DLSS 5 preview

[CAT-THE-FIFTH]

@Armageus Looks like John Linneman from DF was blindsided by the video and was not consulted about it:

Spoiler

He made some negative comments on bluesky but deleted them.

Looks like DF has made a follow up video:

DF Direct Q+A: The Big DLSS 5 ML Debate + Why We Should Have Waited With Our Coverage​

It's been a wild few days. When we posted about DLSS 5, we were impressed with the technology and what it was doing - but we posted too soon. When generative AI is so divisive and with the future of games development possibly at stake, we should have taken the time to get the story straight and to analyse what we saw more thoroughly drawing on the thoughts of the whole team. In this video, we try to do just that while tackling the big questions raised by our supporters.

Looks like John wasn't the only person who wasn't happy,Alex comments at 7 minutes 30 seconds in detail:

My initial gut reaction and permanant reaction is this isn't the same person.I think that is the key result most people will get out of that and the second thing I get out of that is that I played the game quite a lot,ah,in preparation for a lot of my videos covering it and I know that Grace in that shot there is reluctantly going to the house where her mother was murdered and she is not wanting to go there and is nervous,all these things. She is not even caring about anything other than just getting there and she quite obviously doesn't look like she is wearing things like make-up and what not and I honestly looked at the side by sides here and I thought I saw evidence of,other than the facial features being completely different I see a few things like evidence of mascara,I see evidence of what appears to be pursed lip stick...

I suspect,just like with DLSS1 and DLSS2,there needs to be refinements to the models being used.

 

[Trickle]

Just beat me to it.

Lulz at the mugshot.

 

[GordyR]

Did you actually read anything I said?

Yes. Hence the detailed response.

I didn't state an opinion on this and now you are argueing with me?

Eh?

What arguing? Can you you point to where I've argued anything with you?

I've calmly pointed out that the artistic style preference examples you gave; borderlands versus realistic, 2d isometric versus 3d, minecraft etc. are all without a doubt subjective art preferences, and I agree that that's little more than matter of taste, as are people's preferred genres; but that that's quite different to a measurable change in lighting fidelity, material detail or photorealism, and it's that change that I'm discussing - measurable, objective things; not one about a preference of style.

It's quite possible to think that something is demonstrably more photorealistic and detailed than something else, but to maintain a strong preference for the less photorealistic and less detailed thing.

I don't mean you specifically, but are people in this forum just incredibly over-sensitive when it comes to discussion and debate or something?

Instead of imagining what people are saying LOOK at what people are saying:

Search results for query: *

forums.overclockers.co.uk

I'm afaid I have no idea what you're responding to here or why are you linking to your own comments.

What people want in games is different and you don't seem to understand that.

Good grief man of course I do. Are you certain you actually read my post?

The entire thing was me agreeing with that fact, but explaining to you that we're talking at cross purposes, about two different things.

Moreover,any early tech has issues - so even if someone highlights issues,why is that a problem? This is an early tech demo,so OFC it has issues. It's not a shipping product. Its pretty much Alpha or Beta at best.

DLSS1 had issues,as did DLSS2. It was only after people highlighted these issues that DLSS3,DLSS4 and DLSS4.5 arrived. So highlighting issues,does not mean you are against the tech but would want improvements.

Again, where are you getting these ideas that anybody is taking any kind of umbrage with any of that? It's like you're responding to a completely different post.

Honestly, chill out. I agreed with every single word in your previous post and I still do. I just don't think it was relevant to what I was saying.

 

[humbug]

We would likely need a several-fold increase in GPU power to achieve the same levels of fidelity, lighting and realism using traditional rendering techniques in real-time.

That's why this is an exciting technology. If a developer wants the age to look different, they simply change the geometry, dial down the intensity or modify the settings etc.

Note though, that I still caveat this with the fact that we've yet to see it in fast motion.




In the DLSS 5 off screenshot, the character is completely flat and barely lit.

It looks like a mannequin and I can't make out any kind of potential age whatsoever; just a human shaped blob.




Forget this has anything to do with DLSS or AI and answer this question for me if you wouldn't mind.

Which of the two screenshots below do you think looks more realistic, and which would you say has the better visuals/graphics in the broad, colloquial sense?

Are you honestly telling me that in the before screenshot, you have a clear gauge of the age of the mannequin-like character, so much so that you can perfectly place him at 15, but in the after shot those visuals and the illusion of realism has been completely destroyed for you by the extra detail and better lighting? I mean, if that's really true then I honestly don't know what to say.

I'm beginning to wonder if perhaps the divide is simply due to the Uncanny Valley effect at work:

Uncanny valley - Wikipedia

en.wikipedia.org

Perhaps it's just that an awful lot of people are very sensitive to it, and are experiencing a kind of immediate revulsion to the sudden increase in realism.

if that is the case, then maybe the ridiculously overblown and absurdly hyperbolic YouTube community groupthink-like response that I've taken such umbrage with, makes a little more sense.

Neither of them look realistic, the latter looks like a different art style.

Have you thought that perhaps the game is supposed to look as it does? Hogwarts Legacy doesn't strike me as something that looks like its trying to be photo realistic, its soft look is an artistic choice.

 

[gpuerrilla]

I don't mean you specifically, but are people in this forum just incredibly over-sensitive when it comes to discussion and debate or something?

 

[GordyR]

Neither of them look realistic, the latter looks like a different art style.

Perhaps let's frame it a different way.

Do you think that there is any difference in realism between one or the other whatsoever? Is one more realistic than the other to you, or are they both truly equally unrealistic to you?

Have you thought that perhaps the game is supposed to look as it does, Hogwarts legacy doesn't strike me as something that looks like its trying to be photo realistic, its soft look is an artistic choice.

Of course. But it's interesting that you should say that...

Georgian Avasilcutei, a veteran game artist from Avalanche software, who developed Hogwarts Legacy, had this to say on the matter:

After this whole debate about DLSS 5, I came to the conclusion that most of the people talking about it are completely unaware of what they don't know…they're on the peak of ignorance and don't even grasp how little they understand. They just heard generative AI, and like Pavlov's dog, they just start drooling, thinking it's the same **** as unethical slop image generators…for the love of Christ…go and educate yourself before raging on the internet for no reason.

DLLS 5 is not a prompt-based generator…it's not creating stuff based on someone else's images and hallucinates results. It uses the information from the raster to build up a final render frame with the same information but with better lighting and shading…

I think that's maybe a bit strong, but who am I to argue?

Anyway, it's worth reading the full article for his complete take, especially given that we've just bought up what the artists for Hogwarts Legacy may or may not have originally intended for the game:

Here's a snippet:

In an X post, Avasilcutei attacked most of the people who are bashing NVIDIA's new technology for being ignorant of what's actually going on behind the scenes with it. He noted that, unlike what some folks seem to believe, this is no mere hallucination-prone, prompt-based generator, but instead a model that reuses the exact same information from the game to augment its lighting and shading.

To prove his point, he showed a comparison picture of a character model he personally worked on in two different lighting conditions: regular rasterization and ray traced lighting. The use of more accurate lighting, as well as hair and skin shaders, produces a result that, upon careful analysis, appears to have slightly different features from the original (the nose's ridge, for one). That happens without changing the actual facial shape at all. He also added that any artist would like to see their work in the best possible lighting, but that's usually impossible due to real-time rendering limitations that NVIDIA aims to supersede with DLSS 5.

Avasilcutei also posted a modified Dunning–Kruger effect graph where most of the people talking (negatively) about DLSS 5 are situated on the peak of "Mount Stupid", having absolute confidence but also nearly zero competence in the subject. The artist used it to underline, provocatively, that most DLSS 5 naysayers have failed to fundamentally understand the technology, stopping at a much more superficial level once they learned of AI involvement

I wont post the Dunning-Kruger chart that he did, as I suspect it'll upset a few people.

Anyway, here's the article in full:

"Most People Bashing DLSS 5 Are on the Peak of Ignorance" - Veteran Game Artist Shows Exactly How Much of a Difference Lighting Can Make

A veteran artist behind Dishonored 2 and Hogwarts Legacy fires back at DLSS 5 critics, showing exactly how much lighting can change a face.

wccftech.com

 

[Calin Banc]

I'm probably being dozy, but I'm unsure what your question is?

Edit* it's much easier to upscale a blurry image and add sharpening, many might even claim the final result is better than native I haven't played an upscaled game that has the same clarity as a well optimised game like Alyx.

From the top of my head, those percentages are for quality, performance and ultra performance ratio with DLSS. So, can a modified TAA provide the same or better image quality just by changing some minor settings?

If true, then AMD must have the biggest noobs working for them since FSR hasn't managed to come close until it went "hardware ran" with version 4. Same for Sony.

My gut tells me you can make it look good/decent at native and that's about it.

 

[humbug]

More realistic yes.

Again game developer or not its completely missing the complaint people have of it, "ACTUALLY you're wrong about the technicalities about how it works" as if that makes peoples complaint about it null and void, it doesn't, its deliberately obfuscating the complaint as a way to be dismissive of it because to actually argue with the result validates the argument.

From a studio heads perspective technology like this means you can hire cheaper and less experienced people, less of them, low effort and just have the AI make everything look the same as everything else.

 

[Sc00p007]

From the top of my head, those percentages are for quality, performance and ultra performance ratio with DLSS. So, can a modified TAA provide the same or better image quality just by changing some minor settings?

If true, then AMD must have the biggest noobs working for them since FSR hasn't managed to come close until it went "hardware ran" with version 4. Same for Sony.

My gut tells me you can make it look good/decent at native and that's about it.

I would be really intrigued to see Alyx upscaled from 33% via DLSS. My gut says it would look worse than native. As i said, i have never played a game that uses DLSS that has as clear an image as Alyx at native output. That would be the true test. As i mentioned, if your native output is butchered with underlying unoptimized technique, resulting in blurred image, then DLSS can and will make it look better. But if the render is built on optimised technique resulting in sharp 4k image, then DLSS will have a much harder time.

In VR, there is no hiding when it comes to clarity. Non of the games that look the best use DLSS. The games that do, even when using the latest presets, never look clear.

 

[opethdisciple]

As long as the developers are in control of the final DLSS 5 render of the faces then it's just the same as if they created them themselves the old fashioned way.

 

[Calin Banc]

I would be really intrigued to see Alyx upscaled from 33% via DLSS. My gut says it would look worse than native. As i said, i have never played a game that uses DLSS that has as clear an image as Alyx at native output. That would be the true test. As i mentioned, if your native output is butchered with underlying unoptimized technique, resulting in blurred image, then DLSS can and will make it look better. But if the render is built on optimised technique resulting in sharp 4k image, then DLSS will have a much harder time.

In VR, there is no hiding when it comes to clarity. Non of the games that look the best use DLSS. The games that do, even when using the latest presets, never look clear.

I'm not in the boat of "better than native". I think is DLAA looks best out of all DLLS mods.

My point is that DLSS's strength lays with his ability to improve performance via rendering the image at a lower resolution. So, let's say a 540p image or 720p or whatever, upscaled higher, does it look the same with that modified TAA as it does with DLSS? My guess is... no.

 

[Sc00p007]

I'm not in the boat of "better than native". I think is DLAA looks best out of all DLLS mods.

My point is that DLSS's strength lays with his ability to improve performance via rendering the image at a lower resolution. So, let's say a 540p image or 720p or whatever, upscaled higher, does it look the same with that modified TAA as it does with DLSS? My guess is... no.

yeh, I get what you are saying, but i think you are missing my point. I'm a little lazy right now, but when i have some time later ill have another stab at explaining it.

 

[Calin Banc]

yeh, I get what you are saying, but i think you are missing my point. I'm a little lazy right now, but when i have some time later ill have another stab at explaining it.

I agree with your point of running at native with a tuned TAA and being sharper than upscaled DLSS. I agree. DLAA is also native and looks better than lower presets.

My point is that when you don't have the power to run a game at native resolution, no matter how good the optimization is, then DLSS will be better than TAA. Hope is clearer now.

 

[Personality+]

Reddit - The heart of the internet

www.reddit.com

 

[CAT-THE-FIFTH]

@Calin Banc DLAA was originally what DLSS was meant to be IIRC. The upscaling component was a later addition.

Reddit - The heart of the internet

www.reddit.com

It's quite clear its in alpha/beta so needs some time to bake in the Nvidia oven. I think this was a more a investor/developer type preview IMHO and less for gamers.

Nvidia should have just revealed this during the summer when they had more time to work on it and get it into a better state.

 

[Sc00p007]

I agree with your point of running at native with a tuned TAA and being sharper than upscaled DLSS. I agree. DLAA is also native and looks better than lower presets.

My point is that when you don't have the power to run a game at native resolution, no matter how good the optimization is, then DLSS will be better than TAA. Hope is clearer now.

I'd go as far as to say that a badly tuned 4k is already displaying pq on par with well tuned @ 1440p.

 

[Tinek]

That's precisely how nVidia have stated it works. Why have you got it into your head that that *isn't* how it works?

Because of what I stated already and will state again. And they you for confirming you have absolutely no evidence of how they system actually works.

Once more....

The inputs are the original raster frame, and the motion vector frame. Nothing else.

Once more indeed... Original raster frame is not "just pixel colours", that's the point I'm making. For an AI model to know what it needs to improve and how exactly, it needs much more information than that. And Nvidia very clearly listed what it does (texture improvements, eye brows, eyes, subsurface scattering of the skin, hair, etc.) - ergo it knows where light sources are (to generate proper light, shadows, etc.), what is eyebrow, what is skin, metal etc. This is the bit you are just adamant to completely ignore. There's much more in that "black box" of theirs then they are saying.

Why do you think that isn't true? It makes perfect sense that's how it works.

It makes 0 sense at all the way you describe it. It might as well predict the future by looking at tea leaves.