An enhanced subspace iteration algorithm has been developed to solve eigenvalue problems reliably and efficiently. Basic subspace iteration algorithm has been improved by eliminating recalculation of converged eigenvectors, using Krylov sequence as initial vectors and incorporating with shifting techniques. The number of iterations and computational time have been considerably reduced when compared with the original one, and reliability for catching copies of the multiple roots has been retained successfully. Further research would be required for mathematical justification of the present method.