Fatigue cracking is one of the main distress modes of highway asphalt pavements. In laboratory fatigue tests, haversine or sinusoidal loading waves are commonly applied to characterize the fatigue performance of asphalt mixture samples. The strain wave in field asphalt layer under a moving axle load, however, is noticeably different from the haversine or sinusoidal wave. This research compared the fatigue behaviours of compacted asphalt mixture samples under actual strain waves and the commonly used haversine wave. Two types of actual strain waves induced by the single axle and tandem axle of vehicles were included for analysis. It is found that the fatigue lives of asphalt mixture samples subjected to the single-axle wave are obviously higher than those subjected to haversine wave, while the tandem–axle wave leads to the lowest fatigue life. The changing rates of stiffness, phase angle, and dissipated energy with loading cycles for the mixture samples under single-axle wave are all the lowest, while those under tandem-axle wave are all the highest. This helps explain the discrepancies in the samples’ fatigue lives under different loading waves. The above findings suggest that it is necessary to test asphalt mixture samples using loading waves that simulate the actual ones in the field (i.e., single-axle wave and tandem-axle wave) to improve the reliability of fatigue life prediction of asphalt pavements. In addition, this research reveals that three dissipated-energy-based indicators, namely initial dissipated energy, cumulative dissipated energy and plateau value, all generate a unique relationship with the sample’s fatigue life. The relationship is independent of loading wave type and strain level.