It was seen that performing a DNS in complex geometries is asked for by the problem of thermal fatigue. Performing a DNS in complex geometries calls for the use of methods, the 2nd order accurate FVM, and cell types, extruded polyhydral cells and full polyhydral cells, which are normally not used for a DNS. It was shown that pipe and channel flows can be calculated within a few percent of the references when comparing the RMS of the fluctuations. Moving to complex geometries mesh strategies were described together with compromises in the solver settings. In conclusion it can be said that these methods and cell types can be used with confidence in predicting the channel and pipe flows. If more complex geometries, such as a T-junction, are used, results can still be considered highly accurate. To exactly quantify the accuracy for these geometries more analysis needs to be done.