Support NeoGAF

[winjer]

What do you mean? They aren't biased /s

Incredible how incompetent these outlets are for upscaling analysis. Almost have to wait for Digital Foundry

HU analysis of FSR, DLSS and XeSS were very good. And very impartial.
What they showed today is limited by having to film a monitor, to capture images, instead of direct capture.
But I bet they are going to make proper comparisons, as soon as they get the final hardware and software. For both FSr4 and DLSS4.
And the analysis of HU for these upscalers is significantly better than DF's. At least they don't screw up and do those tests with motion blur enabled.

 

[llien]

Same as DF

Is like saying "same like shit".

HUB is THE SOURCE that had the most detailed DLSS-XESS-FSR analysis that is also very critical of the latter.

It doesn't get any "amd exclusive super preview before anyone else". Just a legit review source.

Unlike that PF shit.

 

[MadchesterManc]

Is like saying "same like shit".

HUB is THE SOURCE that had the most detailed DLSS-XESS-FSR analysis that is also very critical of the latter.

It doesn't get any "amd exclusive super preview before anyone else". Just a legit review source.

Unlike that shit.

Folk are probably still angry HUB generally don't use upscaling in their performance comparisons, unless comparing upscalers.

 

[Buggy Loop]

HU is good. Tim has some of the best content and knowledge when it comes to those upscaling tech.

They can be good, but not going into analysis is basically just a marketing machine for now. Just like Linus' 5090 hands on. You have to trust their word for it, but that's not really what we want.

DF gets a shitload of critique around here but their frame by frame analysis is pretty stellar for upscaling tech.

 

[SolidQ]

Unlike that PF shit.

 

[Buggy Loop]

HU analysis of FSR, DLSS and XeSS were very good. And very impartial.
What they showed today is limited by having to film a monitor, to capture images, instead of direct capture.
But I bet they are going to make proper comparisons, as soon as they get the final hardware and software. For both FSr4 and DLSS4.
And the analysis of HU for these upscalers is significantly better than DF's. At least they don't screw up and do those tests with motion blur enabled.

Yea I know they can be good, I'm pulling legs

This is much like Linus' video on 5090. It's just marketing.

I'm just getting impatient to get real data

 

[llien]

Who is this guy?

 

[winjer]

Yea I know they can be good, I'm pulling legs

This is much like Linus' video on 5090. It's just marketing.

I'm just getting impatient to get real data

Sorry, but it's not like Linus video.
Linus video is just 100% crap, made by a moron. Even the people from Nvidia looked liked they wanted to be somewhere else.
HU show of FSR4, was the best Tim, or anyone else, could do with such limitations.
It was still a professional take on an early view of the tech, done on the floor of a trade show.

 

[Celcius]

 

[SolidQ]

Who is this guy?

Behind FXAA and other thin things

 

[RobRSG]

And I imagine the births result in even more AMD shills, we now know how they multiply.

It's a very small number of shills, according to Steam Surveys.

 

[Radical_3d]

Using offscreen shots for an upscaling tech is beyond retarded. I guess I'll wait for GN.

 

[MikeM]

The only issue I have with FSR4 is that no matter how good it might be compared with FSR3, AMD are just so far behind NVIDIA with AI upscaling that I would personally find it near impossible to consider ever buying one of their GPUs. NVIDIA have a 5 year head start on AI upscaling and several years on ray-tracing too over AMD. It makes AMD GPUs second class products in my eyes. Fine for consoles but I would not want one in my gaming PC, even if they were much cheaper.

Cost matters in a world of ever increasing living costs. You may be immune, but that's not the reality for most.

If AMD undercuts significantly but gets close enough to DLSS, then it could work. Who knows?

 

[Bojji]

Using offscreen shots for an upscaling tech is beyond retarded. I guess I'll wait for GN.

You can still see massive obvious differences vs. 3.1. Looks much better.

 

[RoboFu]

 

[iQuasarLV]

Cost matters in a world of ever increasing living costs. You may be immune, but that's not the reality for most.

If AMD undercuts significantly but gets close enough to DLSS, then it could work. Who knows?

If these cards intro at 399 and 499 its going to be day one for me. They will have to undercut Nvidia that much to outclass their cards in both stack performance and cost. Especially since they are sticking with GDDR6.

 

[Insane Metal]

If that's real it's a massive difference. Glad AMD is catching up, even if it's just like 5% lol

 

[CrustyBritches]

already posted.

 

[Ddox Gameplays]

 

[llien]

You can still see massive obvious differences vs. 3.1. Looks much better.

The question is: did the inputs change?

It is very impressive if mere DLL swap was enough to do this upgrade.

It is a very different story if they feed motion vectors (and whatever other data) differently.

 

[Urban]

wake me up once ghosting is not a thing anymore

 

[Radical_3d]

You can still see massive obvious differences vs. 3.1. Looks much better.

I want a comparison with XeSS and DLSS. Beating non IA FSR is like beating a kid.

 

[Tazzu]

Will this work on Nvidia cards?

 

[winjer]

Will this work on Nvidia cards?

Probably not, as the Tensor units on RDNA4 are different from the units on RTX cards.
Then again, why would you want to use FSR4 on an Nvidia card, when there is DLSS4?

 

[Tazzu]

Probably not, as the Tensor units on RDNA4 are different from the units on RTX cards.
Then again, why would you want to use FSR4 on an Nvidia card, when there is DLSS4?

MFG wont come to 40 series so I thought AMD might have something similar.

 

[SolidQ]

I thought AMD might have something similar.

Who knows, but they was planned it release in FSR3

 

[winjer]