Deep-learning anti-aliasing, often known as DLAA, is effectively DLSS with no image upscaling. The technology maintains the rendering resolution at par with your display’s but employs AI to try to add detail to your game, especially smoothing out edges and other raster graphics forms that could need anti-aliasing. The end result is an image quality that is clearly better than even native resolution and other anti-aliasing techniques like TAA and MSAA. Since everything is taking place at native resolution, there is neither performance gain nor loss that can be compared to MSAA. It’s a fantastic technology, and any game that supports DLSS should automatically expose end users to DLAA in its highest quality settings, but that’s not the case. The uptake of DLAA by game developers has been slow, and NVIDIA has decided to address this.

Updates to DLSS are being developed by NVIDIA, thereby turning DLAA into a quality preset. According to a report from earlier this week, DLSS performance presets have an impact on Reflex and other algorithm aspects in addition to linearly scaling render resolution. In an Epic Unreal Engine 5.2 presentation, a special “G” setting was discovered. It turns out that this setting is a component of the changes NVIDIA is preparing for both DLSS 2 and DLSS 3, making DLAA a superior DLSS preset. When activated, it renders the game at its full display resolution without scaling down, but it also makes extensive use of DLSS’s AI picture reconstruction skills. It is hoped that by doing this, many more game creators will get interested in DLAA and include it into virtually every game that supports DLSS.