Computing Performance: Game Over or Next Level?

Samuel H. Fuller, Analog Devices Inc.
Lynette I. Millett, National Research Council
The end of dramatic exponential growth in single-processor performance marks the end of the dominance of the single microproessor in computing. The era of sequential computing must give way to an era in which parallelism holds the forefront. Although important scientific and engineering challenges lie ahead, this is an opportune time for innovation in programming systems and computing architectures.

Last year, the Computer Science and Telecommunications Board (CSTB) of the US National Academy of Sciences released The Future of Computing Performance: Game Over or Next Level?1 With sponsorship from the US National Science Foundation, the CSTB convened a committee of experts to identify key challenges to continued growth in computing performance and to outline a research agenda for meeting the emerging computing needs of the 21st century. These experts brought diverse perspectives in the fields of semiconductor technology, computer architecture, programming languages and methods, and applications to explore challenges to sustaining performance growth and meeting broad societal expectations for computing now and in the future.

The committee’s observations, findings, and recommendations can be broadly summarized in two categories: energy and power constraints on growth in computing performance, and a proposed research agenda that emphasizes new approaches to software and parallelism to meet future expectations for performance growth.

Read full text