U.S. Invests $258 Million in Supercomputing Race With China
The U.S. government is trying to stave off China and other countries challenging the U.S. for dominance in the next generation of the world's fastest computers.
The Department of Energy on Thursday said it would award $258 million over three years to be shared by six tech companies, as part of a plan to develop new supercomputers that can crunch data at least 50 times faster than the nation's most powerful systems today.
The companies are Advanced Micro Devices Inc., Cray Inc., Hewlett Packard Enterprise Co., International Business Machines Corp., Intel Corp. and Nvidia Corp.
U.S. government leaders warned in a September 2016 technical meeting convened by the National Security Agency and the DOE that the country was in danger of losing its leadership in supercomputers to China. Governments have long used these systems to crack codes and develop nuclear weapons, and the supercomputers have business purposes such as oil exploration and auto design.
The U.S. is in a "horse race" with China, the European Union and Japan as they attempt to outdo one another in processing power, said Steve Conway, senior vice president of research at Hyperion Research.
The fastest U.S. computer, the Titan, built by Cray, can handle 17,590 trillion calculations per second, or the rough equivalent of 11.6 million iPad Pros running at the same time. It is about the size of a basketball court and is said to use enough power to run a small town.
In June 2016, China took the top spot in a twice-yearly ranking of the 500 fastest scientific computers. China's machine, called the Sunway TaihuLight, marked the first time China had taken the top speed ranking without using U.S. semiconductor technology. China also, for the first time, placed more machines than the U.S. on the so-called Top 500 list, by 167 to 165. In a November 2016 ranking, the U.S. and China each had 171 systems on the list.
The DOE funding will be used by the tech companies to further the research and development into exascale computers. Hewlett Packard Enterprise, for example, last month demonstrated a prototype of a new memory-driven computer called The Machine, which it will further develop with the additional funds by the government. The system is HP Enterprise's largest research and development program in the history of the company. That computing system is capable of simultaneously working with the data held in every book in the Library of Congress five times over, about 160 million books.
Hyperion Research estimates that the price to research, develop and buy a U.S. exascale system will total about $300 million to $500 million per system. Oak Ridge National Laboratory spent about $97 million to buy the Titan supercomputer from Cray. While the U.S. government is typically the first buyer for these systems, the technology advances usually trickle down to other computing products at lower prices.
Write to Rachael King at rachael.king@wsj.com
(END) Dow Jones Newswires
June 15, 2017 14:14 ET (18:14 GMT)