The capabilities of optimization solvers Concorde and CPLEX from the NEOS server for solving traveling salesman problems of various dimensions using Euclidean and Manhattan norms for distance matrix calculations are investigated. The problem of finding all optimal traveling salesman routes is formulated based on classical Miller-Tucker-Zemlin constraints that ensure the connectivity of each found route, supplemented by additional constraints that exclude the re-discovery of already known tours. Computational experiments were performed for three test datasets containing graph descriptions with 38, 57, and 1354 vertices, constructed based on the geographical coordinates of settlements in the Republic of Djibouti (first test) and the Kyiv region of Ukraine (two other tests).