A composite refinement approach for stationary iterative methods is introduced. Two new formulas (RJGS and RGSJ) are compared with the classical forms. Rates of convergence of the introduced composite formulas (RJGS and RGSJ) are well established. The efficient performance of the new forms is established theoretically and confirmed through numerical examples. The decrease in the required number of iterations for convergence is established through the calculation of the spectral radius of the iteration matrices. The algorithmic structure of the new formulas is announced. Three numerical examples with different convergent properties are considered. The calculations are performed with the help of computer algebra software Mathematica.