The long-awaited Metro: Exodus game is finally here, with Ray-tracing and Deep-Learning Super-Sampling (DLSS). Furthermore, Battlefield 5 has received an update that allows for DLSS. Nvidia’s RTX series is finally finding a playground to properly shine. The whole point of the Tensor cores in the RTX line of cards was to improve reflections and to make games more “lifelike”. To achieve this, the cards manipulate graphics settings like Ray-Tracing. However, soon after launch, we discovered that these settings affect performance significantly.
Nvidia understood this and was already working on a solution. That solution being DLSS. With DLSS, games could effectively “learn” from patterns and trends and store that data to present it for future use rather than using precious computing performance to re-create the image. A feature like DLSS enables the game to maintain higher framerates and to be more playable at higher resolutions. However, The DLSS that gamers have received so far has a slight issue on hand.
The Issue With DLSS At The Moment
The problem that gamers are facing with the feature is that it distorts or ruins the image quality when the feature is enabled. Whether that is an issue with the way the data is cached or something else is something for NVIDIA to find out and fix. At the moment, it is something players have to deal with to maintain a more fluid gaming experience. In essence then, it is the trade-off of performance for graphics.
The Technical Director of Deep Learning at NVIDIA, Andrew Edelstien, Posted this on NVIDIA’s website. The aim of the upload was to help users better understand why the issue was prevalent. He says that DLSS was meant to be used at or lower than 60 fps and at higher resolutions, otherwise, DLSS won’t prove to be much of a performance boost. Furthermore, as far as the image quality issues were concerned, he said that the issue will be fixed in due time.
A deep learning algorithm is one that requires a significant amount of base data to understand and recreate a situation for future use. It may need to analyze the same scene hundreds of times over before it is able to make a crystal clear replica of it. Perhaps that is what Andrew wished to forward with his upload. Of course, NVIDIA is working to improve the speed and analyzing capabilities of the algorithm. However, in the meantime , we can only give it more time and hope it improves with the passing of such.