originally appeared in the August 19 edition of ORNL Today
Lab key in drive to regain supercomputing lead

Rep. Zach Wamp (left) listens to Dr. Orbach's scientific
computing plan. |
ORNL will be a prime proving ground and centerpoint for a DOE,
industry- and university-wide initiative to reestablish U.S.
supremacy in scientific computing, Office of Science Director Ray
Orbach announced Thursday afternoon at a reception at the Oak Rdge
Chamber of Commerce.
The program, Dr. Orbach says, will be on a scale with the
Spallation Neutron Source, both in size and importance to the
nation. Dr. Orbach, Battelle CEO Carl Kohrt and a host of DOE
officials are in town this week for ORNL's annual on-site review.
The first step in the initiative is occurring at ORNL-the Lab will
test the effectiveness of a new Cray, Inc., supercomputer
architecture by acquiring a 32-processor Cray X1 supercomputer
system. The Cray is the first U.S.-made supercomputer to offer
vector and massively parallel processing in a single
architecture.
Japan leaped ahead of the field last April with its Earth
Simulator supercomputer. It operates with 50 percent efficiency at
40 teraflops. U.S. supercomputers currently operate at a maximum
10 percent efficiency at 12 teraflops. Dr. Orbach explained in a
news conference Thursday that the Japanese accomplished their feat
by designing a computer strictly for science. The U.S. hasn't used
that approach.
"Our computers aren't built for science; they are built for
commercial uses," Dr. Orbach says. "It's one size fits all."
Dr. Orbach emphasized, as he did in Wednesday's "First Light"
ceremony, that scientific computing-simulation-has taken on equal
importance with theory and experimentation in the scientific
process. That's why losing ground to an overseas nation causes
such concern.
"I cannot overemphasize its severity," Dr. Orbach says. "I am
doing everything I can to make it an administration priority."
He predicted it will take four to five years to catch up with the
Earth Simulator. The price of the program over that time,
involving five national labs, industry and universities, will be
around $1 billion, similar in scope to the SNS. The initial '03
funding for the Cray project is in the $2-3 million range. Rep.
Zach Wamp pointed out that ORNL's reputation for science and
leadership plays a key role in bringing much of the work at the
Lab.