Marc Snir, director of the Mathematics and Computer Science Division at Argonne National Laboratory, gave a keynote address titled “The Future of Supercomputing,” at ICS 2014, the 28th International Conference on Supercomputing, held in Munich, Germany, June 10-13, 2014.
Noting that for the past two decades supercomputing has evolved in a relatively straightforward manner, Snir quickly focused on some of the challenges the area now faces. For example, the simple model of assembling supercomputers out of commodity microprocessors and leveraging their exponential increase in performance is clearly under stress. Moreover, the continued shrinking of device size requires increasingly expensive technologies that may not lead to improvements in the cost/performance ratio.
Far from being discouraged, Snir stated that continued progress is possible: more exotic, non-commodity device technology can provide two or more order of magnitude in performance, and better hardware and software architectures can significantly increase the efficiency of scientific computing platforms.
But real progress, Snir argued, will require a significant international research effort and major investments in future large-scale “computational instruments.”
ICS is the premier international forum for the presentation of research results in high-performance computing systems.