Power Analysis on the Intel® Xeon Phi™ Coprocessor
Claude J. Wright Intel Corporation, USA
Abstract
Power has become the limiting factor for scaling many HPC clusters today. We can plug in as many compute nodes as are available, but we may be limited in how many we can supply with enough power. We are limited by the electrical power for which a data center has been outfitted. Some cluster installations today are running upwards of 20,000,000 W (20 MW) of power to run HPC problems at a very large scale. These clusters are all limited in scaling because of the power capacity available to a data center. Power is taking center stage as a key problem we must address in order to scale a cluster to new levels of high performance. There ...
Get High Performance Parallelism Pearls Volume One now with the O’Reilly learning platform.
O’Reilly members experience books, live events, courses curated by job role, and more from O’Reilly and nearly 200 top publishers.
