Atmospheric nucleation is a process of phase transformation, which serves a significant role in many atmospheric and technological processes. To simulate atmospheric nucleation activities, certain molecular models with three-dimensional (3-D) structures are generated. Analyzing these 3-D molecular models can help promote understanding of nucleation processes. Unfortunately, the ability to understand atmospheric nucleation processes is greatly restricted due to lack of efficient visual data exploration tools. In this paper, we present a data visualization solution to visualize and classify 3-D molecular crystals. We developed a novel algorithm for calculating similarity between the 3-D molecular crystals, and further improved the overall system performance with GPU (graphics processing unit) acceleration.