For many real-life systems ranging from financial to population-related to medical, dynamics is described by a system of linear equations. For such systems, the growth rate lambda can be determined as the largest eigenvalue of the corresponding matrix A. In many practical situations, we only know the components of the matrix A with interval (or fuzzy) uncertainty. In such situations, it is desirable to find the range of possible values of lambda. In this paper, we propose an efficient algorithm for computing lambda for a practically important case when all the components of the matrix A are non-negative.