A systematic comparison and benchmark for 9 simulation algorithms of cellular dynamic differentiation in single-cell transcriptomics was conducted, and reliable guideline and reference for developers and users were provided. Various metrics were used for the comprehensive evaluation of 9 algorithms in terms of accuracy, scalability and usability, and a model was established for predicting the time consuming and memory usage. Results showed that 9 evaluated algorithms are not capable of performing well both in data property and cellular differentiation trajectory simulation. Dyngen can simulate data which is more similar to the reference data in topology and cellular differentiation trajectory, but it consumed more time and used more memory. Almost half of the algorithms needed updating versions and maintaining relevant functions. When using simulation methods designed for cellular dynamic differentiation, users are supposed to take different applying situations and features of the tasks into consideration in order to select the most suitable simulation algorithm.