NEWS RELEASE, 8/19/97

Intel Corp. donates $6 million worth of computer equipment to enhance research and teaching at UC Berkeley

Berkeley -- Six million dollars worth of computer equipment will begin arriving on the UC Berkeley campus this month, thanks to a grant from Intel Corp. that will make high-performance parallel computing available to a broad range of departments that have until now had scarce or no access.

The grant was part of Intel's Technology for Education 2000 program, a three-year, $85 million grant program in support of higher education. Intel announced grants of up to $6.2 million to each of 12 universities Monday (8/18) in a first-round selection totaling $67.2 million.

The goal of TechEd 2000 is to bolster university research and curriculum development and help place PCs, workstations, servers and networking hardware based on Intel Architecture in key research universities throughout the United States.

"This vastly improves the computing infrastructure for research and education," said James Demmel, professor of computer science and mathematics and principle investigator for the UC Berkeley project, called Millennium. "It will enable high-performance computing in many departments that previously only dreamed of it."

With fast parallel computers at their disposal, UC Berkeley scientists will conduct more simulations of real world events as an adjunct to theory and experiment. Among those to benefit are business professors interested in complex financial computations, biologists creating evolutionary family trees, astrophysicists modeling star formation, and information specialists creating "digital" libraries for the next century.

"The computing will allow simulation of events that we can't afford to do experimentally," Demmel said.

Demmel and co-principal investigator David Culler, professor of computer science, will take the Intel hardware -- primarily Pentium processors, the guts of high-performance desktop computers -- and develop the hardware and software to build a range of computers for the campus. These include desktop computers for individual users, small 20-processor parallel machines for departments, and one large 250-processor parallel computer to be shared by researchers across the campus.

In addition, they and other computer science faculty members will supply parallel computing software developed at UC Berkeley.

For its part, the campus has agreed to provide $1.5 million of its own, primarily in staff hours, to assemble the computers, develop software and maintain the networks.

Parallel computers are arrays of individual computers that work together to solve a problem, splitting the computation among each computer and continually sharing data and information. Standard, non-parallel computers solve problems sequentially, in a linear, step-by-step process.

One of the great advantages of parallel computers is that they are cheaper than the large mainframes and supercomputers, allowing complex computing for less money.

Computer scientists at UC Berkeley have pioneered a unique parallel architecture called NOW (for Network of Workstations) that until now has been operating with Sun machines using the UNIX operating system. Culler says the NOW architecture will be adapted to work with Intel's Pentium processors using either UNIX or Microsoft's NT operating system.

A significant part of the Millennium project will be conducted in collaboration with Lawrence Berkeley National Laboratory's National Energy Research Scientific Computing Center. For example, the center will adapt sophisticated software tools from its library for use on the campus system.

"One of the most exciting aspects of Millennium is the potential to demonstrate that you can take a large number mass-market commodity PCs and networks, harness them together with some special software, and get a powerful supercomputer for less than the current prices of supercomputers," said Bill Saphir of the center's Future Technologies Group.

Some of the innovative uses include:

• Professors in the Haas School of Business hope to use the new parallel computers to keep pace with and actively participate in the explosive growth in the mathematical and computational sophistication of the problems encountered in today's business world. These are encountered in the areas of financial derivatives, risk management and money management, including such problems as simulating or optimizing complex financial portfolios.
• The new National Center of Excellence for Aviation Operations Research, established last year at UC Berkeley to explore the redesign of the nation's air traffic control system, will use a network of workstations to simulate the National Airspace System as it would operate under conditions of "free flight." Free flight, an air traffic scenario planned for the 21st century that would allow pilots to make decisions now made by air traffic controllers, could save up to $3 billion annually in jet fuel costs.
• UC Berkeley's Digital Library project (one of six nationwide) will use a dedicated NOW for a nationally accessible Web server that would provide high speed access to large image databases.
• Astronomers will expand their work in numerical simulations of star formation. One project involves simulating the collapse of giant molecular clouds into massive stars, using new data such as that from the Hubble Space Telescope.
• Biologists in the Department of Integrative Biology hope to exploit parallel computing to reconstruct the evolutionary history of green plants, mollusks and other organisms. Knowledge of the relationships among species and the ability to predict such relationships has become vital to modern organismal biology. Although recent advances have made reconstruction of such family trees commonplace, the computational expense of reconstructing these relationships has prohibited research from keeping pace with the amount of new genetic data.
• Researchers in the Department of Nuclear Engineering, in collaboration with scientists at Lawrence Berkeley National Laboratory, will use the computers to develop a way to calculate the optimum protocol for a new type of brain cancer treatment, boron neutron capture therapy. High-performance computers are necessary to handle the data from CT, MRI and PET scans, and to calculate the optimum treatment plan and dose.