supercomputer with chips it manufactured locally, topping the U.S. for the first time in the number and performance of installed systems.
The world's most powerful supercomputer is made in China.
The news comes as China tops the U.S. for the first time in both the number and performance of systems on the latest Top500 list.
The Sunway TaihuLight at the National Supercomputing Center in Wuxi hit 93 petaflops/second on the Linpack benchmark and a theoretical peak performance of 125.4 Pflop/s. It uses 40,960 Sunway SW26010 processors designed by the Shanghai High Performance IC Design Center.
China’s Tianhe-2 already held the title of the world’s fastest supercomputer for the last three years, but it was built using Intel processors and an interconnect chip designed in China.
“The Sunway TaihuLight is almost three times as fast and three times as efficient as the [Tianhe-2] it displaces in the number one spot,” said Jack Dongarra, a computer science professor at University of Tennessee and co-author of the twice yearly Top500 list. The system is five times faster than the top ranked U.S. supercomputer, the Titan, a Cray XK7 system at Oak Ridge National Laboratory that hits 17.59 pflops/s, making it #3 in the world
Figure 1: The Sunway system in Wuxi used a custom China processor to hit 93 petaflops/s. (Image source: Top 500)
“For comparison, the next large acquisition of a supercomputer for the U.S. Department of Energy which is expected to be approximately 200 Pflops/s will not be until 2017 with production beginning in 2018,” wrote Dongarra in an 18-page report on the Sunway system.
Dongarra sounded a notes of both praise for China and alarm for U.S. technologists in his conclusion of the report:
*The Sunway TaihuLight system demonstrates the significant progress that China has made in designing and manufacturing large-scale computation systems…The fact that there are sizeable applications and Gordon Bell [award] contender applications running on the system is impressive and shows that the system is capable of running real applications and not just a stunt machine.
In 2001 there were no supercomputers listed on the Top500 in China. Today China has 167 systems on the list compared to 165 systems in the US. This is the first time the US has lost the lead…It is clear that they are on a path which will take them to an exascale computer by 2020, well ahead of the US plans for reaching exascale by 2023. The latest rankings are the result of China’s focused spending on supercomputing over the last three to five years while spending in the U.S. slumped, said Horst Simon, deputy director of the Lawrence Berkeley National Laboratory and another co-author of the Top500 list. The good news is the U.S. is getting back in the game with projects such as the National Strategic Computing Initiative, he said.*
In particular, the competition to deliver the first exascale-class system has “never has been as close a race as it is now,” said Simon.
Competition inside and outside China
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.
Figure 2: Horst Simon, Deputy director of the Lawrence Berkeley National Laboratory
“China can do new things,” said Simon. “Sunway’s architecture is unique. An American vendor would not have built it due to [lack of] backward compatibility,” he said.
The current rankings mark an all-time low in the number and percentage of overall performance for U.S.-based supercomputers since the Top500 list was created in 1993.
“All the U.S. systems on the list are getting old,” said Simon, 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.
“There was no urgency in Washington to get things funded,” he said, noting a U.S. exsacale project is now “firmly established” in the 2016 federal budget.
On the positive side, the U.S. has “a breadth and depth” in supercomputing expertise “that will take China a whole generation to match,” he said, noting more than 30 years of government funding for computational science.
“It will take a while for China to catch up in this area, but they have spent a lot of money with a focused goal and that focus has been missing in U.S. for last several years,” he said. “This will change going forward but there was a three or four year lack of investment that has gotten us to the point where we are now,” he added.
Web giants such as Amazon, Google and Facebook depend on the kinds of technologies developed for supercomputers, in part with government help, said Simon.
After attending a recent talk by a Facebook data center networking expert, “it’s clear they are looking at the same technology base as the high performance computing community,” said Simon. “It’s the same big processors and fast memories, just a different usage model,” he said.
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.
Take a look at the history of the most powerful supercomputers since 1993 in the gallery.*