UCSD Supercomputer Nips at the Frontiers of Science : Mathematics: Device processes equations impossible to solve with pencil and paper, and helps the transfer of military technology to the private sector.

It’s been seven years since physics professor Herbert Hamber motored down from Irvine to become the first user of the Cray XMP supercomputer at UCSD’s San Diego Supercomputer Center.

Just days after crews gently moved the powerful computer off the loading dock and into its climate-controlled setting, Hamber was pumping complex physics equations into the Cray, which performed 100 million calculations per second.

The UC Irvine physics professor no longer has to drive to San Diego to use the center. Like nearly 2,000 other users nationwide, Hamber simply hooks into the Cray via a telephone modem.

The Cray XMP leased by the center in 1985 has been replaced by a faster and more powerful Cray YMP supercomputer. And, if a planned expansion occurs, Hamber will soon be hooking up with a C90, Cray’s newest supercomputer.

Despite his years of experience with Crays, Hamber is still in awe of them.

“We’re working with more and more complex (physics) equations that, for the moment, we have no clue as to how to solve with pencil and paper,” Hamber said. “We don’t even see a way to do (these problems) the old-fashioned way. . . . The pencil-and-paper approach has failed to produce a single number.”

The National Science Foundation had civilian researchers such as Hamber in mind in the early 1980s when it agreed to fund supercomputer centers at UC San Diego, Princeton University, the University of Illinois and Cornell University.

At that time, a handful of supercomputers were available only to scientists with security clearances at military research facilities. NSF officials feared that the nation’s competitive edge would erode unless supercomputers were more readily available to academic and industrial researchers.

Since 1985, an estimated 15,000 researchers have used center supercomputers, most of them linking up through personal computers and simple desktop programs.

“We’ve done it,” said Larry Smarr, director of the Illinois center. “We played a historic role in the transfer of technology from the military to civilian sector.”

Civilian researchers are using the powerful machines to delve into some of nature’s most complex secrets.

“People are trying to understand the structure of the universe,” said Wayne Pfeiffer, deputy director for research of the supercomputer center at UCSD. “At the very opposite end, people are . . . trying to determine the basic properties of fundamental particles.”

One of the highest-profile uses of the supercomputer in San Diego came in 1991, when a group of researchers used the center’s computational power to make the first “map” of a protein’s structure. A computer-generated likeness created by the supercomputer center’s $6-million visualization laboratory was featured on the cover of Science magazine.

Some researchers are using the supercomputer and the visualization lab to determine how smog travels through the atmosphere over the vast Los Angeles basin, while others use the center to model weather data for the whole of Western North America.

Industrial clients have used the center in the study of molecules, which could one day lead to the treatment of potentially deadly diseases. Solar Turbines of San Diego is using the computers to design better turbine blades.

Supercomputer research is moving forward quickly because civilian researchers now enjoy access to operating programs and software developed by the military since the late 1970s. Hamber and other researchers tap into the supercomputer through relatively simple programs that operate on personal computers at home and the office.

But researchers won’t find it as easy to use “massively parallel machines,” the next generation of highly powerful computers that is now making its way into the supercomputer centers.

The machines are called massively parallel because they use many processors, speeding computations by tackling separate parts of problems at the same time. In contrast, traditional supercomputers perform billions of calculations in sequence, a slower process.

Initial research on the parallel machines will be slowed by the lack of operating programs and software. Existing programs are complicated and not as useful as the MacIntosh and IBM-compatible programs that physicists, physical scientists and other researchers use at home and the office.

But, as researchers harness the power of massively parallel machines, researchers will start to “explore the grand challenges of computational science and engineering,” Smarr said. “We’re going to explore problems of fundamental importance to basic research, industrial competitiveness and society as a whole.”

Massively parallel computing machines “are potentially 10 times more powerful than the Cray YMP we have now,” said Sid Karin, director of the San Diego Supercomputer Center since its inception.

But power isn’t the parallel machine’s only allure. The devices are also dramatically less expensive than traditional supercomputers.

Karin is excited by the potential the parallel computers offer.

“We’re not only going to be at the leading edge with these machines . . . we’re going to be pulling that edge back,” he said. “That makes it very exciting.”

In an era of tight budgets, the San Diego Supercomputer Center is struggling to keep pace with technological changes sweeping through the supercomputer industry.

The center will lease its $25-million Cray YMP from the Minneapolis-based manufacturer, but it is purchasing a new parallel machine--it eventually will have three--from Santa Clara-based Intel for less than a tenth of the cost of a traditional supercomputer.

The center will take delivery of its third parallel processing machine later this month. Last week, center officials began negotiating lease terms for a C90, Cray’s latest-generation supercomputer. And, because the bigger and more powerful machines generate ever-increasing data streams, the center is spending $1.7 million for additional computer memory capacity.

Despite the state’s bleak economy and budget woes, UCSD officials last month initiated a search for an architect to design a $2-million expansion at the three-story building that opened in 1986.

Construction funding has not yet been approved, but the center hopes to build a 12,000-square-foot expansion to house 35 researchers, faculty and teaching assistants, along with a new graphic-visualization classroom.

Karin said the addition is needed because the center’s only building is overcrowded as record numbers of researchers seek time on the supercomputer center’s machines.

Because demand is growing and time is limited, less than half the time requested by researchers is being granted. In many cases, scientists make do with just a portion of the time requested.

Only 10% of the scientists who logged on to center computers in October did so from the UCSD campus. The remainder use telephone modems to reach the center from personal computers around the continental United States and Hawaii.

But top-ranked academics aren’t the only ones using the supercomputer center.

Younger research assistants are given time on the machines with the expectation that one day they will use their computing skills in academia or industry. Hamber recently created a “computational physics” class at UC Irvine that teaches undergraduates to use supercomputers.

Industrial clients, including Rohr and Solar Turbines of San Diego, are using the center’s machines to develop everything from improved turbines to new drugs.

The National Science Foundation allowed centers to lease time on computers to industrial customers, in part to spread supercomputing to industry and in part to generate revenue for the centers.

Budgetary cuts have forced the Illinois center to shrink its staff by 25%, but the San Diego center has continued to grow. “We didn’t grow as rapidly as (Illinois) did in the past, so we’ve had no need to cut back,” Karin said. “Our growth has been more continuous and slower.”

The center has a $20-million annual budget and a full-time staff of about 100. When it opened in late 1985, the center had about 60 employees and a $15-million annual budget.