Ray tracing is a well known algorithm for the generation of realistic synthesis images. It is also applied in the acoustic domain. The computational cost of the full 3D algorithm was previously an obstacle leading to approximations like the 2D 1/2 ray tracing. Despite the fact that this method significantly reduces the needed computations, it also limits the realism of the approach. This article proposes a study of the 3D ray tracing algorithm in the acoustic context. It presents several methods which can be used to reduce the computation time using different acceleration structures. Design decisions and optimisations made on the general ray tracing engine are explained. Then the choice of the acceleration structure and the propagation methods to use is clarified. The implantation was tested in the TYMPAN environment provided by EDF R&D using the NMPB08 method. To conclude, different results and takeaways are discussed.