This paper provides a comprehensive survey of dynamic power management (DPM) techniques for system-level design. DPM is a methodology that dynamically reconfigures systems to minimize active components or load when they are idle or underutilized, thereby achieving energy efficiency. The authors describe how systems employ power-manageable components and the impact of dynamic reconfiguration on overall power consumption. They analyze implementation issues in electronic systems and recent initiatives to standardize hardware/software interfaces for software-controlled power management. The paper models power-managed systems as interacting components controlled by a power manager, focusing on the interaction between components and the environment. It discusses the characteristics of power-manageable components, such as multiple modes of operation and transition costs, and introduces the power state machine (PSM) model. The authors also explore the applicability of DPM, considering the break-even time for inactive states and the exploitability of these states based on workload statistics. The paper reviews predictive techniques, including fixed timeout policies and adaptive techniques, and stochastic control approaches. Predictive techniques exploit past workload data to predict future idle periods, while stochastic control addresses uncertainty in workload and system response. The authors discuss the limitations of static predictive techniques and the need for adaptation in nonstationary workloads. Overall, the paper aims to provide a detailed overview of DPM techniques, highlighting their benefits and challenges, and discussing the trade-offs between performance and power savings.This paper provides a comprehensive survey of dynamic power management (DPM) techniques for system-level design. DPM is a methodology that dynamically reconfigures systems to minimize active components or load when they are idle or underutilized, thereby achieving energy efficiency. The authors describe how systems employ power-manageable components and the impact of dynamic reconfiguration on overall power consumption. They analyze implementation issues in electronic systems and recent initiatives to standardize hardware/software interfaces for software-controlled power management. The paper models power-managed systems as interacting components controlled by a power manager, focusing on the interaction between components and the environment. It discusses the characteristics of power-manageable components, such as multiple modes of operation and transition costs, and introduces the power state machine (PSM) model. The authors also explore the applicability of DPM, considering the break-even time for inactive states and the exploitability of these states based on workload statistics. The paper reviews predictive techniques, including fixed timeout policies and adaptive techniques, and stochastic control approaches. Predictive techniques exploit past workload data to predict future idle periods, while stochastic control addresses uncertainty in workload and system response. The authors discuss the limitations of static predictive techniques and the need for adaptation in nonstationary workloads. Overall, the paper aims to provide a detailed overview of DPM techniques, highlighting their benefits and challenges, and discussing the trade-offs between performance and power savings.