• 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

Soldato
Joined
6 Feb 2019
Posts
17,466
AMD says FSR is coming sometime this year. They've given some updates

https://videocardz.com/newz/amd-fidelityfx-super-resolution-fsr-to-launch-this-year-for-pcs

* FSR is due to launch in 2021

* FSR does not use machine learning like DLSS or DirectML, nor fixed function AI cores, it's software based

* AMD wants to launch on PC first then consoles

* Even though it's due this year, a lot of work is still needed (AMD's words)

* A working version of FSR is already up and running, but it AMD isn't happy with the image quality, between now and launch AMD is working on improving the FSR image quality
 
Last edited:
Permabanned
Joined
2 Sep 2017
Posts
10,490
I think AMD must find a way to release Infinite detail technology to the conventional APIs.
The PlayStation 5 has already started to receive breakthroughs in this direction.
 
Caporegime
Joined
4 Jun 2009
Posts
30,926
Sounds like quite the wait still......

Also, sounds like it will be rather different to directml/dlss, be interesting to see the outcome!
 
Permabanned
Joined
2 Sep 2017
Posts
10,490
It's very difficult - it simply means that they must find ways to increase the performance by at least 30% but better >50% while keeping the same image quality.
 
Soldato
Joined
16 Jun 2004
Posts
3,215
A software based solution is always going to be fundamentally inferior to that of a hardware based approach.

People who think this is going to achieve DLSS 2.0 levels of performance ( & IQ levels) are in for a big disappointment, just as is the case with their weak hybrid software/hardware based Ray Tracing performance levels.
 
Caporegime
Joined
4 Jun 2009
Posts
30,926
Achieving good IQ can be done via a software based approach but the problem is the lack of hardware to enable good performance.

If it is anything like madvr (video renderer with up and downscaling for media content) and any device out there that does some form of image processing i.e. google pixel phones then you need the hardware to see the best results in terms of performance.

Think a lot of people will be disappointed to hear this though as many, including myself thought this was going to be based on directml (the real competitor to dlss) but seems not....
 
Associate
Joined
20 Jun 2016
Posts
1,308
AMD should just copy Sapphire TriXX Boost feature and be done with it.

Looks like it will be a long wait. They are probably prioritising consoles before they introduce it to PC.
Now there is a suprise....:D

Its well know Sony and Microsoft have AMD by the short and curlys over any decisions.
Starting with the manufacture all the way down to software lvl.
 
Caporegime
Joined
18 Oct 2002
Posts
29,679
People who think this is going to achieve DLSS 2.0 levels of performance ( & IQ levels) are in for a big disappointment, just as is the case with their weak hybrid software/hardware based Ray Tracing performance levels.

Here's a novel idea, maybe we can actually wait for its release before slagging the **** out of it?

:p
 
Soldato
Joined
30 Aug 2014
Posts
5,960
It sounds a bit disappointing so far, I was hoping for some machine learning involvement even as a final step to improve quality (even if this final step was only on RDNA 2, the rest could work on older architectures). Still, if it's in all or most games that's something so I'll wait and see.
 
Caporegime
Joined
18 Oct 2002
Posts
29,679
He’s right though, a software based approach is without doubt far inferior to a hardware based one

No doubt, but I'd like to actually see it in action and compared first. I can see the DLSS 'smearing effect' (most can't) so it's not really that useful to me (it's the high latencies involved in using the Tensor cores apparantly).
 
Associate
Joined
8 Oct 2020
Posts
2,282
Sounds like a wasted effort, otherwise it could be a tactical approach given the lack of an available platform to build a more intelligent solution on top of e.g. Direct ML.
 
Soldato
Joined
8 Jan 2003
Posts
3,692
Location
Scotland
He didn't entirely rule out ML, just suggested that ML isn't the only way to achieve it.

You don’t need machine learning to do it, you can do this many different ways and we are evaluating many different ways. What matters the most to us is what game developers want to use because if at the end of the day it is just for us, we force people to do it, it is not a good outcome. We would rather say: gaming community, which one of these techniques would you rather see us implement so that this way it can be immediately spread across the industry and hopefully cross-platform.
 
Soldato
OP
Joined
6 Feb 2019
Posts
17,466
Won't Nvidia utilise DirectML in the future? If so, why won't AMD?


AMD is not ruling out machine learning, my interpretation of what AMD said is that right now FSR does not utilise it but could in the future. This is more than likely because RDNA2 doesn't contain any Tensor like cores, but if RDNA3 does then FSR could be updated to use it - they're just giving themselves some wiggle room for the future
 
Back
Top Bottom