Terrain modeling and rendering is essential in construction of realistic virtual environment, for interactive computer games application. Advanced high resolution scanning technologies such as LIDAR enable us to obtain dense and accurate irregular terrain dataset. However, unlike the regular sampled dataset such as DEM and DTED, irregular dataset cannot be rendered in real-time due to random distribution of samples. Moreover, because they generally have huge amount of point data, it is very hard to manipulate them in consumer PC. We propose a fast and efficient terrain rendering method using large-scale irregular dataset for computer games application. First, it reconstructs the geometry by converting irregular dataset into regular one, through resampling input data with regular interval considering the spatial distribution of sample points. It can generate animated scene by applying quad-tree based rendering method to converted dataset. Experimental results show that its image quality is not deteriorated in comparison to the result of rendering reconstructed geometric models.