DLSS 3 personnel.5 with tracing paths are more real than native rendering with a rack, says NVIDIA

To celebrate the launch of NVIDIA DLSS 3.5 in Cyberpunk 2077: Phantom Liberty, Digital Foundry spent a video chain at a round table with Brian Katanzaro (vice president of deep-learning NVIDIA and Yakub Knapica (vice president and global art director of CD Project RED).

Discussing new technology, Brian Katanzaro from Nvidia said Dlss 3.5 Not only more beautiful than native rendering, but in a sense, its personnel are more realistic in combination with tracing of the path than native rendering with a traditional raster approach.

You know, actually I think DLSS 3.5 makes Cyberpunk 2077 even more beautiful than native rendering. This is my belief. The reason for this, again, is that AI is able to make more reasonable decisions on how to visualize the scene than what we knew how to do without AI. I think this will continue to develop.

Cyberpunk 2077 shots using DLSS (including personnel generation) are much “more real” than traditional graphic personnel. If you think about all the graphic tricks, all occlusions, shadows, artificial reflections, screen effects. The raster in general is just a bag of falsehood, right? So, we can abandon this and start tracing the path and really can get real shadows and real reflections.

The only way to do this is to synthesize many pixels with AI. It would be too time-consuming to perform the rendering of track tracer without any tricks. So, we change the methods that we use, and I think that ultimately we will get more real pixels with DLSS 3.5 than without him.

Yakub Knapik from the CDPR agreed with him, calling the rasterization of the “bunch of khakov” imposed on each other.

It’s strange to say this, but I agree with you. This is a very interesting point of view, and you ask: what compromise is here? On the one hand, you have an approach to rasterization, which is a set of hacks. You have many rendering layers that do not balanced among themselves, you simply apply them on each other to create personnel. Each separate layer is a hack, a approximation of reality: reflection in the screen space and all that, and you create such pixels, unlike a more accurate definition of reality by tracing the path, where you generate something using DLSS. It used to be said that using DLSS, you donated some quality for the sake of performance, but with DLSS 3.5 The image undoubtedly looks better than without it.

The discussion of “fake personnel” first arose in connection with the creation of DLSS personnel, since it generates one frame regardless of the conveyor of rendering and inserts the “imitated frame” after each “real frame”. Nevertheless, the increase in performance provided by generated personnel is undeniably large, especially in games with the loading of the processor, where an ordinary upciate can do very little. It is no coincidence that AMD follows this example with FSR 3, which will soon debut in Forspoken and Immortals of Aveum.

As for specifically DLSS 3.5, the new Ray Reconstruction function provides great advantages in quality when the scaling and trace of rays actively.