Speaker: Stanley Williams, Hewlett Packard Labs, USA
Abstract: With the end of Moore’s Law within sight (really!), what opportunities exist to continue the exponential scaling of computer performance and efficiency into the future? The primary consumers of energy and time in computers today are not the processors, but rather the communication and storage systems in a data center. Even if it were possible to perform computation infinitely fast with zero energy consumption, there would be little change in the power and time required to perform a calculation on the types of ‘big data’ sets that are beginning to dominate information technology. I will describe a path forward with two stages. The first is to change the fundamental structure of a computer from the processor-centric von Neumann paradigm that has dominated for the past seven decades to a memory-driven architecture based on a flat nonvolatile memory space, high bandwidth photonic interconnect and dispersed system-on-chip processors. This is the vision behind The Machine, a major research and development effort currently in progress in Hewlett Packard Enterprise. Once this transformation has been completed, a new era of hybrid computation can begin, which is the motivation behind a new Nanotechnology-Inspired Grand Challenge for Future Computing recently announced by the US White House. One of the drivers for this initiative is the thesis that although our present understanding of the brain is limited, we know enough now to design and build circuits that can accelerate certain computational tasks; and as we learn more about how brains communicate and process information, we will be able to harness that understanding to create a new exponential growth path for computing technology.
28 дек 2015
