<![CDATA[Modern computing demands have driven the use of processing devices capable of handling large workloads quickly and efficiently. The CPU plays an important role as the central controller for general-purpose instructions, whereas the GPU provides parallel-processing capabilities that are better suited to large-scale numerical operations. This study aims to evaluate performance differences between CPU and GPU architectures in Google Colab-based parallel computing. A quantitative experimental method was employed using a matrix multiplication test scenario implemented with the PyTorch library. Tests were performed on five matrix sizes: 500 x 500, 1000 x 1000, 2000 x 2000, 3000 x 3000, and 4000 x 4000. The measured parameters included CPU execution time, GPU execution time, and speedup. The results show that the GPU was not optimal for the small 500 x 500 matrix, with a speedup of 0.47 times. However, for larger matrices, the GPU delivered substantial performance gains, achieving speedups of 23.25 to 27.30 times over the CPU. These findings indicate that GPUs are more effective for large-scale parallel processing.]]>
Copyrights © 2026
