The reliability of the AMBER force field is tested by comparing the total interaction energy and dispersion energy with the reference data obtained at the density functional theory-symmetry-adapted perturbation treatment (DFT-SAPT)/aug-cc-pVDZ level. The comparison is made for 194 different geometries of noncovalent complexes (H-bonded, stacked, mixed, and dispersion-bound), at the equilibrium distances as well as at longer distances (up to a relative distance of two). The total interaction energies agree very well with the reference data and only the strength of H-bonded complexes is slightly underestimated. In the case of dispersion energy, the overall agreement is even better, with the exception of the stacked aromatic systems, where the empirical dispersion energy is overestimated. The use of AMBER interaction energy and AMBER dispersion energy for different types of noncovalent complexes at equilibrium as well as at longer distances is thus justified, except for a few cases, such as the water molecule, where the dispersion energy is highly inaccurate.