Pumping coolants through microchannels with well-defined structures along micro-electronic devices is a typical approach to remove the heat. It has been found recently that so-called nanofluids, i.e. dilute water-Cu or water-Al2O3 suspensions with a particle diameter of 100-150 nm, are highly efficient coolants. Numerical simulations can help to optimize the microchannel structures and typically homogenous single-phase models are applied. However, these underestimate the experimental results. An alternative approach is two-phase models based on an Eulerian approach for the base fluid and a Lagrangian description of the suspended particles. In this paper we follow that route and solve the three-dimensional governing equations including continuity, Navier-Stokes and energy equations with the well-known SIMPLE method. The governing equations for particles are solved by a 4th order Runge-Kutta algorithm. We focus on a wavy microchannel structure and demonstrate that the disagreement between the two simulation approaches is due to the non-homogeneous particle distribution in the domain. We also find that the Nusselt number increases with the increase in volume fraction and the decrease in particle diameter and that it. is about three times higher for a nanofluid in a wavy microchannels as compared to water in a straight microchannels.