Sandia National Labs/Intel System First to Shatter Computing High-Performance Mark
SANTA CLARA, Calif., -- December 17, 1996 – The United States Department of Energy's Sandia National Laboratory and Intel Corporation, the world's largest computer chip company, have built a parallel supercomputer that has reached the one trillion- operations-per-second performance mark using the industry standard Linpack measurement method, the two organizations announced today. The achievement of 1.06 teraflops -- or trillions of floating point operations-per-second -- shatters the previous performance record of 368.2 gigaflops (billion-operations-per-second) by over 250 percent.
"Today's accomplishment is computing's equivalent to breaking the sound barrier," said Dr. Craig R. Barrett, Intel executive vice president and chief operating officer. "Just a few years ago, a teraflop was an intellectual barrier that nature dared us to cross. Now that we've surpassed that barrier, we have the computing horsepower needed to address the Grand Challenges of Science. We could be at the threshold of robust scientific discovery, triggered by access to teraflop-level computing performance."
The Grand Challenges of Science include issues in Applied Fluid Dynamics, Meso- to Macro-Scale Environmental Modeling, Ecosystem Simulations, Biomedical Imaging and Biomechanics, Molecular Biology, Molecular Design and Process Optimization, Cognition, Fundamental Computational Sciences, and Grand-Challenge-Scale Applications. Utilizing only three-fourths of the final supercomputer -- 7,264 of the planned 9,200 Pentium® Pro processors -- the Intel/Sandia Teraflops System is the first computer in the world to break the teraflops barrier.
The Intel/Sandia Teraflops Computer is currently under construction at Intel's Beaverton plant. The system will be installed in stages at Sandia National Laboratories in New Mexico in the first half of 1997. At completion, the computer will have 9,200 Pentium Pro processors and is expected to perform at sustained rates of 1.4 teraflops and peak rates of approaching two teraflops. Just 25 years ago, Intel introduced the world's first microprocessor, which delivered 60,000 instructions per second.
U.S. Secretary of Energy Hazel R. O'Leary said, "This achievement firmly re-establishes U.S. computer industry leadership in developing high-end systems. Four years ago the U.S. government, industry and academia set a goal. At that time it was not clear how or even if a trillion-operations-per-second computer could be achieved. Now thanks to U.S. innovation, it's not only possible, it's being done."
"The Intel/Sandia teraflop computer is built from commercial, off-the-shelf products and technologies including the same Pentium Pro processor in many of today's workstations and servers," said Ed Masi, general manager and vice president of Intel's Server Systems Products Division. "Using commercially available technology has enabled the government to utilize the R&D muscle of the marketplace, focusing tax dollars on combining these standard building blocks into the world's most powerful computer."
The Linpack measurement method is the most widely recognized single benchmark for measuring sustained floating-point performance of high-end computers. It gives an accurate picture of the performance of a given computer on applications that require the solution of large, dense linear systems – a category that includes a very wide range of technical applications.
* Other names and brands may be claimed as the property of others.