Nvidia has been training DLSS with a supercomputer for 6 years

DLSS (Deep Learning Super Sampling) is considered by many to be one of Nvidia’s biggest assets as of late, as it can increase the frame rate and improve image quality in games. This requires long, continuous development carried out on a massive supercomputer.

In an interview with the website TechSpot Recently at CES 2025, Bryan Catanzaro, Vice President of Deep and Applied Learning Research, explained: DLSS training continues since its launch six years ago. All this happens at once The supercomputer was equipped with the company’s latest graphics cards in a seamless process.

Using machine learning and large neural networks, Deep Learning Super Sampling is trained based on errors. When images cannot be created, developers try to understand the reasons for the error and restart training with a new datasetDo this in cycles until you can deliver the source game content in the best quality possible.

DLSS 4 comes with changes

Although it may seem simple, working with a dataset of this scale is a huge challenge. DLSS doesn’t just work on one game or another and Available in more than 600 games with realistic textures and various special effectsIn addition to games that have not yet been released, Nvidia is already working on this process.

It is useful to remember this in recent days. Nvidia announces new DLSS 4 for GeForce RTX 50 This also brings improvements to the RTX 40 series. This technology can deliver higher frame rates with upgrades and improvements. Introduces Multi-Frame Rendering, which uses AI to create up to three artificial frames and catapult performances.

DLSS 4 will be available on January 30, the launch day of the RTX 5090 and RTX 5080, and will initially be seen in 75 games.