Mobile computing devices such as cellular phones, tablets, and laptop computers have become permanent fixtures of modern-day life. Moreover, users have become accustomed to consistent advances in technology that continue to improve features and usability. For decades, such technological improvements could largely be attributed to Moore's Law, which predicts consistent doubling of transistor count and reductions in costs. For almost as long, people have repeated the now well-known saying, "no exponential lasts forever," and also predicted the eventual end of Moore's Law. It appears that the end might finally be near, which motivates research that identifies inefficiencies and finds opportunities at the intersection of traditionally disparate research areas to improve performance and energy efficiency of devices that span the computing spectrum from mobile to servers in the cloud. This project builds on existing technologies that control the voltage and frequency of microprocessors, but takes them to higher levels of integration and granularity.
There are several key challenges that stand to obstruct continued growth in computing. Limitations in power budgets, whether due to high cooling costs in high-performance computing systems or to fixed energy capacity in mobile devices, require innovations that will allow future systems with larger numbers of cores to dynamically adapt to time-varying needs of modern workloads. Integrated voltage regulators provide one of the most promising approaches to address energy and scalability constraints. Integrated voltage regulators offer advantages of nanosecond-scale voltage transition times, more efficient current delivery to the load at high power levels, and significant benefits in form-factor and system-level power management. This project aims to develop a systematic approach to answer questions surrounding the benefits and overheads associated with embedding integrated voltage regulators in future chips.
