GPU-Memory Coordinated Energy Saving Approach Based on Extreme Learning Machine

GPU can offer more powerful computing capabilities compared to that of the CPU, but the problem of the energy consumption is particularly prominent and affects its application in a broader filed. In solving this problem, DVFS (Dynamic Voltage Frequency Scaling) becomes an effective solution. The previous works use the linear approaches and ignore other component characteristics in system at software runtime. This paper assumes that functional relation between the software runtime characteristics and the appropriate frequency which corresponds to the GPU and memory as nonlinear and proposes a GPU and memory coordinated energy saving approach (EDVFS) based on extreme learning machine. Experiments demonstrate the effectiveness of the approach and reasonableness of assumption and that EDVFS can get maximum energy savings of 10.63% and average energy savings of 2.68% compared to the traditional DVFS.