INDUSTRY

The supercomputing system, a cluster of Unix servers connected with a high-speed network, will use more than 2,000 processors. About half are in a collection of 145 p655 servers being installed now that use IBM's Power4 processor, and the remainder are Power5 processors in servers scheduled to arrive by the end of the year, GM and IBM executives said today. The supercomputer performs at 9 teraflops and is equivalent to the fourth-fastest system on the most recent Top 500. The system, twice as fast as an earlier IBM supercomputer at GM, is "most likely the most powerful system of any industrial company," said Frank Roney, IBM's lead salesman for the GM account, in a news conference. It is expected to slash the amount of time it takes to get a vehicle to market from four years to 18 months. The GM supercomputer will be used to improve automotive safety research by simulating crashes with mathematical models rather than performing real-world crash tests, said Robert A. Kruse Jr., GM's executive director of vehicle integration. "At a cost of $500,000 per (crash test) vehicle, this has resulted in substantial savings," as well as the ability to test more designs than was previously possible, Kruse said. GM received the first phase of the supercomputer network in March and will receive a second phase later this year. Neither GM nor IBM would reveal how much the computer cost. "The automobile industry is increasingly reliant upon computers to develop and test vehicles. Our state-of-the-art computing platform allows us to shorten the time it takes to bring a vehicle to market from 48 to 18 months for some vehicles and significantly reduce structural costs in the process," said Jim Queen, vice president, GM North America Engineering. "Thanks to high performance computing GM can move faster, introduce more new vehicles and keep pace with changing consumer preferences." The following are examples of recent high performance computing work within GM: Innovative New Products -- Thanks to high-performance computing, GM can develop new, low-cost vehicle architectures faster than ever before. When the call came out to turn the Pontiac Solstice concept into a reality, GM created the new Kappa platform and the Pontiac Solstice in record time and at a low cost by using GM's computing network. With a heavy reliance on digital design and validation, the team was able to shorten the time it would take to develop and test this important new product. They were able to deliver a completely new architecture and product in 27 months and at a price point of $20,000. Global Collaboration and Product Development -- Creating products globally requires a high reliance on computing resources and virtual collaboration tools. The GTO team, working with the engineering and manufacturing staff at Holden Ltd. in Australia, benefited from GM's ability to share large data files between the 16 global engineering centers. This allowed the engineering team to conduct work in both Michigan and Australia seamlessly. Increased Cost Savings -- The increased use and reliance on digital testing has allowed GM to eliminate a large part of the cost of building and testing vehicles. By relying upon virtual crash tests, GM has been able to reduce the number of crash vehicles needed by more than 85%. At a cost of $500,000 per vehicle crash test, this adds up to significant savings. IBM won the GM contract based upon product performance, physical packaging and its commitment to deliver additional capacity in 2004 to meet GM's rising engineering and analytical demands. With this deal IBM becomes the major supplier of supercomputers to GM worldwide and establishes its technology as a global standard for high performance computing. IBM is amid a concerted attack on the high-performance technical computing market, which includes not just systems such as GM's but also a large number of much smaller systems.