APPLICATIONS

Silicon Graphics and Livermore Software Technology Corp. (LSTC) today announced that SGI Altix servers deliver record performance on LS-DYNA, the widely used impact simulation and analysis software from LSTC. In tests conducted by SGI and certified by LSTC, SGI Altix systems outperformed similarly configured high-performance computing (HPC) systems from IBM, Cray and HP on a standard three-car model crash simulation. SGI Altix turned in the fastest run times among all systems running from 16 to 96 processors. Incorporating automobile models developed by the National Crash Analysis Center, the three-car LS-DYNA model simulates a van crashing into the rear of a compact car, which in turn crashes into a mid-size car directly in front of it. This crash simulation is used by most vehicle manufacturers to determine safety issues such as rate a vehicle's safety or the risk of injury in the event of a crash. "Major steps forward in an HPC system's memory and file I/O capabilities, like those built into SGI Altix, make a real difference to engineers seeking to cut time-to-solution and enhance their ability to analyze and understand ever-growing simulation requirements," said Dr. John Hallquist, president, LSTC. "Today's achievement shows that our long history with SGI to leverage the greatest possible performance from both hardware and LSTC software continues to deliver benefits for LS-DYNA users." Based on the powerful and scalable Intel Itanium 2 processor family and featuring the industry-leading performance of the SGI NUMAflex architecture, the Altix 3000 server was able to outpace its closest competitor, the Cray XD1 supercomputer, on all tested configurations. "For years SGI and LSTC have worked closely together to deliver record-breaking performance which is invaluable to engineers and scientists the world over," said Jeff Greenwald, senior director of marketing and product management, Server and Platform Group, SGI. "Time and again we've heard how our customers have benefited from our combined investment in solutions to help them get higher quality products to market in record time." SGI Altix is particularly well suited to engineering and scientific applications, due in large part to SGI's high-throughput system architecture. SGI Altix features the industry's most scalable memory I/O, powered by a memory bus capable of moving information between CPUs and memory at 10.4GB/second. Via SGI NUMAlink, SGI Altix systems also offer data I/O bandwidth of more than 6.4GB/s per link, well beyond the typical Linux barrier of 500MB/s. Data crosses over an SGI NUMAlink switch, round-trip, in as little as 50 nanoseconds -- less time than it takes a beam of light to travel 50 feet -- compared to 10,000 nanoseconds or more with many commodity interconnects. Another key advantage for the Altix family is SGI's third-generation NUMAflex global shared-memory architecture. With NUMAflex, researchers can hold entire data sets in memory, allowing for faster and more interactive data analysis, and resulting in more incisive conclusions. As computer-aided engineering (CAE) applications grapple with larger and larger data sizes, SGI's unique NUMAflex architecture allows users to increase productivity by maximizing the application's use of available memory and minimizing latencies due to data transfer. In addition to running shared memory codes very well, Altix is versatile enough to run well as a clustered system. Earlier this year SGI unveiled the SGI Altix 1350 cluster, based on the acclaimed mid-range SGI Altix 350 server. Altix 1350 clusters were designed for large node capability plus cluster capacity-with nodes that can scale to 32 processors and 384GB of memory. Unlike conventional clusters, the SGI Altix 1350 cluster is easy to deploy and administer, with exceptional flexibility, scalability and usability at unparalleled price/performance over competing cluster solutions. Complete LS-DYNA three-car model results are available at: its Web site.