APPLICATIONS

SEATTLE-- Global supercomputer leader Cray Inc. today announced the Cray X1(TM) system, the world's most powerful supercomputer product. The new product is available with up to 52.4 trillion calculations per second (teraflops) of peak computing power and 65.5 terabytes of memory. U.S. list pricing starts at about $2.5 million. The high-efficiency, extreme-performance system is aimed at the critical computing needs of classified and unclassified government, academic research, and the weather-environmental, automotive, aerospace, chemical and pharmaceutical markets. Cray also announced today that it is the first company to accept the challenge, as stated in a 1999 report of the President's Information Technology Advisory Committee, to provide actual, sustained (not merely "theoretical peak") petaflop computing speed -- 1,000 trillion calculations per second -- for critical next-generation applications by 2010. Along with previously announced initial orders for Cray X1 systems, Cray expects substantial additional orders in the year-end 2002 timeframe. The company recently reported that five early-production Cray X1 units passed acceptance tests at the U.S. Army High Performance Computing Research Center (AHPCRC) and undisclosed customer sites, and that Spain's National Institute of Meteorology (INM) placed an $8.4 million, multi-year order for a Cray X1 system. Cray plans to ship the first production system later this year, with the manufacturing ramp-up scheduled to begin in early 2003. The AHPCRC is the first site in the DoD High Performance Computing Modernization Program to acquire a Cray X1 system. AHPCRC scientists already have implemented widely used weather forecasting, computational mechanics, and computational fluid dynamics application codes on their two early production systems. In August 2002, the U.S. Department of Energy announced that Oak Ridge National Laboratory (ORNL) had been selected to test the effectiveness of the Cray X1 system in solving important scientific problems in climate, fusion, biology, nanoscale materials and astrophysics. Dr. Raymond L. Orbach, director of the department's Office of Science, said the program is part of an effort to provide the U.S. scientific community with computing resources to match or exceed those of the new Japanese "Earth Simulator," which has an effective speed more than 20 times that of the fastest U.S. civilian supercomputer. Under the program, ORNL initially will acquire a 32-processor Cray X1 supercomputer for evaluation. "The Cray X1 system is designed to scale to deliver performance for key scientific applications greater than the performance of currently available U.S. computers. It is the first U.S. computer to offer vector processing and massively parallel processing capabilities in a single architecture," said ORNL's Associate Director for Computing and Computational Sciences, Thomas Zacharia. "We are pleased to partner with Cray to provide U.S. scientists with the computational power to solve the most demanding science applications." Sustained Petaflop Goal "The Cray X1 product is a major milestone en route to our goal of enabling the world's most creative scientists and engineers to run challenging applications at sustained petaflop speed by the end of this decade," said Cray Chairman and CEO James Rottsolk. "Our product roadmap is focused on this commitment. Other key initiatives, including our previously announced $90 million contract to develop and deliver the 'Red Storm' supercomputer to Sandia National Laboratory, our DARPA HPCS contract for advanced research on trans-petaflops computer systems, and further experience with our multithreaded MTA-2(TM) system will accelerate our progress toward this goal." Rottsolk said that in meetings such as the Petaflops I (1994) and Petaflops II (1999) conferences, experts identified crucial problems requiring petaflop speed. "These problems range from basic science to severe weather forecasting, full-fidelity automotive crash testing, advanced aircraft and spacecraft design, rapid detection of wildfires, virtual surgery planning, national economic modeling, combating pandemics and bioterrorism, improving electrical power generation and distribution, advanced drug discovery methods, and many more." The Cray X1 system will allow users to begin addressing these next-generation applications, he said. World's Most Powerful Supercomputer Product With innovative processors more than four times faster than in today's typical cluster products, and seven to 40 times more memory bandwidth for solving big, complex problems efficiently, the Cray X1 supercomputer promises to help advance the frontiers of science and engineering across a broad spectrum of disciplines. Cluster products are collections of individual servers or PCs, rather than tightly integrated supercomputers, and are adept at handling smaller, less complicated problems. "Current 'clustered' systems provide a low initial purchase price and can be broadly deployed, but they do not necessarily provide the highest levels of performance for next-generation scientific and engineering problems," said Debra Goldfarb, IDC group vice president, Worldwide Systems and Servers. "The Cray X1 is masterfully designed to address these strategically important requirements and, based on early indications, has the potential to return Cray to leadership in the high-end supercomputer segment." The new product features ultra-fast (12.8 gigaflops) individual processors, up to 819 gigaflops (billions of calculations per second) of peak computing power in a single chassis, and a high-bandwidth, low-latency interconnect for substantially more efficient scaling than on clustered SMP systems. Cray X1 systems are available with 8-4,096 processors (2-1,024 nodes) and 32 gigabytes-65.5 terabytes of memory. "The Cray X1 supercomputer is the successor to Cray MPP and vector products," said Steve Scott, chief architect of the Cray X1 system. "Programmers with experience using clustered systems will also find the Cray X1 environment familiar: MPI, cache-based, distributed memory, 32- and 64-bit IEEE support. Shmem, Co-array Fortran and UPC programming models are available for leading-edge work benefiting from the very low latency and communication overhead of true shared memory." Broad Applicability "It is not just large government projects or major university research sites that desire an alternative to standard parallel systems built from off-the-shelf processors," according to Rich Partridge, Enterprise Systems analyst at D.H. Brown Associates. "Automotive design/analysis, biotech research such as protein folding, petrochemical seismic analysis, and other compute-intensive fields eagerly await the arrival of Cray's new high-bandwidth, low-latency X1 design. Cray has not succumbed to the attack of the low-priced commodity providers, but instead is in the process of re-establishing its leadership as the premier high-performance computing provider." "A lot of people have argued that cluster systems are cheaper than semi-custom systems such as the Cray X1. The fact of the matter is, they're not cheaper," said Paul Muzio, AHPCRC Support Infrastructure Director. "When we did our total life-cycle cost analysis, it turns out that clusters are fairly expensive -- a lot more expensive than people would think, especially when you look at the relatively low level of sustained performance that you get out of the processors on a cluster system, and compare that with the higher levels of sustained performance on the Cray X1. The Cray X1 will allow AHPCRC, Army, and DoD researchers to address a set of key problems in computational mechanics, battlefield weather forecasting, and biological process modeling that are of great importance to the defense of the United States." Rottsolk said he also expects the Cray X1 product to benefit the automotive and aerospace industries. "For example, the computational capability of the Cray X1 system is very applicable to the types of problems Ford Motor Company is trying to solve. Ford has utilized Cray capability-class systems in the past, and we look forward to partnering with Ford in the future on their most challenging problems." "In the next eight years, the Cray X1 and its successors are going to enable users for the first time to do very, very complex simulations that are biologically and medically relevant," said Dr. Frederick Hausheer, chairman and CEO of BioNumerik Pharmaceuticals, Inc. As part of a previously announced agreement, Germany's Konrad-Zuse-Zentrum fur Informationstechnik Berlin (ZIB) and Bielefeld University will be given access to the Cray X1 system to develop advanced software solutions for some of the most challenging problems in the new field of bioinformatics. "The new Cray X1 system will dramatically increase the computing capability of INM for operational numerical weather prediction and for research in climate and atmospheric modeling. The new system will play a major role in the improvement of short and very short range forecasting in Spain in coming years," said Ricardo Riosalido, Head, Observations and Instruments at INM. Ohio Supercomputer Center helped Cray evaluate I/O node technologies and data archiving tools for the Cray X1 product. Cray recently announced an agreement for Iowa State University to test software products developed by the company for the Cray X1(TM) system, and that it will use the StorNext® File System from Advanced Digital Information Corporation (Nasdaq:ADIC - News) as the primary SAN file system for the new supercomputer. Separately today, Samsung Semiconductor, Inc., announced that Cray selected Samsung RDRAM as the memory technology for the Cray X1 product.