Power analysis for decoding of the digital audio encoding format MP3: Decoding the central processing unit and the graphics processing unit.

Abstract

Power consumption is one of the major design considerations for recent computing systems. MP3 players require MP3 decoding performed on the processor. We analyze the power consumption for MP3 decoding varying the processor type [central processing unit (CPU) and graphics processing unit (GPU)]. In general, the MP3 decoding is performed on CPU . GPU, which is a specialized processor that offloads three dimensional graphics rendering from the CPU, can be used for MP3 decoding since it is used for various computations as well as graphics rendering. At first, we analyze the power consumption for the MP3 decoding performed on CPU by using Inspector tool. And then, we analyze the power consumption for the MP3 decoding performed on GPU. To analyze the power efficiency of the GPU for MP3 decoding, we use CUDA for translating the MP3 decoding process from CPU to GPU. Computer unified device architecture (CUDA) is the computing engine in NVIDIA GPUs that is accessible to software developers through industry standard programming languages. According to our experimental results, the MP3 decoding performed on GPU saves power consumption by 23% compared to that on CPU. [Work supported by the MKE, Korea, under the ITRC supervised by the IITA (IITA-2009-(C1090-0903-0008).]