U.S. researchers are bracing for the prospect that China will once again have the world's fastest computer─and could hang on to the speed crown for some time.
美国的研究人员正准备迎接中国将再次拥有世界上运算速度最快的电脑这一现实，而且中国的这一地位还有可能保持相当一段时间。
China's National University of Defense Technology last week informed visiting scientists about a massive machine in Changsha whose initial test results appear likely to top the next ranking of the 500 largest supercomputers, scheduled for release later this month.
中国的国防科技大学(National University of Defense Technology)上周告知来访的科学家们，中国一台巨型计算机的初次测试结果显示，它有可能在定于本月晚些时候公布的最新一期全球超级计算机500强排行榜上位列第一。
Such systems, used heavily by government agencies for purposes such as weapons design and intelligence-gathering, have long been seen a symbol of national competitiveness─and the focus of recent leapfrogging moves by the U.S., China and Japan.
巨型计算机主要被政府机构出于武器设计和情报收集等目的而使用，它们一向被视为国家竞争力的象征，世界运算速度最快计算机的宝座近年来一直被美国、中国和日本这三个国家交替占据。
A supercomputer in China unveiled in 2010 briefly topped the global speed ranking. But U.S. research institutions recently have fielded the fastest hardware, led by a machine called Titan at Oak Ridge National Laboratory.
中国2010年对外公布的一台超级计算机曾短暂成为世界运算速度最快的计算机。但是美国的研究机构又推出了更快的硬件，即田纳西州橡树岭国家实验室(Oak Ridge National Laboratory)的一台叫做“泰坦”(Titan)的计算机。
The new Chinese system─called the Tianhe-2, or Milkyway-2─uses two different kinds of computer chips from Intel Corp. INTC -2.60% as well as some homegrown circuitry, according to Jack Dongarra, a U.S. supercomputer expert who saw the system last week. He published details about the machine that were reported earlier by the publication HPCwire.
根据上周亲眼看到这一系统的美国超级电脑专家唐加拉(Jack Dongarra)说，中国开发的新系统叫做“天河2号”，使用了来自英特尔的两种不同电脑芯片以及一些国产的电路系统。他发表了这台计算机的具体信息，这些信息早些时候已经被HPCwire网站报道。
'It's a real system,' said Mr. Dongarra in an interview Wednesday. 'It seems to be ready for production.'
唐加拉周三接受采访时说，这是一个真正的系统。它似乎已经可以投入生产。
Mr. Dongarra, a professor at the University of Tennessee who is also on the Oak Ridge research staff, said Chinese officials put the system's theoretical peak performance at 54.9 quadrillion operations per second─or petaflops, in industry parlance. It reached 30.65 petaflops in a standard test known as Linpack, he said, compared with 17.59 petaflops for the Oak Ridge system.
唐加拉是田纳西大学(University of Tennessee)的教授，也是橡树岭国家实验室的工作成员。他说，中国官员称，该系统理论上的性能峰值是每秒运算54.9千万亿次。他说，在Linpack标准测试中可达到30.65千万亿次，相比之下，橡树岭实验室的系统可达到17.59千万亿次。
The U.S. has previously responded to such challenges, including a system from Japan that temporarily held the top speed ranking. But the next U.S. supercomputer that could challenge the Tianhe-2 isn't expected until 2015, said Horst Simon, another supercomputer expert who is deputy director at Lawrence Berkeley National Laboratory.
美国之前曾经回应过此类挑战，包括运算速度曾经排名第一的来自日本的系统。但是另一名超级计算机专家西蒙(Horst Simon)说，美国能够挑战天河2号的下一台超级计算机预计到2015年才会出现。西蒙是劳伦斯伯克利国家实验室(Lawrence Berkeley National Laboratory)的副主任。
'If anybody believes that the Chinese are just in there for a stunt, this machine has proven them wrong,' Mr. Simon said.
西蒙说，如果有人觉得中国人研究超级计算机只是噱头，这台计算机可以证明他们错了。
He argued that the U.S. government has invested sporadically in supercomputers lately, while Chinese agencies have spent steadily. 'In some areas where there has been a tradition of U.S. leadership, the Chinese are out there to really be in the race and win,' Mr. Simon said.
他说，美国政府最近对超级计算机进行了零散的投资，而中国的投资一直很稳定。西蒙说，在美国长期占据领导地位的一些领域，中国人已经在参与竞争并且取得成功。
No official price has been released for the Tianhe-2 system, which was constructed with help from the Chinese company Inspur. But some U.S. researchers put the likely cost at $200 million to $300 million.
天河2号系统的官方价格还没有公布。天河2号系统是在中国公司浪潮集团(Inspur Co.)的协助下建造的。但是美国的一些研究人员认为价格可能在2亿美元到3亿美元之间。
Tianhe-2 is a big endorsement for a new Intel product line. The Silicon Valley company's standard microprocessors have been used in supercomputers since the 1990s. But some big machines lately have been enhanced with chips called GPUs, or graphics processing units, from companies such as Nvidia Corp. NVDA -2.14%
天河2号是对英特尔一个新产品系列的巨大认可。自从上世纪90年代以来，这家硅谷公司的标准微处理器就一直被用于超级计算机中。但是一些大型计算机近来使用了叫做“图形处理单元”(GPU)的芯片，并因此得到性能上的提升，这些GPU来自英伟达(Nvidia Corp.)等公司。
Intel has responded with the Xeon Phi, a specialized chip with as many as 61 calculating engines. The new Chinese system has 48,000 of those chips, as well as 32,000 conventional Xeon chips based on a forthcoming design not yet generally available, according to Mr. Dongarra's report.
英特尔的回应是推出了Xeon Phi，这是一种拥有多达61个计算引擎的专业芯片。根据唐加拉的报告，中国新的系统拥有4.8万个Xeon Phi芯片，以及3.2万个传统的Xeon芯片，这些芯片基于一种即将推出的设计，这种设计目前还没有全面普及。
In addition, the system has about 4,000 chips developed in China that are based on a standard design called Sparc. Mr. Dongarra said Chinese officials appear determined to reduce their dependence on U.S. components.
除此之外，这一系统拥有大约4,000个在中国开发的芯片，这些芯片基于一种叫做Sparc的标准设计。唐加拉说，中国官员似乎下定决心减少对美国零部件的依赖。
He doesn't discount the possibility that government agencies in Japan will field a system to rival the Tianhe-2, but he regards the Chinese system as particularly impressive. 'In some sense, it's a wake-up call to the U.S.,' Mr. Dongarra said.
他不否认日本的政府机构也可能推出一个与天河2号相竞争的系统，但是他认为中国的系统格外令人印象深刻。唐加拉说，在某种意义上，这对美国是个警醒。