We report on the performances of two second-moment turbulence closures in predicting turbulence and laminar-to-turbulent transition in turbomachinery flows. The first model considered is the one by Hanjalic and Jakirlic (HJ) [Comput. Fluids, 27(2), pp. 137–156 (1998)], which follows the conventional approach with damping functions to account for the wall viscous and nonviscous effect. The second is an innovative topology-free elliptic blending model, EBM [R. Manceau and K. Hanjalic, Phys. Fluids, 14(3), pp. 1–11 (2002)], here presented in a revised formulation. An in-house finite element code based on a parallel technique is used for solving the equation set [Borello, Comput. Fluids, 32, pp. 1017–1047 (2003)]. The test cases under scrutiny are the transitional flow on a flat plate with circular leading edge (T3L ERCOFTAC-TSIG), and the flow around a double circular arc (DCA) compressor cascade in quasi-off-design condition [Zierke and Deutsch, NASA Contract Report 185118 (1989)]. The comparison between computations and experiments shows a satisfactory performance of the HJ model in predicting complex turbomachinery flows. The EBM also exhibits a fair level of accuracy, though it is less satisfactory in transition prediction. Nevertheless, in view of the robustness of the numerical formulation, the relative insensitivity to grid refinement, and the absence of topology-dependent parameters, the EBM is identified as an attractive second-moment closure option for computation of complex 3D turbulent flows in realistic turbomachinery configurations.