• 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

Just watched a video from JayZ where he basically spitballs about FSR. The only facts confirmed/alleged we can take from his video are as follows.

It will work on a larger userbase (confirmed fact)
It is easier to implement than DLSS. (see video for why)
10x developers have already signed up for it (not confirmed and no idea who)
AMD have claimed it will give similar clarity as DLSS and is easier to implement (not confirmed and I consider this BS until proved).

Now all we have to do is wait until it is released to see how it actually looks. Because nobody gives a crap if it doesn't use deep learning if it gives similar or close to similar results. If it is open source and works on cross platform and gets wider acceptance than the woefully supported and proprietary DLSS, well then that's a win.

I will consider it a win if it is better than DLSS 1.0 with room for improvement.

 
ICDP said:
Just watched a video from JayZ where he basically spitballs about FSR. The only facts confirmed/alleged we can take from his video are as follows.


It is easier to implement than DLSS. (see video for why)
Click to expand...

DLSS is just a click of a button for any game using unreal or unity engines, and for other games it has been shown just to taker a couple of days to add. We know that FSR requires developer support in the game, so it is very hard to beleive that FSR is going to be much easier than a click of button/2 days work.

The one caveat is that DLSS 2 requires motion vectors, so the game should already support TAA. FSR doesn't
 
Why do you keep ignoring the fact that DLSS is only limited to a very small subset of GPUs even if it is a click of a button in some (not all) game engines?

So how does that negate what I said for every other GPU and console and for every other game engine? It's like you ignore what's actually being said and go straight on the DLSS defensive. You actually proved my point by showing it is not widely implemented.
 
Aegis said:
DF are a bit more skeptical - they reckon at best it'll be DLSS 1.0-like.

https://youtu.be/JKXkbAI4Cek?t=664
Click to expand...


This is an obvious truth form the fact that AMD public stated that FSR does not use temporal accumulation , so being restricted to spatial data only the state of the art is using deep learning. Nvidia implemented a state of the art DL method in DLSS 1.0.

There were many problem with DLSS 1, such as an original goal of per game training, and lacking any additional post processing filtering to increase acuity. But the general method is about as good as you can get with spatial only data. So DLSS 1 is the performance bar for FSR. AMD might be able to get simlar quality, potentially better performance, and not make the mistake if going for per a game training regime.
 
ICDP said:
Why do you keep ignoring the fact that DLSS is only limited to a very small subset of GPUs even if it is a click of a button in some (not all) game engines?

So how does that negate what I said for every other GPU and for every other game engine?
Click to expand...


I didn't ignore it, I was just responding to one of these misplaced rumours that DLSS is still hard to implement.

The actual real-world number of GPUs that will support FSR is totally unknown. It is now clear AMD won't support FSR on Nvidia GPUs, and there is no way Nvidia will, so there is 80% of the PC space gone. Intel haven't confirmed anything but stated they are working on a deep learning based solution. Unreal 5 engine supports a very good TSR (and DLSS2) so there is likely no need to add inferior FSR support, which then impacts console usage etc. The PS5 has hardware accelerated checkerboarding, so FSR would have to be better than that to get good traction.
 
DLSS 1 isn't a very high bar quality wise with lots of bizarre artifacts/missing or smeared textures. I personally think FSR will retain more texture detail but with softer image.

We will see.
 
D.P. said:
I didn't ignore it, I was just responding to one of these misplaced rumours that DLSS is still hard to implement.

The actual real-world number of GPUs that will support FSR is totally unknown. It is now clear AMD won't support FSR on Nvidia GPUs, and there is no way Nvidia will, so there is 80% of the PC space gone. Intel haven't confirmed anything but stated they are working on a deep learning based solution. Unreal 5 engine supports a very good TSR (and DLSS2) so there is likely no need to add inferior FSR support, which then impacts console usage etc. The PS5 has hardware accelerated checkerboarding, so FSR would have to be better than that to get good traction.
Click to expand...

I'm putting you on ignore because you just keep posting the same guesswork nonsense masquerading as facts over and over. You have no idea if FSR will be added to Unreal or not but keep saying it won't as if it is a fact. You keep saying FSR has no application in consoles as if it is a fact, yet we already have rumours (some say confirmation) AMD are working with MS to add it to Xbox X and S consoles. I have even posted links showing this before and you keep ignoring them as they don't suit your Nvidia biased agenda.

https://www.techquila.co.in/amd-fsr-coming-to-xbox-series-x-s-consoles/

https://www.neowin.net/news/xbox-is-also-getting-amd-fidelityfx-super-resolution-confirms-microsoft/

https://www.geeky-gadgets.com/amd-fsr-technology-07-06-2021/
 
Last edited:
ICDP said:
I'm putting you on ignore because you just keep posting the same guesswork nonsense masquerading as facts over and over. You have no idea if FSR will be added to Unreal or not but keep saying it won't as if it is a fact. You keep saying FSR has no application in consoles as if it is a fact, yet we already have rumours (some say confirmation) AMD are working with MS to add it to Xbox X and S consoles.

https://www.techquila.co.in/amd-fsr-coming-to-xbox-series-x-s-consoles/

https://www.neowin.net/news/xbox-is-also-getting-amd-fidelityfx-super-resolution-confirms-microsoft/

https://www.geeky-gadgets.com/amd-fsr-technology-07-06-2021/
Click to expand...


You seem to have a reading comprehension problem. I never stated FSR has no application in consoles, in fact I was very clear and made no such statement and said the exact oposite that it is totally unknown what will happen on console since consoles already use technologies TAA, checkerboarding and potentially TSR. AMD can;t force game developers to use a technology, especially one that requires active development. It will be entirely up to the developer. If FSR is really fanastic then they will likeyl use it, but it is isn't as good as their own image scaling then they simply won't
 
Scougar said:
DLSS 1 isn't a very high bar quality wise with lots of bizarre artifacts/missing or smeared textures. I personally think FSR will retain more texture detail but with softer image.

We will see.
Click to expand...


DLSS 1 quality was generally OK but did have many artifacts. It was better than the prior state of the linear interpolation techniques overall, but with some added artifacts. It also didn't include any sharpening on the output which helped.

DLSS 1's problems were more related to the complex integration required; modifying code to extract additional data, sending an executable to Nvidia to then do the training and waiting for a model to come back and ship with the game. The DLSS process itself was relatively costly as well, which took a way some of the performance gains and meant it only made sense on very demanding games/settings.

So overall it was pretty terrible, but had a lot of potential given the known state of the art in deep learning.

It should be noted than DLSS 2.0 is not just DL based image reconstruction, but the temporal accumulation provides a massive quality improvement. DLSS 2 is not at all an evolution of v1.
 
Scougar said:
DLSS 1 isn't a very high bar quality wise with lots of bizarre artifacts/missing or smeared textures. I personally think FSR will retain more texture detail but with softer image.

We will see.
Click to expand...

Indeed, I would hope that AMD clear that low bar and keep improving as Nvidia did with DLSS2.0. If it ends up DLSS 1.0 for ever I would consider it a worthless technology because we can get better results from just upscaling and sharpening. What I am hoping for in FSR is a simple graphical option in game settings, that gives a decent FPS boost without too much loss in quality.

I mean even now if I try DLSS 2.0 in Cyberpunk I can see a blur effect compared to 4K native, especially in the desert areas. Though it's still a better compromise than running with RT off. In games like Death Stranding without RT, DLSS 2.0 is worse than 4K native and 4K native already runs very fluid for me. Though I can see the application for having it as it allows someone with a lower end GPU to get playable framerates.
 
@D.P. I am pretty impressed with the improvement with DLSS 2.0 over 1.0. my gripe is how they market it, and try to push the performance option when it looks terrible as soon as you move. Obviously then the lack if games that support it as well. Plus 4k gaming in meant to be about quality, I don't see the point in going there if we are sacrificing that just to say we are 4k.

FSR won't be able to compete with 2.0, and I am going to be just as critical of it as I was with 1.0 if FSR sucks. AMD need to start pumping money into their software department.

I have low expectations of FSR and I don't think they will be exceeded.
 
Scougar said:
@D.P. I am pretty impressed with the improvement with DLSS 2.0 over 1.0. my gripe is how they market it, and try to push the performance option when it looks terrible as soon as you move. Obviously then the lack if games that support it as well. Plus 4k gaming in meant to be about quality, I don't see the point in going there if we are sacrificing that just to say we are 4k.

FSR won't be able to compete with 2.0, and I am going to be just as critical of it as I was with 1.0 if FSR sucks. AMD need to start pumping money into their software department.

I have low expectations of FSR and I don't think they will be exceeded.
Click to expand...

I have a similar thought process to you. When I went 4K it was for clarity and sacrificing that clarity is not an option for me in most cases. DLSS 2.0 was a compromise but when you play a game like WDL and see the RT reflections I thought is was worth it. I see FSR succeeding where DLSS has failed, in that it will reach a bigger customer base and have a bigger adoption rate. Having said that it needs to give better results than DLSS 1.0 for quality.
 
DLSS has always been a fudge for not having the ability to display properly. I downscale (render at 6k and play at 4k) for better visuals. And no, DXR is not in the games i play,
 
Scougar said:
@D.P. I am pretty impressed with the improvement with DLSS 2.0 over 1.0. my gripe is how they market it, and try to push the performance option when it looks terrible as soon as you move. Obviously then the lack if games that support it as well. Plus 4k gaming in meant to be about quality, I don't see the point in going there if we are sacrificing that just to say we are 4k.

FSR won't be able to compete with 2.0, and I am going to be just as critical of it as I was with 1.0 if FSR sucks. AMD need to start pumping money into their software department.

I have low expectations of FSR and I don't think they will be exceeded.
Click to expand...


The performance option provides good marketing material, so it is easy to see why.

I get the issue with 4k, but the fact is that Moor's law is dead so big advances in GPU raw performance are gone, and we will be looking at smaller performance increases and less frequent GPU releases. DLSS is a way of jumping 2-3 generations, which then allows the increase in graphics quality. It is really a tradeoff, that to enable RTX for example then to get playable framerates at 4K then DLSS is necessary, otherwise we end up waiting another generation or 2.

DLSS also has to be seen in light of what consoles do - all the major games use image scaling to get to requisite frame rates. DLSS brings the same concept to PCs but does much better than any linear scaling technique. Similarly, modern game engines can't apply MSAA, so if you want AA you have to use TAA. AND many games are designed around TAA as a requirement because there will be significant pixel inconsistencies frame to frame . TAA is not easy and has its own problems, DLSS is a combination of TAA and image sclaing that often does better than the underlying TAA.


Of course in a perfect world we would just get a far more powerful GPU with the grunt to do everything at native resolution. But that simply isn't possible now. And even if it was, DLSS essentially gives you that 2-3 generation advantage.
 
ICDP said:
I see FSR succeeding where DLSS has failed, in that it will reach a bigger customer base and have a bigger adoption rate.
Click to expand...


Do you have a time machine for that?

All well and good posters complaining about speculation of FSR, but then the biggest advocates of FSR are the ones not using facts.

FSR might have a bigger customer base, but we won't know. And this depdns on many caveats that FSR is any anyway half decent, even on Nvidia hardware.
 
D.P. said:
Do you have a time machine for that?

All well and good posters complaining about speculation of FSR, but then the biggest advocates of FSR are the ones not using facts.

FSR might have a bigger customer base, but we won't know. And this depdns on many caveats that FSR is any anyway half decent, even on Nvidia hardware.
Click to expand...

Console markets huge.
 
You must log in or register to reply here.
Back
Top Bottom