A 3-dimensional (3-D) probabilistic model which has been developed previously for the prediction of grain structure formation during solidification is applied to thin superalloy plates produced using the investment-casting process. This model considers the random nucleation and orientation of nuclei formed at the mold surface and in the bulk of the liquid, the growth kinetics of the dendrite tips, and the preferential growth directions of the dendrite trunks and arms. In the present study, the grains are assumed to nucleate at the surface of the mold only. The computed grain structures, as observed in 2-dimensional (2-D) sections made parallel to the mold surface, are compared with experimental micrographs. The grain densities are then deduced as a function of the distance from the mold surface for both the experiment and the simulation. It is shown that these values are in good agreement, thus, providing validation of the grain formation mechanisms built into the 3-D probabilistic model. Finally, this model is further extended to more complex geometries and the 3-D computed grain structure of an equiaxed turbine-blade airfoil is compared with the experimental transverse section micrograph.