Nvidia launches Deep Learning Super Sampling 2.0, an advanced platform for rendering AI-based graphics

In addition to the RTX Global Illumination SDK, Nvidia recently launched Deep Learning Super Sampling 2.0, AI technology that promises incredible graphics processing power for next generation GPUs. new, enhanced rendering capabilities based on AI.

At the heart of DLSS 2.0 is an artificial neural network that uses Nvidia RTX TensorCores to increase frame rates and produce sharp frames that can be asymptotic or even for better quality than graphics rendering. original.

Video introduction DLSS 2.0

DLSS 2.0 has trained over tens of thousands of high-resolution images, displayed offline in supercomputers at very low frame rates, with 64 patterns per pixel. With training weights for neural network aggregation, DLSS 2.0 takes low-resolution images as input and builds high-resolution images. Nvidia will then distribute this trained deep learning model to RTX-based PCs through NVIDIA drivers and OTA updates.

Nvidia launches Deep Learning Super Sampling 2.0, an advanced platform for rendering AI-based graphics Picture 1

By using the Turing TensorCores engine that provides specialized computing power up to 110 teraflops, DLSS 2.0 can run twice as fast as its predecessor. According to Nvidia's statement, the platform can handle an intensive 3D game along with a deep learning network simultaneously in real time. In addition to further enhancing performance, DLSS 2.0 will use transient feedback techniques to display only 1/4 to 1/2 of the number of pixels while still being able to provide image quality equivalent to original resolution.

Unlike the previous version, which requires neural networks to be trained separately for each new game, DLSS 2.0 is trained based on the general data warehouse, works on many different games, last together provide better game compatibility and support for more DLSS titles.

DLSS 2.0 has 3 picture quality modes for the internal rendering resolution of the game, including Quality, Balanced, and Performance. In particular, Performance is the most powerful option, allowing up to 4 times 'super resolution' (i.e. from 1080p → 4K).

Not stopping there, DLSS is now integrated into a custom branch of the Unreal Engine 4 code base. If you're interested in developing code with DLSS, you must first connect your Epic Games and GitHub accounts. and then access the development branch on GitHub. Details can be found on developer.nvidia.com/unrealengine.