Like it or not, but technology upscaling modern titles are becoming more and more dependent. The most popular feature that allows you to “squeeze out” additional FPS (in exchange for more input lag) is NVIDIA DLSS. This is only available on Nvidia RTX 20 and newer graphics cards and uses native Tensor cores specifically designed to work with artificial intelligence and machine learning.
DLSS is currently available in two main versions – 2.0 and 3.0 – with the newer one able to offer even higher frame rate (FPS) gains, which was excellently demonstrated, for example, in Cyberpunk 2077. However, it is only available for game models created on Ada Lovelace architecture, i.e. RTX 40. According to Videocardz, Nvidia intends to take the version introduced last September to an even higher level and should focus on improving ray tracing first.
Videocardz represents a technology already announced by Nvidia, but probably the only source so far to release information directly from the green team. At the time of this writing, no press release has been issued that DLSS 3.5 presented. However, we have a number of details needed to discuss the new scaling feature.
DLSS 3.5 will be released in the fall and will add a new feature called Ray Reconstruction.. A model that promises a much better visual experience when ray tracing is enabled has allegedly been trained on quintuple data was used to create DLSS 3.0. According to Videocardz, the technology should work better than standard denoisers, which often unsuccessfully try to rid ray tracing of various lighting inaccuracies (such as noise). In addition, denoisers worsen global illumination and reflections, which Ray Reconstruction also has to deal with.
Source :Indian TV
