NVIDIA DLSS 5 Uses 2D Data To Boost Image Quality: What We Know

NVIDIA's DLSS 5 has proven to be one of the more controversial announcements so far this year. When initially announced, NVIDIA CEO Jensen Huang said it was using new "neural rendering" techniques and "generative control at the geometry level". As we mentioned when covering Jensen's rejection of all criticism to DLSS 5, these claims seem to be in conflict with statements from some NVIDIA engineers who have been working on DLSS 5, with NVIDIA's own official slides confirming that DLSS 5 works from the same data as previous DLSS iterations, rather than being more deeply built into the engine.


Further statements from NVIDIA's Jacob Freeman made to YouTuber Daniel Owen also point toward this being the case. Freeman confirms that "Yes, DLSS 5 takes a 2D frame plus motion vectors as input", that all of its lighting calculations are made based on a single frame (i.e. screen space lighting/reflection calculations), and that "materials are inferred from the rendered frame". In response to criticism pointing out how DLSS 5 sometimes seems to add unintended facial features, like makeup, Freeman re-asserted that "The underlying geometry is unchanged" and that "this is a very early preview of the tech."

He also asserts that "Developers will have detailed controls such as intensity and color grading. Artists can use these controls to adjust blending, contrast, saturation and gamma, and determine where and how enhancements are applied to maintain the games' unique aesthetic. Developers can also mask specific objects or areas to be excluded from enhancement. We continue to talk to developers to understand all the ways they would like to control the technology. Ultimately, we see DLSS 5 as a tool for them to achieve their artistic vision, rather than be limited by the capabilities of traditional real-time rendering."


While criticism of DLSS 5 has been widespread, it is still important to remember that this technology is still in development and has several prominent game development studios are on-board. If unintended alterations to character designs can be mitigated by the time it is released, DLSS 5 could potentially be a good way to visually enhance games that don't use more advanced techniques, like Starfield and its fairly poorly-lit environments.

Fortunately, gamers are under no obligation to enable DLSS 5 in their games--if you don't like what DLSS 5 does to a particular game, don't enable it. Hopefully by the time DLSS 5 actually launches later this year, developers will have had enough time with the technology to ensure that key aspects of the designers' artistic vision are not compromised by enabling the technology and game developers view it as a tool to improve the visual fidelity of their games and not an AI crutch to alter lackluster initial artwork.