Torque converter is widely used in all kinds of transmissions such as cars, buses, wind turbines and engineering machinery. Due to poor accuracy and adaptability, the traditional design method no longer suits the needs of torque converter design and optimization, and 3D flow field simulation is widely used in torque converter design. With the increasing interest in the coupling of computational fluid dynamics and modern optimization techniques, an urgent need in finding an appropriate algorithm in torque converter design has risen. To this end, this paper employed four popular algorithms to optimize the performance in terms of blade numbers on a response surface. We compared and contrasted the algorithms in terms of the ability of searching for global optimum and handling constraints. The results indicated that AMGA exhibited the best search ability which is mainly attributed to the use of a large external archive. NSGA II outperformed the other algorithms in handling constraints. Due to the use of neighborhood crossover, NCGA has difficulties converging to global optimum as well as preserving diversity. No optimizer is clearly best overall, but this study demonstrated the merits and drawbacks of each optimizer and this may help the designer in choosing an appropriate algorithm considering their actual needs.