This paper compares two approaches to implementing the Marching Cubes algorithm, a method of extracting a polygonal mesh from a 3D scalar field. One possible application of this algorithm is as a procedural terrain generation technique for use in video game development. The Marching Cubes algorithm is an easily parallelizable task, and as such benefits greatly from being executed on the GPU. The reason that the algorithm is so well suited for parallelization is that it breaks the problem of mesh generation into a large group of similar sub-problems that can be solved completely independently.