Physically based rendering algorithms are typically easy to parallelize. The aim of my thesis is to study the common algorithms of this area, and analyze the usability of the GPU to accelerate them. For the efficient operation and usage of these algorithms some further tools are needed - like model handling and accelerating structures. These are somewhat distinguishable from the concrete implemented algorithms, therefore, can be applied to either of them. The thesis discusses the basics of ray tracing and a few algorithms used for the more accurate approximation of the shading equation. These are the Monte-Carlo integration, and some of its specific variants, that differ either in the quality of the generated image or the computational requirements.
Furthermore the usage of the GPU with these algorithms is discussed, and two possible applications are shown through the implementation of two programs. For this part the thesis builds on the NVIDIA OptiX Ray Tracing Engine, and discusses its two differently aimed libraries, the OptiX and the OptiX Prime.