GOVERNMENT

System Built on Cray XT Infrastructure Represents a Milestone for Cray's Adaptive Supercomputing Vision: Cray today announced the availability of its next-generation massively parallel processing (MPP) system, the Cray XT4 supercomputer. The powerful new supercomputer, previously code-named "Hood," is designed to easily and efficiently scale to a peak performance of more than one petaflops (1,000 trillion floating-point operations per second). The Cray XT4 supercomputer debuts with several large system orders announced earlier this year from leading organizations, including the Oak Ridge National Laboratory (ORNL), the National Energy Research Scientific Computing Center (NERSC) and the Finnish IT Center for Science (CSC). The new Cray XT4 system is designed for superior scalability, incorporating the latest generation of AMD Opteron processors matched with expanded local and interconnect bandwidth that provides the industry-leading system balance necessary to optimize application performance at scale. The system is equipped with AMD Opteron dual-core processors that can be easily upgraded to AMD's quad-core processing technology in the future to deliver balanced petaflops performance. "The Cray XT4 system is a crucial element in our ongoing partnership with Cray, which calls for us to increase the performance of our Jaguar system to 100 teraflops by the end of this year, and again to 250 teraflops in 2007 or early 2008," said Thomas Zacharia, associate laboratory director, Oak Ridge National Laboratory. "Our users are experiencing sustained performance well in excess of 10 teraflops on a range of applications from Astrophysics/Cosmology, Fusion Science and Materials Science to name a few on our current 54 teraflops system. With our Cray XT4 system upgrade, we are on the path to a system that will eventually perform significantly faster on a broad spectrum of important scientific applications." The Cray XT4 supercomputer is the first product in Cray's Rainier program, the initial phase of the company's Adaptive Supercomputing vision. The Rainier program will integrate all of Cray's current processor technologies onto a common Cray XT infrastructure, allowing a diverse and challenging application workload to run on a single, highly scalable infrastructure without compromising performance or usability. Evolved from the successful Cray XT3 supercomputer, the Cray XT infrastructure provides a common, scalable environment for login, compilation, resource management, work scheduling and I/O. This environment also includes a unique globally shared, high-performance parallel file system, as well as network interfaces to other systems. "Massively parallel processors play and important role in the high-performance computing market," said Dr. Earl Joseph, IDC program vice president, Technical Computing Systems. "Their internal systems interconnects are designed to provide significantly higher bandwidth and lower latencies relative to general purpose networking technologies often used in clusters. This stronger internal performance can allow MPP systems to address a broader range of applications with greater levels of scalability than commercial off the shelf products." "While the theoretical peak speed of supercomputers may be good for bragging rights, it is not an accurate indicator of how the machine will perform when running actual research codes which is what our 2,500 users are most interested in," said Horst Simon, director of the NERSC Division at Berkeley Lab. "To better gauge how well a system will meet the needs of our users, we developed SSP, a sustained system performance benchmark suite. Under this real-world performance test, the new Cray XT4 system will deliver over 16 teraflops on a sustained basis." Architected for the Future The Cray XT4 supercomputer uses up to 30,000 AMD Opteron dual-core processors running a highly scalable operating system and interfaced to the Cray SeaStar interconnect chip to provide unsurpassed scalability and performance. Unlike typical cluster architectures, in which many microprocessors share one communications interface, each AMD Opteron processor in the Cray XT4 system is coupled with its own interconnect chip. Providing six links in three dimensions, the unique Cray SeaStar2 chip uses its embedded routing capability to take advantage of HyperTransport technology and significantly accelerate communications among the processors. "After an extensive selection process we chose the Cray XT4 supercomputer to replace a cluster system that could no longer keep up with the computing demands of our research groups," said Kimmo Koski, managing director of CSC, the Finnish IT Center for Science. "We determined that the Cray XT4 system matched our needs by delivering the best combination of performance and value." "Unlike commodity clusters, the Cray XT4 supercomputer is built from the ground up to provide a scalable, balanced system in order to support the most demanding applications," said Cray President and CEO Peter Ungaro. "The Cray XT4 system builds upon our highly successful Cray XT3 supercomputer in almost every aspect, providing enhanced scalability, performance and reliability in a system that is easily upgradeable to protect a customer's investment for years to come." Key Advantages for Cray Customers The Cray XT4 supercomputer offers customers significant advantages: -Advanced MPP architecture;the system is capable of peak petaflops performance, scaling upward to as many as 30,000 AMD Opteron processors. The design delivers balanced performance using single-socket processor nodes with local memory and direct interprocessor communications, as well as a low-jitter operating system. This dramatically reduces the bottlenecks and connection difficulties associated with large cluster systems built from off-the-shelf commodity server components. -High-bandwidth communications - the second-generation Cray SeaStar2 interconnect chip utilizes HyperTransport™ technology and built-in routers to boost bandwidth and reduce latency for data moving among processors or fetched from memory. This high-speed interconnect eliminates the cost and complexity of external switches, while significantly improving scalability and reliability. -Scalable operating system (OS) - the OS is designed to run very large applications, scaling efficiently to handle the most complex compute tasks. A microkernel runs on the compute processors, providing a computational environment that minimizes system overhead, which is critical to a highly scalable system. Separately, the service processors run a full Linux operating system. -Open, scalable programming environment;the system facilitates development of scalable applications, allowing programmers to focus on the applications themselves and separately deal with global system services. It also offers multiple compilers and Cray Apprentice, the industry's most complete set of performance analysis tools that help identify and correct performance bottlenecks at scale. -Exceptional manageability and reliability;the Cray XT4 system maximizes reliability, availability and serviceability (RAS), with built-in redundancy for critical components and judicious use of socketed components, minimizing single points of failure. The Cray RAS and Management System (CRMS) functions independently with its own control processors and supervisory network so as not to diminish application performance. The system also provides a single interface to manage the entire system as one supercomputer, not as a cluster of independent systems. These characteristics contribute to lowering a customer's total cost of ownership. -Highly scalable I/O; the supercomputer features an input/output scheme that can accommodate the bandwidth needs of the most data-intensive applications. Users can select the appropriate number of RAIDs (redundant array of independent disks) and service elements to store all their data in a single, global file system. -Flexibility and expandability;the system infrastructure allows for easy upgrades to upcoming processor technologies from AMD, as well as future custom processing technologies from Cray such as multithreading, vector and reconfigurable co-processing being developed as part of Cray's Adaptive Supercomputing vision. This further preserves customers' investments in their Cray XT4 system and makes the system flexible enough to handle increasing demands for capability and capacity. The Cray XT4 supercomputer is available and shipping now. Go to www.cray.com/products/xt4 for more information. The new system will be unveiled at the Cray booth #1515, during the opening gala of SC06 in Tampa, Florida, later this evening.