AMD's FSR3 possibly next month ?

[mrk]

The solution is just get a 4090 and laugh at all the trivial stuff

 

[CAT-THE-FIFTH]

The solution is just get a 4090 and laugh at all the trivial stuff

Then spend another £2000 on a PC not to bottleneck it!

 

[mrk]

That's not necessary Mine isn't a £2000 PC (omitting the GPU) and isn't bottlenecked beyond the reach of 100fps+

 

[CAT-THE-FIFTH]

That's not necessary Mine isn't a £2000 PC (omitting the GPU) and isn't bottlenecked beyond the reach of 100fps+

£1999 does not count!

 

[GhostDog1981]

I’ve been waiting for FSR 3 for sooo loooong it’s unreal.

When are we going to get it? And will it be retroactive on all past AAA games?

 

[Troezar]

I’ve been waiting for FSR 3 for sooo loooong it’s unreal.

When are we going to get it? And will it be retroactive on all past AAA games?

I'm hoping I'll never need any of the image downgrade software

 

[humbug]

I'm hoping I'll never need any of the image downgrade software

Its baked in now for both, you get less shaders and more image downgrading to make up for it.

Because like sheepole we affirm everything Nvidia do in downgrading our quality of GPU's and then criticise AMD for not being as good as Nvidia at it.

We should have said very loudly and clearly no to DLSS the moment it appeard, instead we criticised AMD for not having it, i knew at the time this would lead to slower more expensive GPU's, i warned of it.

Now here we are, i have a first gen DLSS card, for the same $500 it cost me i get 20% more 'native' performance, 4 years later.

 

[mrk]

What do you mean by downgrade though? Because DLSS Quality/XeSS Quality, FSR Ultra Quality etc can and do yield in better than native sharpness and detail - Although this does depend on your working resolution, 1080p is not a resolution for getting good upscaler results for example. 1440p at the minimum.

Using upscalers in performance modes is only viable at 4k resolution whereby the internal render res is higher so the AI system has a higher LOD to work with and produc e a better than native (or same as) output reconstruction.

This comparison has been done to death by the community and pros alike, DF have gone into great depth about it showing side by side and there is no argument (other than some people on forums) that upscaling can and does produce as good as, if not better than, native resolution image. Upscaling also offers a superior form of AA without the big performance impact of say MSAA, or the woeful TAA implementations of internal methods in most engines which result in loss of sharpness and other stability issues.

Of course this all depends on the developer implementing an upscaler properly, which not even vendor sponsored games seem to adopt a lot of the time - The latest being Starfield where there is still some aliasing and image artefacting on water reflections on the ground etc when moving the camera fast which is amplified if upscaling is enabled.

Bottom line is that upscaling works, it's free fps gains at no cost to image quality /if/ implemented right. The if is the big part, because it requires developers to actually do some work, instead of just make the feature available on a toggle.

 

[Mesai]

Its baked in now for both, you get less shaders and more image downgrading to make up for it.

Because like sheepole we affirm everything Nvidia do in downgrading our quality of GPU's and then criticise AMD for not being as good as Nvidia at it.

We should have said very loudly and clearly no to DLSS the moment it appeard, instead we criticised AMD for not having it, i knew at the time this would lead to slower more expensive GPU's, i warned of it.

Now here we are, i have a first gen DLSS card, for the same $500 it cost me i get 20% more 'native' performance, 4 years later.

You're assuming game studios wouldn't still be releasing unoptimised beta products that need upscaling tools. Stronger GPUs rendering poorly made games at native resolution is just kicking the can down the road. Either way companies are going to look to extract every penny they can from the customer.

AI and ML in gaming is a useful tech as it's adaptive.

 

[Troezar]

Its baked in now for both, you get less shaders and more image downgrading to make up for it.

Because like sheepole we affirm everything Nvidia do in downgrading our quality of GPU's and then criticise AMD for not being as good as Nvidia at it.

We should have said very loudly and clearly no to DLSS the moment it appeard, instead we criticised AMD for not having it, i knew at the time this would lead to slower more expensive GPU's, i warned of it.

Now here we are, i have a first gen DLSS card, for the same $500 it cost me i get 20% more 'native' performance, 4 years later.

Yes I'm very surprised at anyone thinking it was a good idea. If you GPU can't run it native then the hardware is the issue not the software. I can see a use case for lower end older GPU's where it helps them out but not new flagships ones. My plan is to just stay behind the curve. Stick with 1440p and older reduced price GPU's to run it, as the gaming gods intended

 

[CAT-THE-FIFTH]

Yes I'm very surprised at anyone thinking it was a good idea. If you GPU can't run it native then the hardware is the issue not the software. I can see a use case for lower end older GPU's where it helps them out but not new flagships ones. My plan is to just stay behind the curve. Stick with 1440p and older reduced price GPU's to run it, as the gaming gods intended

The same PC crew were laughing at consoles saying the hardware sucks because they need upscaling because of poor hardware. Now PCMR saying consoles were right all along!

Some of us told people,upscaling will used to sell less for more. The RTX4060TI proves this. But it's "poorly" optimised games. So poorly optimised my RTX3060TI can beat it,and so can an RTX3060!

 

[mrk]

The consoles can only upscale from an already lower resolution though that's the issue, which results in a maximum possible quality level which when looked at closely, can scrutinise image artefacts and stuff (again as shown by the pro reviewers). You can't upscale at native 1440p and use FSR Quality on consoles to get 60fps, for example. You have to factor in the variances at play rather than brush every criticism with the same paint.

 

[Troezar]

Of course this all depends on the developer implementing an upscaler properly,

There is the elephant in the room KISS is what I say.

 

[Zarax]

To see the absurdity of people arguing upscaling is better than native one should simply try something similar in a simpler environment:
Would you think that MP3 + some kind of reconstructing filter would sound better than uncompressed audio?

Go ahead and try to win that battle in any kind of audiophile community then come back and tell us the results. Mind you, this is a far simpler computational problem than what AA is, so according to the "better than native" community it should be easy peasy to argue.

 

[CAT-THE-FIFTH]

The consoles can only upscale from an already lower resolution though that's the issue, which results in a maximum possible quality level which when looked at closely, can scrutinise image artefacts and stuff (again as shown by the pro reviewers). You can't upscale at native 1440p and use FSR Quality on consoles to get 60fps, for example. You have to factor in the variances at play rather than brush every criticism with the same paint.

But consoles are played on TVs which are metres away - the PC is right in front of your face. So you are more likely to notice issues IMHO. But even if we take what you say at face value,Houston we have a problem!
The lower resolution affects PC too:

In terms of pure TFLOPs,the PS5 is around the RX6600XT and the XBox Series X is around the level of an RX6700XT:

https://tpucdn.com/review/amd-radeon-rx-6700-xt/images/relative-performance_1920-1080.png

So only the RTX3070 and RTX3080 are really significantly faster,and perhaps the RTX3060TI is broadly comparable.

The rest are not that great,so will have the same issues as any console even if upscaling is enabled. They lack dGPU horsepower. Also all those GTX cards are locked out of DLSS,so have to rely on FSR!

The RTX4060/RTX4060TI/RX7600 are a backwards leap.

 

[Nexus18]

What do you mean by downgrade though? Because DLSS Quality/XeSS Quality, FSR Ultra Quality etc can and do yield in better than native sharpness and detail - Although this does depend on your working resolution, 1080p is not a resolution for getting good upscaler results for example. 1440p at the minimum.

Using upscalers in performance modes is only viable at 4k resolution whereby the internal render res is higher so the AI system has a higher LOD to work with and produc e a better than native (or same as) output reconstruction.

This comparisons has been done to death by the community and pros alike, DF have gone into great depth about it showing side by side and there is no argument (other than some people on forums) that upscaling can and does produce as good as, if not better than, native resolution image. Upscaling also offers a superior form of AA without the big performance impact of say MSAA, or the woeful TAA implementations of internal methods in most engines which result in loss of sharpness and other stability issues.

Of course this all depends on the developer implementing an upscaler properly, which not even vendor sponsored games seem to adopt a lot of the time - The latest being Starfield where there is still some aliasing and image artefacting on water reflections on the ground etc when moving the camera fast which is amplified if upscaling is enabled.

Bottom line is that upscaling works, it's free fps gains at no cost to image quality /if/ implemented right. The if is the big part, because it requires developers to actually do some work, instead of just make the feature available on a toggle.

Sadly, these ones will never accept it despite EVERY single tech press (TPU, computerbase, pcgamershardware, DF, gamer nexus, HUB, oc3d.net etc.) and comparison out there showing exactly where, when and how upscaling (well more so dlss) can produce as good as native or better than native, I have yet to see anyone who claims otherwise posting some of their own substantial evidence to debunk all the material out there now.....

 

[Freddie1980]

Starfield could have been the poster child for FSR 3 given how demanding it is. AMD really should have made sure it was ready for release at the same time as Starfield, can't help but feel AMD missed an open goal here.

 

[humbug]

Yes I'm very surprised at anyone thinking it was a good idea. If you GPU can't run it native then the hardware is the issue not the software. I can see a use case for lower end older GPU's where it helps them out but not new flagships ones. My plan is to just stay behind the curve. Stick with 1440p and older reduced price GPU's to run it, as the gaming gods intended

Its getting ever more difficult to A, extract more performance per-shader and B, pack more shaders in.

GPU's are concepted years before they land on shelves, 5 years perhaps, AMD and Intel had the idea that to get around this the answer was MCM GPU's, remember ARC was meant to be MCM, Intel failed and gave up, no criticism there, that #### is hard yo, not everyone can do it, AMD achieved a first step in that, like a Zen 1 sort of thing, RDNA 4 was meant to stage 2, like Zen 2, they have it working in labs but have decided to delay it until RDNA 5 because it needs more engineering work, however that also means RDNA 4 will only get one of the MCM chiplets making it a very mid range card again.... its likely to be short lived as RDNA 5 is pushed forward.

Seriously hats off to Intel for trying, its the right thing to do, but they can't even get it working on CPU's and GPU is far more difficult, they will keep going and i have no doubt will get there.

Nvidia decided they didn't want to put the engineering work in, or simply don't have the skill to do it, instead they decided to draw on their talents, software and marketing, the result is DLSS, now AMD have to follow suit or get left behind, and they just aren't as good as Nvidia at this software stuffs....

Suffice to say i prefer the AMD / Intel solution.

 

[mrk]

To see the absurdity of people arguing upscaling is better than native one should simply try something similar in a simpler environment:
Would you think that MP3 + some kind of reconstructing filter would sound better than uncompressed audio?

Go ahead and try to win that battle in any kind of audiophile community then come back and tell us the results. Mind you, this is a far simpler computational problem than what AA is, so according to the "better than native" community it should be easy peasy to argue.

You can't compare audio streams to visual streams. You can't reconstruct audio layers the same way you can a predictive visual model like an image which looks identical or better than the native version. Plus with compressed audio at a certain bitrate, the human ear is incapable of telling the difference anyway, whereas image reconstruction is only limited to the vision capability of each individual person.

 

[CAT-THE-FIFTH]

Starfield could have been the poster child for FSR 3 given how demanding it is. AMD really should have made sure it was ready for release at the same time as Starfield, can't help but feel AMD missed an open goal here.

Another Marketing Disaster.

You can't compare audio streams to visual streams. You can't reconstruct audio layers the same way you can a predictive visual model like an image which looks identical or better than the native version. Plus with compressed audio at a certain bitrate, the human ear is incapable of telling the difference anyway, whereas image reconstruction is only limited to the vision capability of each individual person.

Well you can compare it to visual streams. I am interested in photography. I could go and take a picture of a complex scene with dSLR or mirrorless camera at 24MP. This is an image formed by light reflecting off the object onto the sensor. You can take a picture with film and get and the same image.

Then I could take the same scene at 6MP,and use some of the machine learning algorithms to reconstruct it to 24MP. Are you going to get a better than a native version of a complex scene? No you will be getting a "version" of the image which is an approximation of the image. You can always do a simple subtractive comparison to see if they are exactly the same. If they are it will give you a blank output.

Even with modern digital sensors,the colour is predictive(colour arrays)- but in the technical imaging I worked in years ago,we would rather use full colour filters because it causes issues(even most space probes use colour filters).

The same goes if you took 4K video,which has random movements of people,etc in it. You could do reconstruction there and even frame insertion,but machine algorithms will still have to predict what is happen. They are not so great at random movement. I remember friends working on machine learning years ago - there are still fundamental limitations to what you can do.

Its getting better,but again its always a good estimate.