This study evaluates the effect of crossover operators in many-objective evolutionary algorithms (MOEAs). We consider NSGA-III , ε-MOEA, MOEA/D and IBEA as major MOEAs. In each MOEA, we apply SBX, DE, SPX, PCX and UNDX as crossover operators. Test problems are DTLZ2 and DTLZ3. In this study, Generational Distance (GD) metric is used to evaluate not only the convergence performance of finally obtained pareto optimal set but also the running convergence performance. When we change the number of objective functions for each test problem, we investigate the difference in the convergence performance among crossover operators in each MOEA. As a result, we can confirm that the effective crossover operator is changed by both the problem property and the number of objectives.