Font Size: a A A

Adaptive Voltage Management Enabling Energy Efficiency in Nanoscale Integrated Circuits

Posted on:2017-02-06Degree:Ph.DType:Dissertation
University:University of RochesterCandidate:Shapiro, Alexander EFull Text:PDF
GTID:1458390008995331Subject:Electrical engineering
Abstract/Summary:
Battery powered devices emphasize energy efficiency in modern sub-22 nm CMOS microprocessors rendering classic power reduction solutions not sufficient. Classical solutions that reduce power consumption in high performance integrated circuits are superseded with novel and enhanced power reduction techniques to enable the greater energy efficiency desired in modern microprocessors and emerging mobile platforms. Dynamic power consumption is reduced by operating over a wide range of supply voltages. This region of operation is enabled by a high speed and power efficient level shifter which translates low voltage digital signals to higher voltages (and vice versa), a key component that enables communication among circuits operating at different voltage levels. Additionally, optimizing the wide supply voltage range of signals propagating across long interconnect enables greater energy savings. A closed-form delay model supporting wide voltage range is developed to enable this capability. The model supports an ultra-wide voltage range from nominal voltages to subthreshold voltages, and a wide range of repeater sizes. To mitigate the drawback of lower operating speed at reduced supply voltages, the high performance exhibited by MOS current mode logic technology is exploited. High performance and energy efficient circuits are enabled by combining this logic style with power efficient near threshold circuits. Many-core systems that operate at high frequencies and process highly parallel workloads benefit from this combination of MCML with NTC.;Due to aggressive scaling, static power consumption can in some cases overshadow dynamic power. Techniques to lower leakage power have therefore become an important objective in modern microprocessors. To address this issue, an adaptive power gating technique is proposed. This technique utilizes high levels of granularity to save additional leakage power when a circuit is active as opposed to standard power gating that saves static power only when the entire circuit is powered off. This technique provides significant savings in static power in addition to standard benefits from classical power gating.;Improvements in energy efficiency are enabled by reducing both static and dynamic power consumption utilizing adaptive and near threshold circuit techniques. These advanced power reduction techniques will enable the greater energy efficiency required in modern portable systems.
Keywords/Search Tags:Energy efficiency, Power, Voltage, Circuit, Modern, Adaptive, Techniques
Related items