Abstract. Detached, double-lined spectroscopic binaries that are also eclipsing provide the most accurate determinations of stellar mass, radius, temperature and distance-independent luminosity for each of their individual components, and hence constitute a stringent test of single-star stellar evolution theory. We compile a large sample of 60 non-interacting, well-detached systems mostly with typical errors smaller than 2% for mass and radius and smaller than 5% for effective temperature, and compare them with the properties predicted by stellar evolutionary tracks from a minimization method. To assess the systematic errors introduced by a given set of tracks, we compare the results obtained using three widely-used independent sets of tracks, computed with different physical ingredients (the Geneva, Padova and Granada models). We also test the hypothesis that the components of these systems are coeval and have the same metallicity, and compare the derived ages and metallicities with the ones obtained by fitting a single isochrone to the system. Overall, there is a good agreement among the different determinations, and we provide a comprehensive discussion on the sub-sample of systems which either present problems or have estimated metallicities. Although within the errors the published tracks can fit most of the systems, a large degeneracy between age and metallicity remains. The power of the test is thus limited because the metallicities of most of the systems are unknown.