2002 · 59 citations · 0 references
EngineeringComputer EngineeringComputer ScienceComputer ArchitectureLow-power DesignMobile SystemsEnergy-efficient Mobile ComputingPower ManagementVoltage ScalingEmpirical EvaluationEnergy-efficient BenchmarkingHigh PerformanceHigh-performance Computing SystemsPower Management StrategiesPower Management SystemsPower Management TechniquesPower Measurement And Benchmarking
Energy efficiency is increasingly critical for mobile and server systems, yet current power‑management algorithms assume lower performance points are always more efficient, an assumption that may not hold. The authors introduce and extend the concept of critical power slope to model and predict the energy‑performance trade‑offs of systems with power‑management features. They evaluate three systems—a clock‑throttled Pentium laptop, a frequency‑scaled PowerPC platform, and a voltage‑scaled system—using empirical measurements and public data to demonstrate the benefits of the critical power slope and its extension to voltage scaling. Critical power slope shows that the Pentium laptop is most efficient at its highest frequency, the PowerPC platform at its lowest frequency, and that voltage scaling may not be beneficial below a certain point, as confirmed by a web‑serving benchmark.
Energy efficiency is becoming an increasingly important feature for both mobile and high-performance server systems. Most processors designed today include power management features that provide processor operating points which can be used in power management algorithms. However, existing power management algorithms implicitly assume that lower performance points are more energy efficient than higher performance points. Our empirical observations indicate that for many systems, this assumption is not valid.We introduce a new concept called critical power slope to explain and capture the power-performance characteristics of systems with power management features. We evaluate three systems - a clock throttled Pentium laptop, a frequency scaled PowerPC platform, and a voltage scaled system to demonstrate the benefits of our approach. Our evaluation is based on empirical measurements of the first two systems, and publicly available data for the third. Using critical power slope, we explain why on the Pentium-based system, it is energy efficient to run only at the highest frequency, while on the PowerPC-based system, it is energy efficient to run at the lowest frequency point. We confirm our results by measuring the behavior of a web serving benchmark. Furthermore, we extend the critical power slope concept to understand the benefits of voltage scaling when combined with frequency scaling. We show that in some cases, it may be energy efficient not to reduce voltage below a certain point.