Five years have passed since Nvidia revealed DLSS technology supported by Tensor cores coinciding with the announcement of the RTX 20 cards, and since then the gaming experience with high frame rate and high quality has become possible with all RTX cards, and of course the green giant launches a newer version of the technology exclusively with every A new generation of RTX cards supporting the new generation of Tensor cores.
The DLSS 3.0 technology, exclusive to RTX 40 cards, was the latest frame rate booster in the industry, with its awesome ability to create frames from nothing, and here comes DLSS 3.5 to keep the advantages of the previous version and add to it, but this time it was not limited to improving the frame rate only, Rather, improving the quality of ray tracing is also known as “Ray Reconstruction”, so what magic does Nvidia hide for us with DLSS 3.5 technology?!
Ray Tracing is the cornerstone of DLSS 3.5:
We can’t talk about Ray Reconstruction without knowing how ray tracing works. First, the game engine creates the materials and geometry of the scene, which have physical properties that affect their appearance and the way they interact with light. Then the rays are captured from the camera’s perspective in order to determine the properties of the light sources and the way the light interacts when it hits the materials. If the light hits a mirror, parallel reflections are formed.
You may think that the matter is going smoothly, but there is a slight problem, which is that capturing the rays emanating from each individual pixel in the scene requires extraordinary hardware and a long time, even if this process occurs according to the rendering of the scenes through prior training that takes a few minutes or hours. So it is necessary to use ray sampling, that is, to shoot a bunch of rays at different points across the scene to represent the lighting, reflection, and shading of the entire scene.
In the end, the output of the previous process is scenes full of noise, gaps, and holes, but it is enough to determine what the scenes will look like after adding ray tracing. To complete the pixels that didn’t get their share of ray tracing, hand-tuned denoisers use two different methods to do this. Combine pixels from different frames, or snap them together contiguously to blend them into a scene, turning messy scene output into ray-tracing-enabled images, that simple!
There are challenges facing Ray Tracing without the latest DLSS technology:
The aforementioned “Denoisers” are hand-tuned and specially processed for each type of ray tracing within the scene, which increases the complexity and cost of the development process, and also reduces the frame rate of games that rely heavily on Big on ray tracing technology, multiple noise reduction tools work at the same time to push ray tracing image quality to its limits!
During the creation of the “Ray Tracing” process, the “Denoisers” either fetch pixels from previous frames to increase detail by stealing rays from previous scenes, which threatens the problem of “Ghosting” shadows, removing dynamic effects and reducing the quality of other effects, or Alignment of adjacent pixels, which threatens to mix graphic details and create contrasting lighting effects within the same scene.
After adding “Ray Tracing”, there is only one last stage left in the light ray tracing path, which is the “Upscaling” process, which is the difference element for the most graphically powerful and demanding gaming experience at a high frame rate. But with the removal of some effects and the deterioration of the quality of others, the limitations of “Denoiser” noise reduction tools are evident in removing the fine details or what is known as high-frequency information required to create clear and pure scenes. Fortunately, DLSS 3.5 has come to solve these problems.
DLSS 3.5 technology takes ray tracing to incredible levels for all RTX cards!
DLSS 3.5 avoids all the problems of denoisers thanks to its Ray Reconstruction process, which is a part of the AI Neural Renderers, by replacing the Hand-tuned Denoisers with the AI Network trained Nvidia supercomputers, for creating high quality pixels between beam samples, for all RTX cards including non-DLSS 3.0 cards!
DLSS 3.5 is trained on five times more data than DLSS 3.0, enabling it to identify various ray tracing effects to make smart decisions about using temporal and spatial data and keeping high-frequency data. To provide an advanced level of “Upscaling” quality operations. Pre-trained on scene rendering, which requires a lot of computing power compared to what is available during real-time rendering of game scenes, Ray Reconstruction identifies light patterns from training data and reconstructs them during gameplay, delivering highly detailed scenes. splendor.
When playing Portal without DLSS enabled, the Denoiser will go awry during pixel alignment as it cannot mix enough pixels, resulting in speckle-like effects. The situation is no better with pixel binning of previous frames where good pixels are not enough, infecting the light with effects that look like boiling fumes. But once DLSS 3.5 is enabled, it identifies patterns associated with reflections while keeping scenes stable, collecting precise pixels from previous frames and blending neighboring pixels to create realistic reflections.
We move on to the game Cyberpunk 2077, when the DLSS technology is not activated, the light is emitted from the headlights of the car inaccurately, as it surrounds it on both sides, as a result of the “Denoiser” tool bringing light effects from previous frames without caring about the unrealistic light. But when the DLSS 3.5 technology is activated, the light looks very realistic, as it comes from the front of the car and reflects off the buildings around it. The same goes for all of the stunning reflections within Night City, which are further enhanced by DLSS 3.5 technology, sharper and more realistic.
It is not possible to enjoy a smooth gaming experience for the game Cyberpunk 2077 in Overdrive mode, which offers “Ray Tracing” fully, except by activating the DLSS technologies supported by artificial intelligence, even if the computer is with an RTX 4090 card! While DLSS 2.0 processes images with lower quality and then upscales them with artificial intelligence, which greatly increases the frame rate, DLSS 3.0 supports frame generation, while DLSS 3.5 improves ray tracing effects by reconstructing ray Reconstruction”, and adding all these technologies increases the frame rate by five times at 4K settings compared to when DLSS is not enabled!
The impact of DLSS 3.5 technology is not limited to games only, but creative applications have a lot of luck with it, as these applications include a wide range of content that is difficult for “Denoisers” to deal with, since it requires “Hand-tuning” for each scene, which is What makes the quality of the content below par. Thanks to the neural network, DLSS 3.5 technology is able to identify a wide range of scenes to create high quality images. Among the creative applications expected to support DLSS 3.5 this year is D5 Render, and the impact of the technology is clear below.
RTX 40 owners will be able to enjoy Super Resolution, DLAA, Frame Generation and Ray Reconstruction for a high-performance, high-quality gaming experience. , which is what owners of RTX 30 and RTX 20 cards will get as well, except for frame configuration that requires fourth-generation Tensor cores. DLSS 3.5 will arrive this year for Cyberpunk 2077, Portal with RTX, Chaos Vantage, NVIDIA Omniverse apps, and D5 Render.
Nvidia announces DLSS 3.5 technology and supported games
Tags
DLSS 3.5 Nvidia Ray Tracing RTX 40 Slider
