|June 29, 2015||Presentation: Towards energy efficient computing|
|Room: Drienerburght hotel, room E&F||Masoud Daneshtalab|
We are facing with a key technological problem in the many-core era, called the utilization wall. The utilization wall says that, with each process generation, the percentage of transistors that a chip design can switch at full frequency drops exponentially because of power constraints. This phenomenon is known as dark silicon — large portion of a chip’s silicon area that must remain mostly passive to stay within the chip’s power budget, while the power budget is becoming exponentially more valuable. In this talk, some of the possible solutions, emphasizing on approximation and neuromorphic, will be discussed in order to trade this cheap resource, dark silicon, for the more valuable resource, energy efficiency. Although silicon is getting darker, for researchers the future is bright and exciting; dark silicon will cause a transformation of the computational stack, and from that transformation will come many opportunities for investigation.