Parallel Computing with GPU: An accelerator for Data-Centric High Performance Computing

In recent years, the demand for high-performance computing in various disciplines of ICTs has affected the dependency on single core processes due to its decline in efficiency. Whereas, this requirement of advance in processing power has seen the emergence of Graphical Processing Units (GPU). In this paper, the role of GPUs in parallel computing along with its advantages, limitations and performance in various domains have been presented. The massive parallelism feature of GPUs can break down complex problems into smaller bits of computation with greater efficiency. The ability to compute complex distributed processes is ideal for implementing Machine Learning (ML) and Deep Learning (DL) algorithms. The evaluation of GPUs performance in medical physics, large scale data mining, CUDA workload analysis and applications of ML and DL has been deliberated on the high-performance computing. As GPUs continue to evolve, processing performance and efficiency should improve, and stand out as a cost-effective and powerful solution for optimizing faster and reliable large-scale data processing. Future research is expected to optimize GPU architectures for developing computing paradigms to keep them relevant in high-performance computing's fast-changing industry.