In engineering practice, a layered soil with sand overlying clay is often witnessed in naturally deposited or artificially backfilled grounds. However, limit insight has been gained into the failure mechanism and bearing capacity of two-layer soil. A numerical model on the basis of a finite difference method is developed for studying the influences of the footing roughness, dilation angle and surcharge on the failure mode and ultimate bearing capacity of the soil foundation. Based on the numerical results, the existing practical methods are evaluated, including the weighted average method, the load spreading method and the punching shear method. Our results indicate that the influence of the footing roughness on the bearing capacity decreases with the increase of the friction angle of soil. The dilation angle has remarkable effect on ultimate load when the dilation angle is small. Besides, the influence of surcharge becomes significant for low-strength clay. The weighted average method overestimates the bearing capacity, since it overestimates the weight of the strength of sand due to a conservative prediction of the depth of the failure slip. The load spreading method is considered to underestimate the bearing capacity when the depth of sand is large due to neglecting the shearing resistance of sand. The punching shear method, in which the vertical side sand block is utilized and shearing resistance of sand is taken into account, gives a reasonable solution for small sand depth and low-strength clay. In addition, the accuracy of the calculation can be improved if the local shear failure in the clay subgrade is taken into account.