Wednesday, December 29th 2021
AMD Rumored to Introduce Radeon Super Resolution (RSR) Upscale Tech in Early 2022
The image upscaling wars keep grassing, with AMD and NVIDIA claiming as many integrations as possible for their respective FSR (FidelityFX Super Resolution) and DLSS (Deep-Learning Super Sampling) technologies in a bid to achieve maximum market share for their respective technologies. While the entire world was now focusing on Intel's own addition to the image upscaling wars with its XeSS (XE SuperSampling) tech, AMD is apparently looking to introduce a new upscaling tech as early as January 2022. Enter Radeon Super Resolution (RSR).
Right off the bat, do not expect RSR to be AMD's answer to the perceived image quality advantage of NVIDIA's deep-learning-powered DLSS compared to AMD's more open (and cross-hardware compatible) FSR. Instead, AMD seems to be targeting RSR as a game-agnostic upscaling solution that's based on FSR, but which can be enabled at the Radeon driver level for any game that supports exclusive full-screen rendering. AMD is seemingly moving its image upscaling technique further up in the graphics pipeline, which should impact upscaling quality (as there's less information for the image upscaler to work with). What this does enable, however, is an agnostic solution that can be deployed in any game - provided you're rocking one of the two rumored architectures that will support RSR (RDNA and RDNA2, in the form of AMD's RX-5000 and RX-6000 series). Considering the expected release of RSR, it's likely that AMD will have an official announcement around CES 2022, despite the fact that the company won't be physically present due to COVID-19 and logistics concerns.
Source:
Videocardz
Right off the bat, do not expect RSR to be AMD's answer to the perceived image quality advantage of NVIDIA's deep-learning-powered DLSS compared to AMD's more open (and cross-hardware compatible) FSR. Instead, AMD seems to be targeting RSR as a game-agnostic upscaling solution that's based on FSR, but which can be enabled at the Radeon driver level for any game that supports exclusive full-screen rendering. AMD is seemingly moving its image upscaling technique further up in the graphics pipeline, which should impact upscaling quality (as there's less information for the image upscaler to work with). What this does enable, however, is an agnostic solution that can be deployed in any game - provided you're rocking one of the two rumored architectures that will support RSR (RDNA and RDNA2, in the form of AMD's RX-5000 and RX-6000 series). Considering the expected release of RSR, it's likely that AMD will have an official announcement around CES 2022, despite the fact that the company won't be physically present due to COVID-19 and logistics concerns.
65 Comments on AMD Rumored to Introduce Radeon Super Resolution (RSR) Upscale Tech in Early 2022
Basically making a chandeliers appear as some obscure statue, maybe this will improve the horror factor of RE?
Conclusion for FSR review in RE Village (DSOGaming):
AMD FSR improved performance by 20fps on our NVIDIA GeForce RTX 3080. Still, we don’t really recommend using it. It makes more sense to dial back some other settings than to use FSR (which brings a “vaseline” effect to the whole screen).
There is no free lunch and the artificial intelligence won't help..
Also Anti Aliasing is not the same thing as upscaling. FSR require a form of AA anyway and work with TAA.
My main issue with DLSS is it's Nvidia exclusive. I mean even if AMD implemented Tensor Core, they wouldn't be allowed to run DLSS. I think that is against the PC philosophy. I am not a big fan of any Lock-in like that.
But that is just a matter of time before an open alternative will be available for everyone. The technique have huge potential and we are still at the early stage in that technology.
But I strongly think those technique will be moved to the game engine itself instead of being something that come from a third party feature.
Instead, they transformed it into a Tensor Accelerated Temporal Upscaling with DLSS 2.0. We could debate that the use of tensor core mean AI but in the way it's being used with DLSS, it's not in the modern sense of AI with machine learning and stuff.
There are AI imagine upscaling around, it's just not fast enough to be able to be used in a real time rendering pipeline to improve framerate. The tradeoff they made with DLSS 1.0 was just not good enough. They kept the name for the version 2.0 but it's not really AI.
But it would be a very bad marketing move for Nvidia to admit it and rename it Nvidia temporal upscaling. AI is still the trend like Nanotechnology was the trend last decade...
If other GPU vendors don't have DLSS alternative, well they can lower their prices and not use rasterization as the sole performance indicator.
Others vendors will eventually have alternative. But then, we will have to wait few years until the Open one win the market like it always happen. After that, DLSS will just be another PhysX, Glide, Mantle.
Yeah let just stop playing games for a few years and hope for DLSS to die out, how is that a plan LMAO.
Not sure how "the Open one win the market like it always happen", how is Linux vs Window marketshare going? are you gaming on Linux? I'm definitely not.
Why should anyone care about propriety vs open source nonsense, just use what is best at the moment and upgrade when better stuff comes out. If DLSS is gone in the future, it would be because there are better Deep Learning upscaler around (definitely not some spatial upscaler :rolleyes: ), and not because of its propriety-ness.
I still do not see your point at all and you seem to make correlation that do not exist.
And i talk about open standard, not open source. You can install Windows on any hardware that support it.
Also there is no need for a better product, just one that is supported by more hardware. Intel one by example will run on any recent GPU including console if people want to port it there. Unless Nvidia specifically paid to have DLSS into their game, Game dev will have very few reason to not use the one that is supported on all platform.
Heh, one useless upscaler like FSR being support by every hardware is still pretty useless, especially when even bicubic upscaling is better :rolleyes: . Well let hope that Intel XeSS can provide some competition and not run like crap on every GPU beside Intel's.
As for FSR being worst than Bicubic, lol, I think you being a bit too much Nvidia fan boy there.I think you are missing the point, I do not mean that we no longer need any innovation for upscaler. I mean a similar level of performance and feature from an standard solutions would be enough to kill DLSS over time.
The technology will continue to evolve no matter what and i am not against that at all. On the contrary. But i am not really in for closed solution on PC no matter from who it come.
And yeah, I don't see DLSS being the premier reconstruction technique in 10 years time either, but I am glad that it exists in its current, past and perhaps future states to push this arm of technology forward.