Comparative study of Batch and Instance rendering for static geometry in OpenGL

Abstract: Graphical rendering has use cases in many fields. When rendering in real time, such as in simulations or games, the time to compute each frame is of high importance. An important factor to reducing frame time is decreasing the amount of calls between the processor and graphics card. This is especially important when rendering a large amount of meshes, which is a collection of vertices representing some geometry. Two ways of achieving this are Batch and Instance rendering. This thesis aims to analyze and provide valuable insight into the factors that influence the decision between the techniques for static geometry, which is geometry that does not move. More specifically the thesis will investigate how vertex and mesh count affects the frame time. Both techniques were implemented using C++ and OpenGL and then compared against each other. The results showed that for low vertex amounts Batch rendering has lower frame time. Subsequently, at higher vertex amounts Instance rendering has lower frame time. The exact cut off depends on hardware, but in general the conclusion was that Batch rendering is better for simpler meshes while Instance rendering is better for more complex meshes.