A Novel GPU Power Model for Accurate Smartphone Power Breakdown

Abstract

As GPU power consumption in smartphones increases with more advanced graphic performance, it becomes essential to estimate GPU power consumption accurately. The conventional GPU power model assumes, simply, that a GPU consumes constant power when turned on; however, this is no longer true for recent smartphone GPUs. In this paper, we propose an accurate GPU power model for smartphones, considering newly adopted dynamic voltage and frequency scaling. For the proposed GPU power model, our evaluation results show that the error rate for system power estimation is as low as 2.9%, on average, and 4.6% in the worst case.

Modular Modelling of an Embedded Mobile CPU-GPU Chip for Feature Estimation

“…Others track kernel queries to determine power consumption (Pathak et al, 2011). Other techniques, like the one presented in (Kim et al, 2012b;Kim and Chung, 2013;Kim et al, 2015) involve the construction of polynomial models to estimate the power consumption as a function of the frequency and of the load. In this case study, we use battery readings.…”

“…It was shown in the previous subsection that the voltage is itself a function of frequency. For the GPU, voltage measurements are not available in this case study, thus its frequency is used as input of the model (Kim et al, 2015). The last input is the Memory Occupation Rate (MOR)-the ratio of the occupied memory to the full memory-which will help include the memory power consumption in the model, since it is a part of the SoC.…”

“…For instance, Zhang et al propose a model called "PowerBooter" to the estimate power consumption of a smartphone through its battery sensors (Zhang et al, 2010). Another study used system calls to estimate the power consumptions (Pathak et al, 2011), while others used polynomial and regression models in a series of works to estimate the power consumption of smartphone components including the CPU and the GPU (Minyong Kim et al, 2012;Kim et al, 2012b;Kim and Chung, 2013;Kim et al, 2015). These studies offer simple modelling techniques and also show a clear dependence between frequency and power consumption.…”

A Novel Easy-to-construct Power Model for Embedded and Mobile Systems - Using Recursive Neural Nets to Estimate Power Consumption of ARM-based Embedded Systems and Mobile Devices

“…Regression-based models, as is the case for PowerBooter (Zhang et al, 2010), the models by Dong et al, (Dong and Zhong, 2011), Kim et al (Kim et al, 2012), Kim et al (Kim et al, 2015), Shukla et al (Shukla et al, 2016), and Xu et al (Xu et al, 2013) amongst others, are quite popular. However, these models assume-by definition-that the variations in power consumption are linear causing an increased estimation error when it is not the case (Hoque et al, 2015).…”

“…The authors later continued to enhance their model with better power estimation for the GPU (Kim and Chung, 2013), and the display (Kim et al, 2015). More recently, several new models were also published, either for the system as a whole, as it was the case for (Di Nucci et al, 2017a), (Shukla et al, 2016) and (Chowdhury and Hindle, 2016), or just the Central Processing Unit (CPU) like the models presented in (Yoon et al, 2017) and (Walker et al, 2017).…”