Readme File for the dataset files support the article entitled "Thermally-Aware Composite Run-Time CPU Power Models" 
Accepted in Power And Timing Modeling, Optimization and Simulation (PATMOS, September 2016)

Run-Time CPU Power Modelling website:

http://www.powmon.ecs.soton.ac.uk/powermodeling/

This website makes available software tools for implementing our automated model building methodology which produces models that are both accurate and stable. We also provide power models for mobile CPUs (quad-core Cortex-A7 and quad-core Cortex-A15) which can be used directly in situations where an accurate reference model is required.

List of Figures:

Fig-2:
Residuals of model b plotted against temperature. Red line shows an equation predicting this trend.

Fig-3:
Residuals of model c (temperature-compensated) plotted against temperature. Red line shows an equation predicting this trend.

Fig-6:
The measured idle quad-core Cortex-A15 cluster power at various clock frequencies (fclk) when all four cores are online (n=4) and switching each off in turn until no core is online (n=0). The cluster voltage is also shown.

Fig-7,8,9,10:
Idle power consumption across different clock frequencies, showing the contibution from the static power and BG dynamic power. Fan switched on (Fig 7). Fan switched off (Fig.8).
Average error (mean absolute percentage error [MAPE]) (Fig 9) and Power consumption (Fig 10) across each DVFS level for all 39 workloads (Cortex-A15 CPU model)