Comparison of Small-Scale Parallelization Between GPU and CPU
There is a tendency to utilize the hybrid processor (CPU & GPU) for high performance computing (HPC) in the
professional field for the mass computing requirement. However, in the general public or small-scale computing field, there
are few studies that illustrate the potential of hybrid processor. CPU has been used in the main stream and smaller scale
computations for a long period of time. In order to support the necessity of the hybrid processor for the computing demands
of the future, this research seeks to investigate the application of GPU and CPU in small- scale computation scenarios.
Programming languages of C++ and CUDA are used to invoke CPU and GPU in the project. Small-scale of arrays are
applied in parallelization calculations separately by only GPU or CPU. Execution time for all the calculation has been
analyzed in order to obtain the performance and potential of using GPU and CPU.
This study shows that CPUs and GPUs have their own advantages and disadvantages. Since the requirements of user
experience and technology are increasing, conjunction of CPUs and GPUs has a promising potential in the application of
Keywords: GPU, CPU, hybrid, parallelization.