NVIDIA has clarified a long-standing point of contention regarding the underlying functionality of its Deep Learning Super Sampling (DLSS) technology. The company confirmed that for DLSS to operate, its input is strictly limited to 2D frames and motion vectors. Crucially, this means the system does not read or process 3D geometry data from the scene.
This clarification addresses widespread debate and speculation among users and developers about how DLSS integrates with game engines and renders visuals, emphasizing its reliance on post-processing techniques rather than direct manipulation of the three-dimensional game world.