12 Trillion, and IBM's Counting

IBM announced today that they have built the world's fastest supercomputer, capable of 12 trillion calculations per second and more than three times faster than the most powerful computer in existence to date.

PRESS RELEASE:


Known as ASCI White, the RS/6000 SP supercomputer covers an area the size of two basketball courts and will be used by the U.S. Department of Energy (DOE) program to help ensure the safety and reliability of the nation's nuclear weapons stockpile without real-world testing.

In testing just completed at IBM facilities in Poughkeepsie, N.Y., ASCI White demonstrated a record computational capability of 12.3 teraflops (trillions of operations/second) — exceeding the performance requirements of IBM's pioneering contract with the DOE by 23 percent. The system is scheduled to be fully installed at the DOE's Lawrence Livermore National Laboratory in California by the end of the year. The Laboratory is operated for the DOE by the University of California.

ASCI White marks a breakthrough in computing. At 12.3 teraflops, the IBM-designed RS/6000 SP system is the first computer to exceed the double-digit teraflop speed barrier. It will be used by the DOE to develop complex 3D simulation tools for use in supporting nuclear stockpile stewardship efforts.

The system, developed by IBM under the DOE's ASCI (Accelerated Strategic Computing Initiative) Partnership, is powered by 8,192 copper microprocessors, and contains six trillion bytes, or terabytes (TB), of memory with more than 160 TB of IBM disk storage capacity — enough to hold six times the entire book collection of the Library of Congress. Delivery of the ASCI White system from IBM facilities in Poughkeepsie, N.Y. to DOE's Lawrence Livermore National Laboratory in Livermore, California will require 28 tractor trailer trucks.

"This is a significant achievement," said U.S. Secretary of Energy Bill Richardson. "Once this system is fully installed it will be an important tool for stockpile stewardship."

"This level of computing power has never been achieved anywhere," said David M. Cooper, associate director for computations and CIO at Lawrence Livermore. "It will open new horizons in scientific computing, as we approach our goal to simulate the aging and operation of a nuclear weapon. This is the second time in our partnership with IBM that they have exceeded contract performance specifications in the delivery of a major supercomputer system."

"The completion of the ASCI White system is a significant milestone for the Department of Energy and for the science of computer simulation," said Nicholas Donofrio, IBM senior vice president and group executive, technology and manufacturing. "The selection of IBM and the RS/6000 SP for a project of this scope and national urgency is a reflection of our unique ability to provide large scale computational power to solve the most demanding business and scientific applications — what we call 'Deep Computing.'"

The DOE's Stockpile Stewardship ASCI project was developed in response to a directive from President Clinton. It integrates the efforts of the three DOE national laboratories: Livermore, Los Alamos and Sandia.


The ASCI project calls for a series of supercomputers — 1, 3, 10, 30 and 100 teraflops in size — to be built over a period of several years.