Statement of

Dr. Victor Reis
Assistant Secretary for Defense Programs
U. S. Department of Energy

Mr. Chairman and members of the Subcommittee, thank you for the opportunity to testify at your hearing on the sale or transfer of supercomputers to foreign entities or governments engaged in nuclear weapons research. Like other members of the President's national security team and this committee, I am concerned about the potential impact these sales might have on U.S. national security. Accordingly, I would like to use this opportunity to put the potential impact of these sales in the context of the stockpile stewardship and management program, a program which relies heavily on supercomputers.

As you know, the President has directed the Department of Energy's Office of Defense Programs to maintain the nation's nuclear weapons deterrence under a comprehensive test ban through the Stockpile Stewardship and Management Program. This program relies heavily on advanced simulation capabilities underpinned by advanced experimental facilities and archived test data.

The pace of our computing effort, known as the Accelerated Strategic Computing Initiative (ASCI), has been set by two factors. First, the simulations must be powerful enough to approach the complexity of "real" weapons; "test­based" designers and analysts must be able to compare their experiences, previous tests and new experiments to the simulations. This requirement for simulation fidelity tells us, among other things, just how fast the computers must be. We estimate that this capability must be of the order of two to five thousand times faster than the fastest production supercomputer available when we started the program. But because experienced weapons designers who have an intimate knowledge of the systems gained by underground testing will retire over the next decade we must reach this value while they are still active. This means we cannot simply wait for the normal, commercially driven improvements in computer speed but must enter into an aggressive program of partnerships with U.S. computer companies and U.S. universities. The computer industry will provide the commercially viable building blocks and the additional engineering to scale these computers to 100 trillion floating point operations per second (TeraFLOPS) and beyond. Universities will provide innovative approaches to achieving large scale, multi­disciplinary simulation as well as validation of the simulations on non­weapon related problems.

We are making excellent progress in meeting our simulation goals. Last December, the Department of Energy with Intel announced breaking the one TeraFLOPS barrier; it is now the worlds fastest computer by a factor of three. This computer is now installed at Sandia National Laboratories and is being used by all three weapons laboratories to run three­dimensional, complex physics simulations needed for stockpile stewardship. Last August we began operating the first phase of what will be a three teraflop IBM computer at Lawrence Livermore National Laboratory, and last November a four teraflop computer from Silicon Graphics Cray Division began installation at Los Alamos National Laboratory. Weapon designers are already seeing and starting to understand effects in the weapons through simulation that they have never been able to explain in the past.

How do these computers compare with those which were allegedly exported to Chelyabinsk?

As I understand it, the machines that we are now using at the one teraflop region are over 300 times faster than the exported SGI machines. And we are planning to have within the decade computers some one hundred times faster than the machines we are now using to support the stockpile stewardship program.

In summary, while potential violations of U.S. export laws are a subject of serious concern, we need to understand these concerns in context, particularly in relation to our own programs. We suggest that should sanctions or punishments for proven violations be considered, such measures be carefully structured so as not to undermine the U.S. stockpile stewardship program.