Machine learning has become increasingly popular as big data and data science applications have grown in demand over the previous decade. GPUs, on the other hand, are ideally suitable for Machine learning challenges due to their tremendous performance. NVIDIA has released the RAPIDS framework, which includes the cuML library, to take benefit from GPU performance for machine learning. A GPU-based framework such as RAPIDS provides fast results on large datasets, whereas a CPU-based framework takes more time. In comparison to the CPU, there is no library other than RAPIDS that provides fast computation speed on the GPU. The purpose of this paper is to compare RAPIDS and Sci-kit clustering, regression, and classification algorithms. As a result, K-means, DBSCAN, Linear regression, Lasso regression, Ridge regression, Random Forest, Nearest Neighbor, and KNearest Neighbor algorithms are used on GPUs and CPUs, and the results are determined by calculating the time taken for execution and speed up. The results indicate that the RAPIDS framework can completely harness the power of the GPU compared to the CPU-based Sci-kit library.