|AMD & Cray News - 02/14/05|
AMD & HP Special
AMD & IBM Special
AMD & Intel Special
AMD & Sun Special
|New Cray XD1 Supercomputer Buoyed by Top Benchmark Scores and Strong Sales Momentum; Demo at LinuxWorld Will Feature New AMD Opteron Dual-Core Processors
Global supercomputer leader Cray Inc (Nasdaq:CRAY) today reported that its Cray XD1 Opteron/Linux-based supercomputer has posted top scores in benchmark tests that include the government-sponsored High Performance Computing (HPC) Challenge, and has achieved worldwide order momentum in the HPC market since its introduction last October. The Company also announced it will demonstrate the Cray XD1 supercomputer equipped with AMD's new Opteron dual-core processors at the LinuxWorld Conference & Expo taking place this week in Boston.
"The Cray XD1 system has surpassed our expectations for benchmark tests and customer enthusiasm at this stage," said Peter Ungaro, Cray senior vice president for sales, marketing and service. "The system has demonstrated breakthrough performance, scalability and manageability compared to conventional cluster solutions. The Cray XD1 system is fulfilling its promise of giving engineers and researchers a high level of supercomputing power priced from under $100,000 US list--well below other supercomputers of comparable size and power."
Later this year, peak performance on the Cray XD1 system will nearly double with the availability of AMD's dual-core Opteron processors, which feature two cores directly connected on a single die. AMD states the chips will deliver the best performance per watt in the industry with little or no increase in power consumption or heat dissipation. The dual-core processors can easily be plugged into the Cray XD1 system as upgrades.
Top Scores on HPC Challenge and Other Tests
On the HPC Challenge benchmarks sponsored by the U.S. Department of Energy, the National Science Foundation and DARPA, the Cray XD1 system significantly outperformed all other HPC systems in its size and price category as of February 11, 2005. In the random-ring latency test, an important indicator of real-world performance, the Cray XD1 system surpassed all the other HPC systems tested regardless of size and price. This test assesses the latency and bandwidth of simultaneous communication patterns that impact small-message-sending codes, such as the Los Alamos National Laboratory's Parallel Ocean Program (POP), Livermore Software Technology Corporation's (LSTC) LS-Dyna engineering software and Acusim Software's AcuSolve(TM) computational fluid dynamics application.
The system performed best in its class and second overall on two tests: performing more than twice as fast as the leading Itanium system on the HPC Challenge random updates to memory test, which is important for adaptive mesh refinement codes such as SAGE, Partisan and CTH; and exceeding the performance of a comparably sized IBM p655 with Federation switch by more than 8 times on the global floating point execution test (FFTE benchmark), which is a useful measure of the total communications capacity of the network and affects codes such as molecular dynamic and climate prediction models.
In addition to the HPC Challenge tests, the Cray XD1 supercomputer has topped the competition in a number of other benchmarks. For example, the system proved 91 percent faster than a Xeon/Myrinet cluster, 31 percent faster than an Opteron/Infiniband cluster and 13 percent faster than on Itanium/Infiniband cluster at 32 processors when tested with Livermore Software's LS-DYNA--a program that is widely used in the automotive industry to simulate real-world automotive collisions.
Popularity in the HPC Marketplace
The Cray XD1 supercomputer has already been adopted by customers in more than 10 countries around the world. Organizations that have recently selected the Cray XD1 system include CINECA, the largest Italian computing consortium comprising 25 universities and the country's national research council, and the U.S. Air Force's Maui High Performance Computing Center, which is one of the top supercomputer sites in the United States Department of Defense.
"We look forward to giving Italy's research community access to the latest Cray technology and to expanding our long-standing relationship with Cray," said CINECA director Marco Lanzarini.
Some other Cray XD1 supercomputer customers include: the Central Institute for Applied Mathematics in Germany's Research Center Juelich (FZJ); the Alfred Wegener Institute in Bremerhaven, Germany; the SAHA Institute of Nuclear Physics in Calcutta, India; the Konrad Zuse Center for Information Technology in Berlin; the Ohio Supercomputer Center in Columbus; Pacific Northwest National Laboratory in Richland, Washington; and the Alabama Supercomputing Authority in Montgomery.
Supercomputing for Departmental Workloads
The Cray XD1 supercomputer is ideal for the highly specialized HPC applications used by government and academia, as well as commercial applications, such as computer-aided engineering (CAE) in aerospace, automotive, marine and other industries. Research areas that can reap benefits include climate and weather forecasting, aeronautics, astrophysics, adaptive optics, computational chemistry, quantum physics and biotechnology, as well as many others. With prices starting from under $100,000 US list the Cray XD1 system brings supercomputing to a broad range of HPC users, handling departmental to enterprise workloads.
A growing number of customers have selected the Cray XD1 supercomputer specifically for its application acceleration capabilities. In the Cray XD1 system architecture, optional field-programmable gate arrays (FPGAs) act as high-speed co-processors that can accelerate critical calculations as much as 30 to 100 times, resulting in significant decreases in the time required to run applications.
"At OSC we are really excited about the Cray XD1 supercomputer," said Kevin Wohlever, director, Springfield Operations, Ohio Supercomputing Center. "Bringing FPGAs into an HPC computing structure will allow us to support a new class of users. It will also reduce the number of platforms that we have to support, and bring together functionalities into a tightly coupled environment. This is extremely useful for our High Level Language and Blue Collar Computing efforts."
For details on the HPC Challenge benchmarks, go here (the Cray XD1 supercomputer is referred to as "Cray AMD Opteron"), or see the February 14, 2005 Cray press release titled "Cray XD1 Supercomputer Outscores Competition in HPC Challenge Benchmark Tests." For LS-Dyna results go to Top Crunch
About the Cray XD1 Supercomputer
The Cray XD1 supercomputer features direct connect processor (DCP) architecture, which removes PCI bottlenecks and memory contention to deliver superior sustained performance. According to the HPC Challenge benchmarks, the Cray XD1 system has the lowest latency of any comparable HPC system, with MPI latency of 1.8 microseconds and random ring latency of 1.3 microseconds.
Tests conducted by the Ohio Supercomputer Center show that the Cray XD1 system ships messages with four times lower MPI latency than common cluster interconnects such as Infiniband and Quadrics, and 30 times lower than Gigabit Ethernet employed in lowest-cost clusters. The Cray XD1 interconnect delivers twice the bandwidth of 4X Infiniband for messages up to 1 KB and 60 percent higher throughput for very large messages.
The Cray XD1 Opteron/Linux system runs x86 32/64 bit codes. Field-programmable gate arrays (FPGAs) are available to accelerate applications, and the Active Manager subsystem provides single-system command and control and high-availability features. The system can nearly double its performance with AMD Opteron dual-core processors when they become available later this year.
A 3VU (5.25") chassis provides 12 compute processors, 58 peak gigaflops, 96 GB/second aggregate switching capacity, 1.8-microsecond MPI interprocessor latency, 84 GB maximum memory and 1.5 TB maximum disk storage. A 12-chassis rack provides 144 compute processors, 691 peak gigaflops, 1TB/second aggregate switching capacity, 2 microsecond MPI interprocessor latency, 922 GB/second aggregate memory bandwidth, 1 TB maximum memory and 18 TB maximum disk storage.
About Cray Inc.
The world's leading supercomputer company, Cray Inc. pioneered high-performance computing with the introduction of the Cray-1 in 1976. The only company dedicated to meeting the specific needs of HPC users, Cray designs and manufactures supercomputers used by government, industry and academia worldwide for applications ranging from scientific research to product design, testing to manufacturing. Cray's diverse product portfolio delivers superior performance, scalability and reliability to the entire HPC market, from the high-end capability user to the department workgroup.
Safe Harbor Statement
This press release contains forward-looking statements. There are certain factors that could cause Cray's execution plans to differ materially from those anticipated by the statements above. These include the successful porting of application programs to Cray systems and general economic and market conditions. For a discussion of these and other risks, see "Factors That Could Affect Future Results" in Cray's most recent Quarterly Report on Form 10-Q filed with the SEC.
Cray is a registered trademark, and Cray XD1 is a trademark, of Cray Inc. All other trademarks are the property of their respective owners.
AMD & Cray Special
Cray XD1 Page
Cray XT3 Page
Sandia & Cray Red Storm
TOP 500 Supercomputers
|Your best source for AMD Athlon 64, barebones, DDR, nForce2, nForce3, nForce4, motherboards, notebooks, Opteron, Sempron & Turion 64 information|