Last updated on December 8th, 2022 at 02:31 pm
AMD is preparing an open cross-platform alternative to NVIDIA DLSS
While NVIDIA focused on RTX (NVIDIA Ray Tracing) and DLSS (Deep Learning Super Sampling) technologies when launching GeForce Ampere, AMD focused more on raw performance, lower power requirements, and groundbreaking memory technology in RDNA 2-based graphics cards. However, this does not mean that AMD is not going to promote similar technologies.
During yesterday’s announcement, there were no details about AMD’s response to NVIDIA DLSS, a smart scaling technology that uses machine learning to dramatically improve graphics card performance by simply “recreating” a low-res AI image that has never been rendered by a graphics card. This sometimes results in an even more detailed and sharper picture than the native resolution.
At the AMD Radeon RX 6000 announcement, Scott Herkelman said, “We’re always looking to help developers with new initiatives and are already working on an ultra-high-resolution feature to empower gamers to improve performance when using ray tracing. ” It’s unclear if the upcoming ultra-high-resolution feature will only work with ray tracing enabled, or if it’s more like two separate technologies that can be enabled separately, as in the case of NVIDIA DLSS.
Tom Warren of The Verge was able to get more information on this ultra-high resolution technology:
“While AMD promises to clash with NVIDIA in 4K gaming and beyond, the big gap in the upcoming battle is the lack of a DLSS equivalent… NVIDIA AI-powered Super Scaling technology transforms the games that support it to deliver the superior image quality and higher frame rates simply by enabling this feature.
But AMD told me that they have a new super-scaling feature in their testing that is designed to improve performance during ray tracing. The company promises that its technology will open and cross-platform, meaning it will appear including the next generation of consoles – PlayStation to 5, the Xbox the Series the X, and the S. AMD is working with a number of partners on this technology and expects strong industry support. Unfortunately, the technology will not be ready when the three new Radeon RX 6000 series cards are launched. “
While AMD has clearly stated this feature as part of the FidelityFX technology, Tom Warren claims that it will be open and cross-platform. It is difficult to imagine NVIDIA supporting FidelityFX smart scaling, so it can be assumed that this technology will soon become part of Microsoft DirectML.
Xbox Series X and S accurately support ML acceleration. Tensor cores of the future console require a very small die area, but at the same time, they can provide a performance increase of 3-10 times, as reported in the Microsoft slide. DirectML Super Resolution is a Microsoft technology that was demonstrated back in 2019 at the Game Developers Conference. It can provide higher frame rates and lower latency.
