Optimization of heat-exchanger manifolds can significantly improve the flow distribution inside their cores, improving the heat exchange and reducing flow obstruction. It also reduces the overall mass of the system and with it, the cost of additive manufacturing. However, during optimization, domains are typically modeled as 2D to minimize computing effort. Likewise, laminar flow is prescribed even when turbulence is expected in operation. The accuracy of such assumptions and their effect on optimized geometry is unclear. In this work, 2D topology optimization was first performed on an inlet manifold for both laminar and turbulent inlet boundary conditions. The resulting geometries were found to be starkly different, illustrating a difference in design concepts for different flow regimes. The laminar flow cases were then topology optimized with a 3D domain that modeled out-of-plane walls. This produced yet more different geometry, showing that these walls cannot be ignored. Experimental validation by testing stereolithography 3D prints proved that 3D optimization involves far more accurate flow modeling and the results are therefore likely to have better flow distribution.