It's the same posters doing the same ole **** and it goes unchecked.
The tipping point for a mod to get involved seems to be calling someone a fanboy.
-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
Fidelity Super Resolution in 2021
- Thread starter Grim5
- Start date
It's the same posters doing the same ole **** and it goes unchecked.
The tipping point for a mod to get involved seems to be calling someone a fanboy.
really don't understand why people are so touchy about this, if its good, its good for everyone, if its poop we can all just happily ignore it like many other "features" that both manufacturers have tried to sell us in the past.
I hope its good, cos that's what's best for us. And for me it fixes the biggest problem with DLSS, which is that DLSS takes a new high end graphics card and allows people to get a little more performance for a small detail hit, when what people really need is to take older low and mid range cards and give them a little more performance for a small detail hit.
The vast majority of people do not benefit from going from 60 to 100 FPS at 4k, what they want is to be able to go from 40 to 60 fps at 1080p
I hope its good, cos that's what's best for us. And for me it fixes the biggest problem with DLSS, which is that DLSS takes a new high end graphics card and allows people to get a little more performance for a small detail hit, when what people really need is to take older low and mid range cards and give them a little more performance for a small detail hit.
The vast majority of people do not benefit from going from 60 to 100 FPS at 4k, what they want is to be able to go from 40 to 60 fps at 1080p
As much as i have no interest in DLSS/FSR. I can get behind this line of thinking. TBF, AMD did demo on the 1060 (?) and they have support for older graphics cards. So they did demonstrate this line of thinking so kudos to them.
Caporegime
During Computex Keynote, Scott Herkelman revealed Radeon’s most anticipated technology FidelityFX Super Resolution (FSR) will be supported on Radeon RX 6000, RX 5000, RX 500, and RX Vega graphics. It will also be supported by AMD APUs which are based on Vega graphics. Computex’s announcement left people wondering why Radeon RX 400 series were excluded from this list, after all, RX 480/470 essentially use the same Polaris GPUs as RX 500 series. AMD was quick to provide an update on this matter, the RX 400 series (RX480 and RX470) will, in fact, support FSR:
I just wanted to pass on the news that we are updating the list of compatible AMD graphics for FidelityFX Super Resolution (FSR) to include Radeon RX 470 and RX 480.
We expect the website to be updated with this information sometime soon and we wanted to make sure you were aware.
Looking forward to having you all try FSR out on the 22nd.
— Mickey Molad, AMD Community Manager
This statement did not confirm, however, whether FSR is coming to RX 560/RX 550 models as well. Things might get interesting once the code becomes available to developers. The technology will be supported by all DirectX 12/11 and Vulkan capable GPUs, this should include modern Intel integrated graphics as well as entry-level GPUs from NVIDIA, should both manufacturers be interested in FSR’s optimization.
The discussion on AMD FSR is ongoing. Some editors raise concerns that FSR will not live to the hype it generated over the past few months. While official videos and images do show that there is visibly lower image quality with FSR enabled we do not know how will FSR work within each game engine, each GPU architecture, and with combination AMD’s own image improving techniques such as Radeon Image Sharpening.
https://videocardz.com/newz/amd-fidelityfx-super-resolution-also-coming-to-radeon-rx-480-and-rx-470
It is strange how vested some posters seem to get, and to the lengths they have to go to defend or attack a company that they have no control over.
The mental gymnastics and irreverent comparisons to try and win points over issues that no one actually cares about is staggering.
It is a shame because it means it is almost impossible just to discuss the technology and examine the pros and cons. Instead it comes down to some mud slinging match when certain posters rabble on about AMD this, Nvidia that, console sales mean X, Y technology is dead. Made all the worse when those who shout the loudest seem to have no understanding of the underlying technologies they are trying to attack or defend.
Personally, I am very excited at what AMD have provided in FSR. Consumers can only win if both IHVs invest heavily in this technology and push the boundaries.
The facts are:
- Temporal accumulation combined with Deep-learning image reconstruction can provide equal or better than native + TAA
- The latest temporal super resolution methods as shown in UE5 looks great, without any ML being used at all
- Deep learning applied to a single image is the state of the art of spatial image scaling and was shown in DLSS 1
- AMD have more or less confirmed that FSR is only spatial, which provides an upper bound on quality. Small chance AMD have screwed up their marketing info
- Temporal methods require a more involved integration with the game engine to collect motion vectors, which may be why FSR doesn't use temporal accumulation. Nvidia had to put in a lot of effort to get plug n play DLSS drop in feature.
- AMD have image scaling patents based on deep learning, so it is possible that FSR is an deep learning based spatial filter like DLSS 1.0
- RDNA (and all GPUs) can do the matrix math required for DL inference, RDNA2 has some additional instructions that provide further acceleration. But the performance is not comparable to Tensor Cores. This has an impact on the model complexity and performance, but does not prevent a DL based super resolution model being present in FSR
- FSR require game engine integration, it is not driver level, just like DLSS. This may hint that in the future FSR could be extended to utilize temporal accumulation (or in fact already does)
- AMD's officially released image, even at ultra quality, are absolutely shocking. No better than the ancient bi-cubic scaling available for the last few decades. One can only hope this is a marketing failure. It is hard to imagine they will release FSR in that state, would be a disaster.
- AMD did support Epic in implementing TSR in UE5. If they have seen the quality that is possible then it is hard to believe they would eb happy with their released FSR screen shots.
- IF AMD limits to spatial information and doesn't use deep learning image reconstruction, then their available options are highly limited to existing linear algorithms with something like CAS applied on top. FSR is then effectively DoA
- If FSR only uses spatial information, then in game TAA will have to be applied for anti-aliasing, and all the pros and cons of TAA will get magnified.
I do start to wonder if FSR is actually not something more akin to UE5's TSR and something went wrong in the marketing material.
Caporegime
I think the mods have gotten fedup.
Soldato
urghh millsim roots
I mean... I somehow managed to get a speedball marker company to make the best magfed marker on the market, so it's not all bad lol.
------------
Definitely need to see more on FSR. Watching some reviews/analysis videos etc, it seems like the upscaling process is done at the very end of the pipeline, so I'm not really sure where it is going to get any extra fidelity from. It seems like it could only really cut down on jaggies from the increased res, plus 'with added smear'.
@D.P: Great list
And I agree. This forum is hard to discuss tech normally without it descending into argument, when it would be wayyy better to be a discussion about tech pro/con's. I have for a long time wished it that way.Going on the released slides, FSR does not look very good. And just like DLSS, they can't justify blurred or crappy image based on performance alone.
(edited 'you' to 'they' to ensure I meant AMD not you DP lol).
Last edited:
I mean... I somehow managed to get a speedball marker company to make the best magfed marker on the market, so it's not all bad lol.
------------
Definitely need to see more on FSR. Watching some reviews/analysis videos etc, it seems like the upscaling process is done at the very end of the pipeline, so I'm not really sure where it is going to get any extra fidelity from. It seems like it could only really cut down on jaggies from the increased res, plus 'with added smear'.
@D.P: Great list
Going on the released slides, FSR does not look very good. And just like DLSS, you can't justify blurred or crappy image based on performance alone.
My last gun was a dm9 or dm10 I got about 2012 but all sold now
The only one I have left is a side fed cooker pump with all belsales internals should really sell it tbh
Not going to judge fsr until its live but not time to wait
Soldato
PCGamer said "When we get all the way up to Performance mode, the uptick is impressive. From that native 49fps, we're seeing a 3X boost to 150fps. And there's no discernable difference in the side-by-side image fidelity."
https://www.pcgamer.com/amd-fidelityfx-super-resolution-announcement-release-date/
I really think they haven't had a good evaluation of FSR, as there are clear differences in image quality, at least to my eye.
Editorial on this and DLSS is all over the place. WCCFTech wrote about an image showing native, DLSS 1 and DLSS 2. "A comparison of DLSS variants vs native taken from Youtube. Notice DLSS 1.0 is far more palatable than AMD FSR and has very little blurring. "
In reality, the image described looks awful with DLSS 1 in control, and is a smeary mess. Not that I think much of wccftech anyway, but trying to trust basic judgement on image quality based on these tech sites is such a pain in the butt as they just make it up.
https://www.pcgamer.com/amd-fidelityfx-super-resolution-announcement-release-date/
I really think they haven't had a good evaluation of FSR, as there are clear differences in image quality, at least to my eye.
Editorial on this and DLSS is all over the place. WCCFTech wrote about an image showing native, DLSS 1 and DLSS 2. "A comparison of DLSS variants vs native taken from Youtube. Notice DLSS 1.0 is far more palatable than AMD FSR and has very little blurring. "
In reality, the image described looks awful with DLSS 1 in control, and is a smeary mess. Not that I think much of wccftech anyway, but trying to trust basic judgement on image quality based on these tech sites is such a pain in the butt as they just make it up.
If you are using the image in that article it is for the 1060 and according to AMD has not been optimised. It looks like DLSS 1.0 in that image and I'm not that impressed. The 3X they were referring too was from the 6800 XT image and quite frankly is ueseless for any real comparison. So we wait until proper independent testing is done.
Soldato
Couldn't agree more
I'm running a 2060, so I doubt I have much to gain, unless it turns out to work right out the box in games and look good.
The problem is AMD has seemingly sent full res uncompressed images to some reviewerd and even with the iltra quality the results were dire. Hopefully that is just some error on AMD's psrt, perhaps the quality names got mixed up.
I agree we need to wait until proper reviews.
However, it id absolutely not up tp Nvidia to improve the inage quality of FSR, performance yes, but if Nvidia have to completely re-write the FsR scaling technology to get usable imsge quality then FsRbis a complete failure. Driver tweskd to get a but nore performance sure, anything to fk with k sge quality, not at all.
I honestly don't get what the fuss is about this "technology".
The ability to boost FPS by reducing the visual quality has been a standard of 3D accelerators since day one. The fact AMD never had DLSS never put me off buying their cards because like FXAA/FSAA it always looked way to bad to ever be worth using (though the lack of PhysX did put me off for a long time, Nvidia really screwed up by abandoning that).
The ability to boost FPS by reducing the visual quality has been a standard of 3D accelerators since day one. The fact AMD never had DLSS never put me off buying their cards because like FXAA/FSAA it always looked way to bad to ever be worth using (though the lack of PhysX did put me off for a long time, Nvidia really screwed up by abandoning that).
Soldato
Dire? Have you seen more than what already been shown?
Because from what I have seen ultra quality on the 6800 looks decent enough.
Obviously I want to see more and even it running in action because end of the day that is the most important goal when playing games.
precisely as I inferred earlier from what I had read on the matter, FSR still needs additional anti aliasing over it - and since most games use TAA that will seriously limit its potential and it's another reason why DLSS has superior visuals.
As for pascal owners getting excited https://wccftech.com/no-amds-fsr-fi...-alternative-and-here-is-why-you-should-care/
AMD has backtracked and clarified that Nvidia cards will not support FSR unless Nvidia puts in work on a game by game basis to make it work, which they probably won't and AMD have said they will not do it - it's just more AMD open source software that Nvidia won't touch
Soldato
Holy **** - FSR isnt out yet with no publicly available white paper , but we have EXPERTS who know all about it already. /facepalm
Soldato
Comments are based on RDNA2 hardware capabilities and what AMD have said/shown of FSR. What is your problem with that?