Call of Duty Modern Warfare 4 graphics with DLSS 5 technology compared to 1440p movie

Enthusiasts demonstrated the capabilities of NVIDIA DLSS 5 neural network rendering technology using the example of the narrative campaign of a military shooter. Call of Duty Modern Warfare 4The key innovation of this generation of algorithms is the 3D guided neural rendering function, which is neural rendering guided by the 3D geometric data of the scene. Unlike previous iterations of super-resolution scaling or interpolation of intermediate frames, the new Frame Generation technology functions as a full-fledged generative coprocessor on top of the finished frame. Using spatial motion vectors, depth maps, and geometric information, the neural network was able to recreate a physically accurate environment, transforming the game’s introductory missions into an interactive cinematic experience without blurring or loss of detail.

Visual changes were more pronounced in complex scenes with a high density of artificial lighting, small objects, and dynamic particles. In the enclosed interiors of a grocery store, the system demonstrated accurate rendering of product packaging, the micro-relief of price tags, the shiny reflections on tile floors and the fabric texture of soldiers’ military equipment. Facial animation and character skin lost their usual polygonal sharpness and rasterization artifacts, gaining reference-quality light diffusion under the skin. In large-scale outdoor scenes with nighttime explosions, the generative coprocessor accurately conveyed the volumetric spread of thick smoke, beams of emergency lights and gradients of flames on wet asphalt, while maintaining the absolute stability of fine lines on wires and panels.

The integration of an advanced visual reconstruction method was carried out in conjunction with maximum graphics settings and hardware ray tracing. From the Call of Duty Modern Warfare 4 engine, by software blocking standard resolution scaling from simultaneously being enabled while ray tracing is active, neural network rendering overcame the limitations of traditional pipelines and delivered sharp 1440p images on a fast organic LED display. The algorithms eliminated flicker from subpixel geometry, ghosting behind fast-moving objects, and geometric noise, demonstrating the superiority of the generative neural network approach over conventional spatial antialiasing and traditional rasterization methods.

Source