As shown in Figure 14, the FF scheduling method shows worse performance than the GPU scheduling method, because the simulated applications except the MP3 decoding program require a lot of computation operations. In the FF method, the application is scheduled to the CPU when the CPU is not busy, resulting in worse performance than waiting for the GPU. The PHS scheduling method shows the best efficiency. It improves the performance by about 65% compared to the FF, since it considers the characteristics of the application based on the execution history. The PHS scheduling scheme also reduces the energy consumption by decreasing the execution time, as shown in Figure 15. Each scheduling method shows different temperature behavior, as shown in Figure 16. The usage of the GPU in the GPU and PHS scheduling methods goes higher than the other methods, resulting in the increased GPU temperature. In contrast, the CPU temperature is more increased in the CPU and FF methods.