In Figure 8, GPU/GPU execution method shows the best performance, since the GPU provides superior throughput than the CPU. CPU/GPU execution method shows better performance than
GPU/CPU and CPU/CPU methods, since more computation intensive application is assigned to the GPU and the other application is assigned to the CPU in the CPU/GPU method. Figure 9 shows the energy consumption according to the execution methods. Figure 9 is similar to Figure 8, because the energy consumption is proportional to the execution time. When computation-intensive applications are executed by using the GPU instead of the CPU, the execution time and the energy
consumption are reduced dramatically. Figure 10 shows the temperature of each processing unit. As shown in the graph, the CPU shows the highest temperature when two computation intensive
applications are assigned to the CPU (CPU/CPU method). The GPU shows the highest temperature when two computation-intensive applications are executed on the GPU. The GPU in the CPU/GPU method shows higher temperature than that in the GPU/CPU method, because a more computation-intensive application between two applications are assigned to the GPU in the CPU/GPU method. Therefore, we can know that the GPU/GPU execution method shows the best performance and
energy efficiency whereas it increases the GPU temperature when two computation intensive applications are executed.