Deep Learning Super Sampling has become a popular feature of NVIDIA graphics cards, as it boosts the performance of the supported games. Now AMD is apparently working on an alternative.
If there are two areas where AMD graphics cards are still lagging behind NVIDIA competition, it’s ray tracing performance and DLSS. The AI-controlled deep learning super sampling of the NVIDIA cards ensures that games can be rendered in lower resolutions without any visible loss of quality. These are more or less fully compensated for by the AI image optimization. The result: higher frame rates with almost the same image quality.
AMD cannot keep up with this so far, but the manufacturer is apparently working on its own solution based on the already existing FidelityFX image optimization. The new feature is called FidelityFX Super Resolution and should, if everything goes well, be released this year. The development is described as one of the most extensive initiatives in internal development, not least to make the use of ray tracing less performance-intensive.
While NVIDIA relies on dedicated tensor cores for DLSS, which take on the AI tasks, AMD apparently wants to rely on a pure software solution, as the current Radeon cards do not have the appropriate requirements. “You don’t need machine learning to do it, you can do this many different ways and we are evaluating many different ways,” said Radeon GM Scott Herkelman. So it will be interesting to see what AMD’s solution will ultimately achieve and whether it can keep up with DLSS.