Software optimization has become very important because of the use of software in the daily activities of society. The lack of maximizing software in the workplace could be caused by a few factors such as the use of inappropriate algorithms and poor program code writing. This report will discuss several optimizations that can be performed upon the Strassen matrix multiplication library using OpenCL. Different types of testing must be performed in order to improve library performance to create a matrix multiplication library that can perform multiplication on both small and large matrices efficiently. These tests utilize matrices of different sizes, and use various test parameters. Data analysis is accomplished by creating tables and graphs from the gathered data, and the conclusion is drawn based on the created tables and graphs. The testing results show that by applying appropriate test parameters then a significant increase in matrix multiplication speed can be achieved. As an example, the calculation of large matrices using optimized Strassen on GPU is 45% faster in term of matrix multiplication computation time, compared to Naive version. However, in the case of smaller size matrices, the Naive algorithm is known to be 17% faster than the optimized Strassen algorithm on GPU.