Optimum design is a key approach to make full use of potential advantages of a parallel manipulator. The optimum design of multi-parameter parallel manipulators(more than three design parameters), such as Stewart manipulator, relies on analysis based and algorithm based optimum design methods, which fall to be accurate or intuitive. To solve this problem and achieve both accurate and intuition, atlas based optimum design of a general Stewart parallel manipulator is established, with rational selection of design parameters. Based on the defined spherical usable workspace(SUW), primary kinematic performance indices of the Stewart manipulator, involving workspace and condition number are introduced and analyzed. Then, corresponding performance atlases are drawn with the established non-dimensional design space, and impact of joint distribution angles on the manipulator performance is analyzed and illustrated. At last, an example on atlas based optimum design of the Stewart manipulator is accomplished to illustrate the optimum design process, considering the end-effector posture. Deduced atlases can be flexibly applied to both quantitative and qualitative analysis to get the desired optimal design for the Stewart manipulator with respect to related performance requirements. Besides, the established optimum design method can be further applied to other multi-parameter parallel manipulators. Optimum design is a key approach to make full use of potential advantages of a parallel manipulator. The optimum design of multi-parameter parallel manipulators (more than three design parameters), such as Stewart manipulator, relies on analysis based and algorithm based optimum design methods , which fall to be accurate or intuitive. Atlas based optimum design of a general Stewart parallel manipulator is established, with rational selection of design parameters. Based on the defined spherical usable workspace (SUW) , the corresponding kinematic performance indices of the Stewart manipulator, involving workspace and condition number are introduced and analyzed. Then, corresponding performance atlases are drawn with the established non-dimensional design space, and impact of joint distribution angles on the manipulator performance is analyzed and illustrated At last, an example on atlas based optimum design of the Stewart manipulator is获得 了 优选 设计 for the Stewart manipulator with respect to related performance requirements. Also, the established optimum design method can be further applied to other multi-parameter parallel manipulators.