Three iterative methods for calculation of n th roots (including one proposed by the author) are compared in two ways: (1) Theoretical convergence estimates are given. (2) A new macro-compiler which estimates machine running time is used to compare the running time of the three methods for a variety of input data.