The kinetostatic analysis of manipulator performances at the end effector (EE) has traditionally been based on the computation of the so-called manipulability ellipsoids rather than on polytopes which are more accurate a representation. A reason why polytopes have received less attention is their being less tractable from a computational viewpoint. This paper presents an effective computation tool based on polytopes, with separation between translational and rotational analyses. This eliminates the ambiguity inherent in the conventional rototranslational formulation. Examples of application of this tool to the kinetostatic performance analysis of a redundant arm are finally provided, with numerical results.