Abstract
CMOS chips are engineered with sufficient performance margins to ensure that they meet the target performance under worst case operating conditions. Consequently, excess power is consumed for most cases when the operating conditions are more benign. This article will review a suite of dynamic power minimization techniques, which have been recently developed to reduce power consumption based on actual operating conditions. We will discuss commonly used techniques like Dynamic Power Switching (DPS), Dynamic Voltage and Frequency Scaling (DVS and DVFS) and Adaptive Voltage Scaling (AVS). Recent efforts to extend these to cover threshold voltage adaptation via Dynamic Voltage and Threshold Scaling (DVTS) will also be presented. Computation rate is also adapted to actual work load requirements via dynamically changing the hardware parallelism or by controlling the number of operations performed. These will be explained with some examples from the application domains of media and wireless signal processing.
Highlights
Power dissipation in a digital CMOS chip is given as:P = N a + N i I leak,iVDD (1)where Na and Ni are the total number of gates in active and asleep logic blocks respectively, a is the activity factor, C is the average capacitance of the net associated with the output of a gate, f is the operation frequency, and VDD is the supply voltage
Where Na and Ni are the total number of gates in active and asleep logic blocks respectively, a is the activity factor, C is the average capacitance of the net associated with the output of a gate, f is the operation frequency, and VDD is the supply voltage
For the active logic blocks, the first term within brackets is the dynamic power due to the capacitances charging and discharging and the second term is the leakage power of all the cells connected to the power supply
Summary
Power dissipation in a digital CMOS chip is given as:. where Na and Ni are the total number of gates in active and asleep logic blocks respectively, a is the activity factor, C is the average capacitance of the net associated with the output of a gate, f is the operation frequency, and VDD is the supply voltage. The physical design of the chip with actual gate choices, their placement and wiring layout, is done to meet the frequency f at worst case process and temperature conditions This determines the total number of gates (and memory cells), the switching capacitances (from gate sizes and wiring parasitics), the leakage currents and the supply voltage VDD in Equation 1. Dynamic power management techniques have emerged, where the power consumption is continuously adjusted during run time of the system [3,4] This is motivated by the fact that since the chip is designed to meet the most demanding application throughput requirements under worst case operating condition, it leads to an excess margin or wastage of power under typical operating conditions. We will discuss techniques which have been recently explored to adapt to varying computation needs via adapting the hardware parallelism or changing the number of operations performed dynamically
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.