Nvidia has officially announced Deep Learning Super Sampling 2.0 (DLSS), which is a massive advance for the now roughly four-year-old project from Nvidia promising upscaling that could look better than native resolution while delivering faster than native frame rates (via TechRadar).
Beyond the benefits for gamers using GeForce RTX 20-series GPUs, perhaps the more important piece of the puzzle is that the new implementation is reportedly considerably easier for developers to integrate into their games making it far more likely that we see the widespread adoption that DLSS 1.0 lacked.
Let's take a quick look at how DLSS 2.0 works and some of the results that NVIDIA is showing off so far.
- Time to upgrade? Plenty of RTX 20-Series among our best gaming laptops of 2020
- Nvidia RTX Super GPUs to deliver 50% faster performance in next-gen gaming laptops
- Xbox Series X, PC games to get graphics boost with new DX12 Ultimate
What is DLSS?
The basic premise of DLSS is that using deep learning and AI the Nvidia RTX technology can offer improved frame rates and visuals while also diminishing the workload for your system. This is achieved using the Tensor Cores in the RTX GPU to execute AI models at high speed and then working in tandem with the DLSS neural network the games are able to produce high-quality models from much lower-quality images.
One of the major differences for DLSS 2.0 is that it is no longer limited to per-game training, with the original implementation of DLSS Nvidia has to do extensive work with each game in order for the developer to make use of DLSS. With the revised AI model being used for DLSS 2.0 any developer should be able to add support into their game without needing to work directly with Nvidia.
While this naturally doesn't mean that every game will offer support for DLSS 2.0, it does remove a substantial pain point of the original implementation that explains why many of the titles that claimed DLSS support was coming never delivered on that promise.
The end results, as you can see in the samples from Wolfenstein: Youngblood and MechWarrior 5 below, are indistinguishable graphics at considerably higher frame rates.
Two games have actually already been making use of DLSS 2.0, Deliver Us The Moon and Wolfenstein: Youngblood, but it is now coming to MechWarrior 5 and Control. The hope is that developers get on board for this new version of DLSS much faster, who doesn't want to be able to run a 4K version of their game while only using the rendering resources for 1440p?
And again part of the benefit of DLSS is that it should keep improving with more data and thanks to some of the revisions this will also happen faster with DLSS 2.0 as it harvests data from games at nearly twice the speed of the original and all of the benefits are delivered to your GeForce RTX GPU via driver updates.
Users are also getting a greater degree of control with DLSS 2.0 with the option to choose between Quality, Balanced, or Performance modes, with Performance, for example, taking a game all the way from 1080p to 4K.
We'll be keeping our eye out for new games claiming support for DLSS 2.0 and watching for those MechWarrior 5 and Control updates to see the difference for ourselves.