This research aims to figure out ; 1) whether the ability of mathematical problem solving of students in the class with Problem Based Learning is better than the one with Direct Learning; 2) whether self-efficacy of students in the class with Problem Based Learning is better than the one with Direct Learning. This is a quantitative experimental research with Post-Test Only Control Design. The population of this research is all students of grade VIII at SMP Negeri 1 Patikraja. Two classes are chosen to be the samples of the research using cluster random sampling. Class VIII F is taught using Problem Based Learning and class VIII D is taught using Direct Learning. The data are analyzed using Independent Sample t-Test for post test and questionnaires with alpha α = 0,05. It is obtained significant score the ability of mathematical problem solving 0,031 and self-efficacy of 0,038. Both scores are less than 0,05. In conclusion, Problem Based Learning is effective towards the abliliy of mathematical problem solving and self-efficacy of the students.