There NVIDIA GTC AI Conference was transmitted ion the occasion of the GTC 2026, where the CEO, Jensen Huang, officially presented the DLSS in its version 5. If until now the DLSS has distinguished between frames so to speak natively generated and those generated by artificial intelligence, the new iteration of the Nvidia technology seems to break down this distinction. DLSS 5 introduces the era of the so-called neural rendering, that is to say a sort of AI-generated overlay. In this scenario, the GPU no longer sequentially computes each individual element of the scene, but uses predictive models to reconstruct and predict all parts of the image. This hybrid approach allows for the management of visually complex environments and characters, drastically reducing computational load and allowing developers to push beyond the current "limits" of the Unreal Engine and leading graphics engines.
The stated goal of DLSS 5 is to achieve photorealism without the need to render every pixel from scratch, optimizing graphics card resources to achieve superior fluidity even in the most technically demanding titles. Although details on backward compatibility with current RTX cards are still lacking, the release window is already set for this fall. Furthermore, the video presented for this new version of DLSS, for now, was captured using two different methods. RTX 5090. Therefore, access to this new version, for now, does not bode well for consumers' pockets. To be continued.