This is how NVIDIA DLSS technology works

With each new generation of graphics cards, manufacturers introduce a number of technological functions. One of the most popular today is ray tracing, which is available on NVIDIA’s RTX series boards. Despite the nice graphics, the feature drops its performance and punishes the GPU with intensive work. Then another important brand new feature comes into play: the DLSS.

The abbreviation DLSS stands for Deep Learning Super Sampling. It’s another technology from NVIDIA specifically designed for heavier games. This time, besides improving the graphics quality, the main goal is to increase the performance of your GPU and increase the FPS rate in several situations.

Next, we’ll show you in detail how NVIDIA’s DLSS technology works, how it uses Artificial Intelligence to accelerate frame rendering on the GPU, and which games are already compatible with the new thing.

This is how NVIDIA DLSS technology works

Deep Learning Super Sampling Technology (DLSS) is an artificial intelligence, a program that is constantly trained to learn how to work properly. Its aim is to scale game images from a lower resolution to a much higher one, always paying attention to the least possible loss of quality and detail. It is applied in real time to every frame generated by the GPU while the game is running.

This process of upscaling an image from a lower resolution source is called upscaling. Although very common in the 4K era, the traditional methods have consequences such as loss of scene detail, blurry images, aliasing, and unwanted artifacts.

This process of upscaling an image from a lower resolution source is called upscaling.

In order for the DLSS-KI to learn to work around these problems and produce a perfect image, it is constantly fed thousands of game images so that it has the ability to detect every difference in pixels and detail between a scene in 720p, 1080p, 4k and. to compare 64k for example.

How DLSS technology came about

DLSS emerged from an NVIDIA internal survey, more precisely from 2016, when the team unveiled an algorithm that specializes in repairing and reconstructing images with damaged or intentionally removed parts. Marco Salvi says they immediately saw the potential for accelerating graphics performance by leveraging this AI.

What sets this technology apart from others is the fact that it is one of the first technologies to be able to process an image and achieve better scaling, free from any graphic artifacts such as dithering, noise and distortion.

This in turn was made possible by an extensive test and training program for the algorithm, which has already been fed with millions of images.

Why does it increase the performance of your graphics card?

In the words of Bryan Catanzaro, VP of the Applied Research Project in NVIDIA’s deep learning, games that use DLSS technology can render their frames almost twice as fast while maintaining the original image quality and detail of the scene.

To achieve this “feat” of increasing performance, NVIDIA uses the so-called tensor cores of graphics cards with Turing architecture to process these artificial intelligence and image scaling processes in real time. The idea of ​​running AI for every frame rendered may seem far-fetched, but we’ve finally got to a point where technology allows us to do that, says the VP.

Since the AI ​​is able to process low-resolution frames and upscale them without loss of fidelity, the DLSS function renders games internally with fewer pixels and uses the color sensor.

The DLSS function renders games internally with fewer pixels and uses the color sensor.

The result is a huge reduction in the load on the GPU and its memory, which now produce more frames per second while DLSS makes the final adjustments. The technology is especially important for those who play on high resolution monitors (like 4K) or models with high refresh rates (like 144 Hz and 240 Hz).

DLSS 2.0: an even better architecture

The research team didn’t stop there. NVIDIA recently announced DLSS 2.0, the advancement of technology and computational model. This new feature brought several improvements, such as a reduction in the training time required for the AI ​​and a significant increase in image quality.

For example, titles like Control had some distortion in elements like grids and reflections on the first iteration of DLSS. This was completely fixed by the 2.0 implementation. In some cases, the technology even goes so far as to render images of higher quality than the native resolution when enabled, even from a lower resolution source.

They even worked on aliasing to make the images cleaner. The algorithm has specialized enough to produce a 4K image from a source in 1080p, which is perfect for monitors and televisions that work at this resolution.

The DLSS 2.0 version is also processed faster by Tensor Cores, which means that the feature has been carried over to all graphics cards in the NVIDIA RTX series.

List of games already supported by NVIDIA DLSS technology

  • anthem
  • Battlefield V
  • Bright memory
  • control
  • Deliver us the moon
  • Final Fantasy XV
  • Mechwarrior V: mercenaries
  • Metro exodus
  • Monster Hunter: world
  • Shadow of the Tomb Raider
  • Wolfenstein Youngblood

Games that implement NVIDIA DLSS technology

  • In the midst of evil
  • ARK: Survival Evolved
  • Atomic heart
  • Cyberpunk 2077
  • Darksiders III
  • Fearless
  • Death stranding
  • Fear the wolves
  • Broken lands
  • Hellblade: Senua’s victim
  • Contract killer 2
  • Islands of Nyne: Battle Royale
  • Justice
  • JX3
  • kinetics
  • Outpost Zero
  • Overkills The Walking Dead
  • PlayerUnknown’s battlefields
  • Leftover from the ashes
  • scum
  • Serious Sam 4: Planet Badass
  • Shearwater
  • The blacksmith arena
  • Vampires: The Masquerade – Bloodlines 2
  • We happy couple