Using physics simulations natural phenomena can be replicated with the computer. The aim is to calculate a physical feature as correclty as possible in order to draw conclusions for the real world. Fields of Application are, for example, medicine, industry, but also games or films. Snow is a very complex natural phenomenon due to its physical structure and properties. To simulate snow, different material properties have to be considered. The most important method that deals with the simulation of snow and its dynamics is the material point method. It combines the Lagrangian particles based on continuum mechanics with a Cartesian grid. The grid enables communication between the snow particles, which are not actually connected. For calculation of particles data is transferred from these particles to the grid nodes. There, calculations are carried out with information about neighboring particles. The results are then transferred back to the original particles. Using GPGPU techniques, physical simulations can be implemented on the graphics card. Procedures like the material point method can be parallelized well with these techniques. This paper deals with the physical basics of the material point method and implements them on the graphics card using compute shaders. Then performance and quality are evaluated.