There is an ever growing industrial demand for quantitative assessment of fatigue endurance of critical structural details. Although FEA-based calculations have become a standard in engineering design, problems involving the Low-To-Medium cycle range (101-104) remain challenging. This paper presents an attempt to optimally choose material data, meshing density and other algorithm settings in the context of recent design of the large offshore windfarm installation vessel, VIDAR. In this study, an attempt is made to assess default FEA-based procedures in RADIOSS software by comparing an experimental test against numerical analyses. Standard slender cylindrical (“I”) samples as well as originally designed “Z”-shaped samples made of A90 (S690)-grade steel have been loaded at various nominal stress ranges with or without local yielding. A good correlation has been found between FEA results and experimental cycles-to-failure in I-shaped samples, provided the softwarematerial data generator is avoided and Smith-Watson-Topper mean stress correction is used. In the case of Z-shaped samples, the calculated cycles-to-initiation of macro-crack is significantly lower (factor of 3) from the experiment. The observed discrepancy is argued to be due to stress gradient influence.