It is experimentally well-known that high anisotropies of the turbulent flow field are dominant inside the tip leakage vortex, which is attributable to a substantial proportion of the total loss and constitutes one of the dominant mechanisms of the noise generation. This anisotropic nature of turbulence invalidates the use of the conventional isotropic eddy viscosity turbulence models based on the Boussinesq assumption. In this study, to check whether an anisotropic turbulence model is superior to the isotropic ones or not, the results obtained from the steady-state Reynolds averaged Navier-Stokes simulations based on the RNG k-$varepsilon$ model and the Reynolds stress model (RSM) are compared with experimental data for two test cases: a linear compressor cascade and a forward-swept axial-flow fan. Through this comparative study of turbulence models, it is clearly shown that the RSM, which can express the production term and body-force term induced by system rotation without introducing any modeling, should be used to predict quantitatively the complex tip leakage flow, especially in the rotating environment.