China's Not-So-Super Computer Program

By BOB DAVIS

SHENZHEN, China?The supercomputer in this southern boomtown is named Nebulae for the interstellar clouds of gas that give birth to stars. The machine symbolizes China's soaring ambition to challenge the U.S. and other developed nations in technology, but also underscores the limitations of what China can achieve.

China's unexpected progress in developing supercomputers, the brains of modern science and an engine of economic development, has caused an outbreak of anxiety over the past two years in the U.S., which has long been the field's undisputed leader. China's advances in computing have been critical to its ability to build spacecraft and advanced warplanes and to its growing prowess in genetics. China overtook Japan in 2010 as the number two investor in research and development, according to Battelle Memorial Institute, a Columbus, Ohio, research outfit. Though China remains well behind the U.S. in R&D spending, it is gaining ground.

But a closer look at China's supercomputers reveals a program that is far less of a threat to U.S. technological dominance than commonly believed. Chinese researchers say decisions about how supercomputers are used are often made by local politicians more interested in local development projects than breakthrough technology.

Associated Press

The Tianhe-1A supercomputer at the National Center for Supercomputing in Tianjin, China.

China's bureaucrats meanwhile haven't figured out how to mount software development projects that come close to U.S. or European standards. Chinese scientists also lack the funding, and freedom, to explore technologies that haven't already been endorsed by the government, which can keep them well behind the cutting edge.

The result is that China's supercomputing projects aren't producing the kinds of breakthroughs that can create new industries. It is instead being deployed to help the country simply catch up with the U.S. and Europe, in areas ranging from health care to automotive design to aviation. That is important economically, but it is also a reminder that China remains a developing country whose main goal is to close the economic and technology gaps with richer nations.

"The strategy has been never to lead, but to follow" technologically, said Qian Depei, a Beihang University researcher, who has worked for decades on China's advanced computing programs. "That was the most economically efficient way to develop."

Richard Suttmeier,a University of Oregon expert on Chinese science policy, said China hasn't figured out "the right formula" to pioneer new technologies in part because researchers are rewarded according to the number of academic papers they publish rather than the quality and novelty of their work.

Supercomputers are largely seen in China as local economic-development tools. City governments play a much larger role in setting China's supercomputer research agenda than they do in the U.S. because Chinese cities finance a larger share of the projects.

Shenzhen, which paid three-quarters of the $1.3 billion cost of the Shenzhen supercomputer center, "doesn't care about climate change and astrophysics"?traditional supercomputer research projects?said Feng Shengzhong, deputy director of a Shenzhen research institute that develops applications for the Nebulae. "They care about local problems."

He is working on a plan to use the Nebulae to improve health care services in South China?a socially important goal but not one that makes use of the power of what is ranked as the world's fourth-fastest supercomputer.

China is now home to 74 of the world's 500 fastest supercomputers?which can make trillions of calculations every second?up from just 10 in 2007. Changes in the way the machines are designed have helped the country.

In the 1980s, when Cray Research in Minneapolis was the world's supercomputer technology leader, the machines were powered by a few enormously powerful processors, whose design was difficult to match. Exports were tightly controlled. Starting in the 1990s, supercomputer researchers began to lash together tens of thousands of off-the-shelf microprocessors to work on a single job. China could buy those computer chips from Intel Corp., Advanced Micro Devices Inc. and other firms and make its own machines.

Analysts say that Chinese scientists have benefited from training at top computer centers in the U.S. and Europe and the availability of computer chips and other parts from abroad, as well as consistent Chinese government funding and support. The supercomputer effort isn't dogged by charges that the Chinese have ripped off foreign technology. Rather, Chinese scientists say U.S. restrictions on some high-technology exports have required them to redouble their domestic efforts.

Still, China remains largely dependent on U.S. made microprocessors?the brains of the computer?which puts it behind the leading edge. Beijing has developed one computer that uses locally designed microprocessors, but doesn't run commercially available software. Another microprocessor in development, the Loongson, would use existing software and could eventually become a competitor to Intel and others.

Some of China's supercomputers have been used to design wings for China's stealth fighter, now in test phase, and to design parts for China's first commercial jet. Beijing lags well behind the U.S. in both efforts.

In Shenzhen, the Nebulae, which is still being tested, is expected to improve storm warning systems and help genetics companies search for the causes of disease. But it is also scheduled to be used for far less demanding tasks, such as processing video animation.

In the U.S., said Steve Conway, a supercomputer analyst at market researcher IDC, in Framingham, Mass., cities and states chip in money for local supercomputer centers, but they generally have little say in setting the priorities for the machines. U.S. supercomputer centers nearly always focus on advanced scientific research, such as designing drugs tailored to individuals.

Research at the edge of technology is risky, but can have big payoffs and leave competitors like China well behind. "American alarmism isn't always well-founded," said Mr. Suttmeier, of the University of Oregon. "The critical point is keep devising strategies in the U.S. to stay way ahead of the game."

One of China's greatest weaknesses is in software development, a potentially crippling problem because the usefulness of the machines depends on the quality of the software applications. Less than 10% of supercomputing funding goes to developing such applications, said Chinese researchers who complain that political leaders press them to build headline-grabbing new machines rather than focus on whether they are used to their full capabilities.

In the U.S., which spends about six times as much on supercomputers as China, the software budget equals about 30% of hardware spending, and computer specialists say even that level isn't sufficient.

The battle for software dollars is so intense in China that researchers rarely work as a team on long-term software projects, Chinese scientists say.

To illustrate the uneven perception of China's supercomputer efforts, Mr. Qian, the veteran supercomputer researcher, holds his palms at hip level. "Generally, we're here," he said, "but everyone thinks we're higher," as he raises his palms to shoulder height.

Write to Bob Davis at bob.davis@wsj.com

Source: http://online.wsj.com/article/SB10001424052702303812904577298062429510918.html?mod=rss_about_china

johnny weir quadrantid meteor shower osu football osu football oklahoma state santonio holmes raheem morris