The purpose of this research is to know: (1) improvement of problem solving ability of mathematics between students taught by using realistic mathematics approach higher than mathematical problem solving ability of students taught by using ordinary learning (2) is there any significant interaction between early ability of mathematics Students and learning to improve students' mathematical problem solving skills. This research is a quasi experimental research. The population in this study were all students of Shafiyyyatul Amaliyah High School Medan, which amounted to 250 students, by taking two class samples, totaling 62 students through purposive sampling technique. The instruments used consisted of: mathematical problem-solving test with Trigonometry material. The instrument is said to have fulfilled the terms of content validity, as well as the realibility coefficients for pretest and posttest of 0.861 and 0.859 for mathematical problem solving abilities. The results of this study indicate that: (1) the calculation result using two path anava that is 12,366 whereas the value of sig 0,001 <0,05 means the improvement of problem solving ability of mathematics between student which is taught by using realistic mathematics approach higher than the ability of problem solving of student mathematic which is taught By using ordinary learning, (2) the result of two-track anova calculation is 0.275 with sig value. 0,761> 0,05 means there is no interaction between students 'early mathematical ability and learning to improve students' mathematical problem solving abilities. Based on research findings, realistic mathematical approach can be recommended to be one of the learning approaches used in the main school to achieve the competence of problem solving ability of mathematics.