NVIDIA’s super-sampling technology powered by artificial intelligence would clearly have a bright future ahead of it.
Currently in version 3.5, inaugurated by update 2.0 of Cyberpunk 2077 And Phantom Liberty, DLSS should continue to benefit from developments in AI. As part of a discussion at the Digital Foundry table, one of the NVIDIA representatives indicated in any case that this process should reach its peak with DLSS 10, no less.
More AI at NVID-IA
NVIDIA does not hide its love of artificial intelligence, given the impressive growth that this booming technology has brought it. A great story that started very early with DLSS, taking advantage of AI (and more precisely a neural network) to improve in-game performance.
Today we are at DLSS 3.5, which improves the quality of ray tracing thanks to the Ray Reconstruction functionality, as well as performance thanks to optimization of image generation, introduced in DLSS 3. This unfortunately on condition to have an RTX 4000 graphics card.
Interviewed by Digital Foundry, Bryan Catanzaro, vice president of applied research on deep learning, indicated that NVIDIA’s proprietary technology will continue to evolve thanks to AI. In the still very distant future, the man is counting on the fact that version 10 of DLSS could take care of the rendering process entirely thanks to a neural system.
AI: the engine of tomorrow’s games?
“ I believe that, in the long term, the graphics industry will turn to generative AI for the rendering process. The reason is simple: we are able to learn much more complex functions by feeding a network with a mass of data than by manually building algorithms. », says Bryan Catanzaro.
According to him, the use of a neural network system interacting with a game engine is what will achieve new heights in visual quality. “ I think we will be able to achieve increased realism and also hopefully make it cheaper to create stunning AAA environments through neural network rendering “.
It goes without saying that such a prediction would suit NVIDIA, which has invested enormous resources in artificial intelligence. Coming back to DLSS, the technology should soon benefit from new features managed by a neural network such as radial caching and texture compression. But this will inevitably require feeding the huge appetite of AI, and therefore adding ever more Tensor Cores to its future graphics cards like the RTX 5000.
Source : Digital Foundry on YouTube
8