Officials at the University of Utah Thursday said its new supercomputer cluster due next month will be based on AMD Opteron processors.
Code-named “Arches,” the system will be Utah’s largest scientific computer for biomedical research and will be the combined product of five clusters comprised of varying numbers of nodes and processors. The current 500-node cluster is being built by Angstrom Microsystems using its Titan64 Superblades and 1,000 Opteron chips.
When completed, researchers say they will try to identify the causes of inherited cancers and other diseases attributed to multiple genes. Researchers are also expected to use Arches for looking at how the body absorbs drugs in an effort to improve treatment of various ailments.
“We chose the AMD Opteron processor because we recognize its capabilities in enhancing the performance of our scientific computing applications,” Julio Facelli, director of the Center for High Performance Computing at the University of Utah said in a statement. “The new system from Angstrom will help us to deeper analyze our extensive amount of biomedical data as well as perform even more complex and advanced simulations.”
To develop the Arches cluster, the Center for High Performance Computing at the University of Utah received a grant from the National Institute of Health’s National Center for Research Resources.
This is the second major installation of Opteron processors coming next month. Last week, Sunnyvale, Calif.-based AMD said the U.S. Department of Energy is building two separate large-scale Linux clusters using more than 3,300 AMD Opteron processors dubbed “Lightning” and “Orange.” The clusters will be installed at Los Alamos National Laboratory in October to support the DOE’s computer projects in medical, environmental and national defense modeling and simulation.
Clustering is quickly becoming a popular configuration in the supercomputing marketplace. According to the latest list of the 500 most powerful supercomputers, a LinuxNetworx cluster running Intel Xeon chips at the Livermore National Laboratory is number three on the list. IBM and Hewlett-Packard also have clusters in the Top Ten.
But analysts say 2004 could be the final make-or-break year for AMD based on how well Opteron does. A report issued Wednesday by In-Stat/MDR says AMD needs to make its 64-bit family a success in order to improve its overall ASP (and top line revenue) and reduce manufacturing cost with 90nm production to the point where it can be profitable.
“AMD’s Opteron will offer key technical advantages such as 64-bit capabilities, an on-chip memory controller, high performance, and HyperTransport links for glueless multiprocessing with up to eight processors,” the analyst firm said. “Despite these advantages, we see an up-hill battle for AMD to gain market share in servers and capture an extensive (volume) commitment from top tier OEM customers.”
Another critical area In-Stat identified with AMD is that the company is moving away from being a vertically integrated company; a transformation may take it so far as to become a fabless company in 2007. The thinking by analysts is that AMD would rely on IBM Microelectronics for its 65nm semiconductor process manufacturing.
“Considering the lead time to build a new fab and bring it into production (AMD’s Fab30 took over three years from announcement to yielding first silicon), AMD is already too late building a new fab to produce 65nm product in 2005,” the analyst firm said.
