October 27, 2003 9:08 AM PST
Cray forecasts Red Storm for masses
Due out sometime in 2004, the supercomputers will use Cray's MPP (massively parallel processing) architecture, which it is using in the design of Red Storm, a 40-teraflop (40 trillion calculations per second) device ordered by the Energy Department's Sandia National Laboratories. Like Red Storm, the new product line will be powered by AMD's Opteron processors and run on the Linux operating system.
Red Storm is being built under the Energy Department's Accelerated Strategic Computing (ASC) initiative to create supercomputers that can, among other complex tasks, simulate nuclear explosions. It is believed that when Red Storm is finished next year, it could outpace the world's fastest supercomputer, Japan's NEC Earth Simulator.
Cray's announcement signifies increasing momentum in the supercomputer arena for AMD, which launched Opteron earlier this year in hopes of increasing its share of the corporate server market. Red Storm runs on 10,000 of the Model 246 processors. AMD has also landed Opteron-based supercomputer deals with the Energy Department's Los Alamos National Laboratory, Japan's National Institute of Advanced Industrial Science and Technology, and Dawning Information Industry of China.
Cray views Opteron as a perfect fit for both Red Storm and the company's upcoming product line because of the processor's ability to run both 32-bit and 64-bit software and its use of HyperTransport interface technology.
Get Up to Speed on...
Get the latest headlines and
company-specific news in our
expanded GUTS section.
Dave Rich, director of high-performance computing marketing at AMD, said the company was hoping that Cray might extend the technology Red Storm uses into its products. Rich said a number of customers already expressed interest in similar high-end Opteron-based machines.
The new Cray product line aims to be more efficient and cost-effective than so-called clustered systems that loosely link multiple servers or PCs that have connections with relatively low bandwidth. The company contends that its more traditional supercomputers are able to make more efficient use of their processors.
Clusters that have "commercial interconnects are fine for handling small problems or big problems that are simple in nature, but their efficiency can drop to less than 5 percent on really challenging problems and work loads, versus five to 10 times better than that for a system like Red Storm," Peter Ungaro, Cray's vice president of worldwide sales and marketing, said in a statement.
Information on configurations, pricing and other details will be disclosed when the company makes its formal product announcement.In related news, Virginia Polytechnic Institute and State University announced last week that it had built a 1,100 dual-processor Macintosh G5 PC cluster that will likely rank with the five fastest machines in the world.
In preliminary performance tests carried out on 2,112 of the system's 2,200 processors, the so-called Big Mac cluster achieved 8.1 teraflops, or trillions of operations per second. While the system is still being tuned, and final results won't be announced until next month, the performance figure would place the Big Mac at No. 4 on the list of the world's fastest 500 supercomputers.