Spatio Temporally-Filtered Adaptive Shading and a Generalized Shading Rate Stabilization Technique

Stefan Stappen

Supervisor(s): Johannes Unterguggenberger

TU Wien

Abstract: We propose a novel method to improve the efficiency and quality of real-time rendering applications, Spatio Temporally-Filtered Adaptive Shading (STeFAS). Utilizing Variable Rate Shading (VRS), a new hardware feature introduced with NVIDIA Turing graphics card micro-architecture released in 2018, and properties derived during Temporal Anti-Aliasing (TAA), our technique adapts the resolution to improve the performance and quality of real-time applications. VRS enables this kind of adaptation of the shading resolution throughout the screen during a single render pass. Our technique examines the color differences over time to find a fitting shading resolution. With each frame, this shading resolution is refined and adapted to the color differences accordingly. In contrast other techniques, STeFAS has very little overhead for computing the shading rate. This is one of the reasons why STeFAS enables up to 4x higher rendering resolutions while keeping up the same performance or 4x better performance at the same resolution. One problem with VRS is that varying shading rates may introduce flickering artefacts into real-time rendering applications. To alleviate these artefacts we focus on their root cause and propose a technique that stabilizes the shading rate. Our technique delivers more stable shading rates and reduces such artefacts significantly. In some cases, it is able to completely eliminate them.
Keywords: Graphics Hardware, Real-time Graphics
Full text:
Year: 2020