Skip to main content
U.S. Department of Energy
Office of Scientific and Technical Information

Parallel designs are making inroads

Journal Article · · Electronics; (United States)
OSTI ID:5850869
There's a revolution brewing in supercomputers, pitting the classical von Neumann architecture against parallel configurations. Parallelism is making inroads in all sizes of computers, but the greater throughput it promises is especially critical in supercomputing. The main thrust in supercomputer architecture to date has been to build the fastest possible vector processor for single- or dual-processor supercomputers. The dual-processor models often team a superfast vector processor with a general-purpose scalar processor. Although many configurations may have only one vector processor, that processor is often highly pipelined for concurrent vector processing. But now there is a clear trend toward multiple-pipeline vector processors and even multiple vector processors in order to boost throughput even higher. For example, Cray is delivering four-processor machines-both X-MP/48 and Cray-2 models-and Alliant is delivering its FX/8 supercomputer with up to eight vector processors.
OSTI ID:
5850869
Journal Information:
Electronics; (United States), Journal Name: Electronics; (United States) Vol. 59:10; ISSN ELECA
Country of Publication:
United States
Language:
English