As network loading forms the basis of many dynamic traffic assignment solution algorithms, we consider in this paper four macroscopic link models that are widely used in DTA studies as the building blocks of certain network loading procedures. For the same generic link, the four models are solved and evaluated numerically for various link inflow profiles under a single-destination framework. And the characteristics of each model is analyzed and compared with others from multiple perspectives such as algorithmic implementation, model calibration, model accuracy, computational time and memory consumption. It is our purpose to find the best link model for the development of realistic and efficient network loading procedures. Moreover, since model discretization and calibration substantially influence a link model's performance, both issues are carefully addressed in the paper.