In directionally solidified components the optimum grain structure consists of columnar grains aligned parallel to the solidification axis. However, in real castings deviations may occur in various forms like e.g. transverse grains, high angle grains or separately nucleated grains. For life assessment of cast turbine components it is of high interest to quantify the impact of these defects on fatigue strength. Parts revealing typical grain deviations were investigated using metallography and EBSD to gather information on grain structure and crystallographic orientation. It was found that grain boundaries can be highly inclined to the solidification axis and grains can reveal a large misorientation with respect to <001>. Therefore, an impact on fatigue life can be expected due to limited strength of grain boundaries and stress concentrations arising from elastic misfit. In order to quantify those effects, fatigue experiments were carried out on nickel-based alloy M-247LC. Cylindrically shaped specimens were cut from bi-crystal plates, being produced by growing two crystal seeds with different orientation in parallel. The bi-crystal specimens were considered as a model material representing typical features of grain structure deviations, such as crystallographic misfit and inclination of the grain boundary with respect to the major loading axis. The fatigue life measured in LCF tests was found to be lower compared to a boundary free and <001> orientated single crystalline specimen. Subsequent to testing, the specimens were investigated using SEM and EBSD to specify the crack initiation and propagation mechanism as well as the defect size causing the failure. It was found that most of the cracks initiated close to the grain boundary, propagating into one of the grains. There was no evidence of low strength of the grain boundary itself. In order to predict the local stress distribution in the bi-crystal specimen and thus the location of failure, the crystallographic orientation measured in the two grains was incorporated into a finite element model. A crystal plasticity material model was applied to account for local plastic deformation. Finally, a non-local damage approach was used to predict fatigue life. The results were found to be in good agreement with the experiments. Thus, it could be shown that local stress concentration induced by the grain boundary and elastic misfit was the main driver for crack initiation and fatigue life of the specimens.