To implement the Mechanistic-Empirical Pavement Design Guide (MEPDG) developed by AASHTO for pavement construction and rehabilitation, it is necessary to evaluate its performance prediction models using actual pavement performance data, material properties, traffic information, and environmental factors. This paper validated the MEPDG models using the performance of typical pavements in the state of Tennessee from pavement management system (PMS). With traffic and pavement structural information collected from PMS, the performance of selected highway pavement sections was analyzed with MEPDG Version 1.100 software. The predicted present serviceability index (PSI) and rutting were compared with the actual measurement values. An initial value for the international roughness index (IRI) was suggested for Tennessee highways through investigation into the actual roughness data. The dynamic moduli of asphalt mixtures for input level 1 were obtained from laboratory testing. Those for input level 3 were estimated with the Witczak model. The results show that rutting of asphalt concrete (AC) pavements was more accurately predicted at input level 1, whereas it was overpredicted at input level 3. Traffic level was found to be an important factor affecting predicted pavement roughness. It was also found that MEPDG software was relatively conservative for highway pavements with low traffic level. However, MEPDG with nationally averaged default parameters was not sensitive enough to differentiate various climates, traffic, and materials in Tennessee for the prediction of PSI.