NVIDIA has further dived into the details of its artificially loaded DLSS 2.3 technology and unveiled a new image comparison tool known as ICAT. The new details include comparisons to the widely used “Spatial Upscaling” methods that have been in use for some time.
NVIDIA Talks DLSS 2.3 – Better quality than “Spatial Upscaling” methods and AMD’s FSR
The video, released by NVIDIA, features Bryan Catanzaro, director of Applied Deep Learning Research, explaining how DLSS 2.3 compares to other spatial scaling methods, including AMD’s FSR, a scaled, optimized, and refined method, but considering that NVIDIA, which has the advantage of AI Tensor Core technologies, DLSS is ultimately a better method.
We already know how NVIDIA DLSS 2.3 helps produce sharper images using motion vectors. It provides better image resolution and reduces ghosting / flicker. NVIDIA stressed that while DLSS 2.3 is better than ever, their own NVIDIA driver package also has its own spatial enhancement technology that produces better image quality than commonly used Enhancement Methods such as Lanczos and FSR. Note that AMD’s FSR is based on an updated and refined version of the Lanczos upscaler, as discussed earlier.
NVIDIA has also announced that they will make the new Image upscaling technology open source for integration into multiple game selections. We have more information about it here!
NVIDIA DLSS 2.3 / Upscaling / FSR Performance Comparison:
NVIDIA doesn’t end there, but compares DLSS performance and image quality to these Spatial Upscaling technologies. The video does not mention which particular spatial scaling method is used (Lanczos or AMD’s FSR), but it seems that DLSS in performance mode offers better performance at the expense of low IQ degradation, while quality mode offers higher IQ with similar or slightly faster performance.
Comparisons of performance and image quality include games such as Necromunda Hired Gun, Back 4 Blood, and Chernonylite (recently added to the DLSS series). You can watch the comparisons in the video below:
Both the NVIDIA DLSS and the AMD FSR are great in their relationship. Overall, DLSS is a superior and robust implementation based on artificial intelligence technology, while FSR strives for a more transparent approach and has a wider deployment rate while being a hardware agnostic. A third GPU player in the form of Intel’s XeSS is also coming to the arena soon. We’ve heard great things about the upcoming XeSS series with artificial intelligence in this interview, and Intel itself has shown some really impressive demos that we can’t expect to see in action when the first GPUs become available to the public by the first quarter of 2022. .