The requirements for test results on vibration authenticity and reproducibility are discussed. Experimental results of longevity and vibration loading (fatigue curves and curves of the "vibration loading") of a 10 deterministic and random modes of vibration, done by a special technique with new sensors and the test equipment, were obtained. The deterministic and random modes equivalence coefficient much increases with the time before the destruction of the sample. It is shown that the "equivalent" replacement of a deterministic model of a random loading when tested on vibration loading is unacceptable. The model replacement when tested on durability is allowed only for polyharmonic mode but is incorrect.