AuthorsK. R. Stokke, H. K. Stensland, C. Griwodz and P. Halvorsen
TitleEnergy Efficient Continuous Multimedia Processing Using the Tegra K1 Mobile SoC
AfilliationCommunication Systems, Communication Systems
Publication TypeProceedings, refereed
Year of Publication2015
Conference NameProceedings of the 7th ACM International Workshop on Mobile Video (MoVid)
Date Published03/2015
ISBN Number978-1-4503-3353-5

Energy consumption is an important issue for mobile devices, as the technological development in battery technology has not kept pace with the power requirements of mobile hardware. In this paper, we use a video rotation filter to study the efects of CPU and GPU frequency scaling in terms of performance and energy. Our platform is the Tegra K1 mobile processor with a quad-core CPU and a CUDA capable GPU. We find that most energy can be saved by minimising CPU frequency while meeting the filter’s framerate requirement. Interestingly, the frequency scaling affects GPUs differently, where the best frequency is always moderately higher than the minimum which meets the framerate requirement. Using these heuristics, it is possible to save up to 10 % energy compared to the standard Linux frequency scaling algorithms, which use processor utilisation to adjust processor frequency.

Citation Key23461

Contact person