At Computex last year, AMD introduced the world to its FidelityFX Super Resolution (FSR) image upscaling technology. FSR was a rival to Nvidia"s Deep Learning Super Sampling (DLSS) and could be run on any graphics card as it didn"t require dedicated Tensor cores for processing. That"s because AMD"s FSR would utilize the Shader cores themselves as it wasn"t any form of AI and machine learning implementation. So while the technology was more democratic, the upscaled image wasn"t always as good as Nvidia"s DLSS 2.0 and newer revisions. However, the next FSR version could change that.
According to a report from VideoCardz, AMD is going to introduce FSR 2.0 in three days time on March 17th and the site also shared some of the alleged details about it.
Apparently, FSR 2.0 will be employing temporal upscaling instead of spatial upscaling on FSR 1.0. If you are wondering what the big deal is, AMD is apparently promising image quality better than native using the temporal approach. Plus there is a new optimized anti-aliasing (AA) in place too.
.jpg){kind=link}
An alleged performance comparison figure is also out which shows FSR 2.0 in the Performance mode delivering near double the frames compared to native rendering.
.jpg){kind=link}
Of course, the per game implementation of FSR will still remain though, AMD is also purportedly releasing its Radeon Super Resolution (RSR) on the same day as FSR 2.0. The great thing about RSR is that it does not require per game implementation although it has its own minor caveat, but its only supported on RDNA-based cards, ie, Radeon RX 5000 and Radeon RX 6000 series GPUs. AMD will likely discuss details about FSR 2.0 and possibly even RSR at the upcoming GDC 2022 event.
While we are on the topic, Intel is also introducing its XeSS upscaling technology and despite being based on AI and ML, XeSS, just like FSR is also expected to run on all hardware.
Source and images: VideoCardz via CapFrameX (Twitter)