Performance-gain Investigation of Foveated Rendering Technique
Modern virtual reality (VR) high-resolution displays render more than 4 million pixels, and each pixel needs to be rasterized, lit, shaded, and colored. High-resolution displays, along with an increased visual quality of VR content, put a high burden on the graphics equipment; thus, increasingly powerful graphics processing units (GPUs) need to cope with the demand. Foveated rendering (FR) is a technique with the potential to reduce the required processing performance for graphics-intensive VR applications significantly. The technique adapts the render resolution dynamically and only renders high-quality content at the user's eye focus point. Previous research already investigated different FR techniques. However, they focused on the technical implementation and produced a variety of results so that it remains inconclusive whether or not the user gains benefits in all situations.
This research aims to investigate the performance gain break-even for foveated rendering for different rendering algorithms. The research approach is experimental. Three different scenes with changing geometrical and algorithmic complexity were analyzed on six different GPU types using rotating FR settings. The performance of the FR renderer and a full-resolution renderer was profiled and compared to a theoretical model. The results indicate that the programmer gains the most from foveated rendering when the scene is complex (>160,000 vertices) with a high-resolution circular region around the gaze point covering 10% of the scene and employs multi-pass algorithms (e.g., shadows) with a minimum of 3 lights.