An AMD patent application, recently published as U.S. Pat. Pub. No. 20260115594, concerns an “AI rendering engine” that could be used to improve the performance of the rendering tasks that a traditional game engine would perform. This development comes on the heels of NVIDIA unveiling its DLSS5 AI filter which (among other features) is capable of re-rendering game assets. AMD’s publication shows that the heavy hitters in the semiconductor space show no sign of slowing down when it comes to developing Artificial Intelligence (“AI”) rendering techniques to build into their hardware.
AMD’s patent application suggests that their take on an AI rendering engine might go far beyond what NVIDIA recently demonstrated with the DLSS5. Through a process the application calls “neural extrapolation,” AMD’s system would be able to generate, upscale, and output detailed images from objects as crude as a stick figure. Graphics that would previously be rendered at low resolution would receive additional “content and frames” after being analyzed and altered by the AI rendering engine.
Illustrative Claim:
10. A system comprising: a memory configured to store state; and
a processor configured to perform operations comprising:
processing the state and user input in a game engine to output a first set of one or more frames; and
processing the one or more frames using extrapolation processing to generate a second set of one or more frames.
