The two-dimensional natural convection and entropy generation within a hollow heat exchanger are investigated. The heat exchanger is filled with CuO-water nanofluid which its dynamic viscosity is estimated by KKL model. In addition, the influence of shapes of nanoparticles on the heat transfer rate is considered in the simulation, and the most efficient shape of nanoparticle is selected to be used in the further investigation. The entropy generation analysis and heatline visualization are employed to present a comprehensive study on the considered heat exchanger. The Rayleigh number in range of 10(3) to 10(6), nanoparticle concentrations in the pure water (0, 0.01, 0.02, 0.03 and 0.04 vol%) and four different thermal arrangements of internal active pipes are the governing parameters. (C) 2018 Elsevier B.V. All rights reserved.