This study presents the development of the stochastic finite element method (SFEM) to investigate the random free vibrations of beams with three-dimensional (3D) random field of material properties. Material properties, such as mass density and Young's modulus, are modeled as 3D stationary univariate random fields, with their correlations considered in the dynamic analysis. The random material fields are discretized into random variables using the weighted integral method combined with the perturbation technique to obtain a first-order approximation of the eigenvalues of free vibration. Statistical quantities, including mean, variance, and coefficient of variation (COV), of the eigenvalues are derived. Monte Carlo simulations (MCs), based on standard FEM and the spectral representation method for stochastic fields, are employed to validate the SFEM solution. The three-dimensional randomness of material properties significantly affects the random dynamic response of the structure. Results reveal that as the correlation distance increases, the dispersion of the eigenvalues around the expected value also increases. A perfect positive correlation between the 3D random fields of Young's modulus and mass density results in a smaller COV, whereas a perfect negative correlation leads to a larger COV. As the correlation distance approaches infinity, the COV approaches the total standard deviation for a negative correlation, while it becomes negligible for a positive correlation. For independent random fields, the COV converges to approximately 70 % of the total standard deviation. The nearly linear relationship between COV and standard deviation enables the prediction of the random response of the structure once the material property randomness is defined.