Nvidia DLSS 1 and 2 antialiasing discussion *spawn*

I'm not sure you understand how it works. Nvidia no longer train specific games as their AI model for motion-based image reconstruction is now seemingly advanced enough that it is game agnostic. That tech is proprietary and Nvidia use it as a selling point for their RTX GPUs, taking advantage of the Tensor cores to accelerate it.

Game devs can't simply "use" directML to achieve the same thing. There's no models available for them to utilize, Nvidia aren't about to release their DLSS model for everyone to use. Could each game dev hire some AI model programmer to train for their game and provide that to directML for any GPU drivers that support it? Sure. I can't see that happening much though, outside of some tech experiments by large publishers.


Hunh...?

No, I don't think you understand. NVidia doesn't make games. And they can't pay every developer to use their model, and NVidia can not AI every game. Otherwise, in over 20 months of DLSS and Turing, we would have more than 6 games..!

So, NVidia has already proven, that they are incapable of doing it for each game (DLSS 3.0 remains to be seen). But if you read the article you will see that Microsoft is using Azure to train AI models in near instant.

Lastly, I do NOT see Game Dev's catering to 5% NVidia owners with $1,200 cards.
 
I'm fairly certain Malo isn't the one misunderstanding things.

It's not a technical achievement, it about a sustainable business model.
Why hasn't DLSS proliferated over the last 23 months..? And why does NVidia get a free pass to work with Game Developer's on AI training, but not Microsoft..?

Explain that, then there will be no misunderstanding.
 
You need to realize things drastically changed with DLSS 2.0, as Malo told you, but you opted to ignore and mock them as being confused. Don't expect anyone else to waste their time trying to explain to you with that kind of attitude.
 
You need to realize things drastically changed with DLSS 2.0, as Malo told you, but you opted to ignore and mock them as being confused. Don't expect anyone else to waste their time trying to explain to you with that kind of attitude.

OK... But I do not see why posing a counter argument & business model, is having an attitude. Malo has NOT explained why Microsoft can't... (for his argument to be viable).

DLSS is proprietary and DX12 U is not. They are both capable means of achieving the same results.
 
OK... But I do not see why posing a counter argument & business model, is having an attitude. Malo has NOT explained why Microsoft can't... (for his argument to be viable).

DLSS is proprietary and DX12 U is not. They are both capable means of achieving the same results.
I agree with you and others here who mentioned MS, mainly Azure and their potential for training for use in directML. It's probably the most viable means of getting something like DLSS to all players. Possible issues have been discussed already so not going into them again.
 
It's not a technical achievement, it about a sustainable business model.
Why hasn't DLSS proliferated over the last 23 months..? And why does NVidia get a free pass to work with Game Developer's on AI training, but not Microsoft..?

Explain that, then there will be no misunderstanding.
The challenge is that there are requirements for DLSS to work. If you don't want to invest the time to change or create a separate rendering pipeline for it, you won't support it. That process is listed here: #1019


It may be true that companies are likely more interested in a vendor agnostic solution before jumping on board, but there is currently no model that supports that. And won't for the foreseeable 6-8 months.
 
As has been mentioned before DLSS is not a mature feature and is continuing to evolve. Supposedly DLSS 3.0 will involve even less developer input and can work with any game, aside from being more performant. The next few weeks should be fairly interesting ....
 
The only problem with MS or Google providing the service is they will likely prioritize exclusives and developers who have agreements to running on their platforms. Not sure competitors like Sony, Tencent, etc... or independent developers would receive the same level of support.
 
As has been mentioned before DLSS is not a mature feature and is continuing to evolve. Supposedly DLSS 3.0 will involve even less developer input and can work with any game, aside from being more performant. The next few weeks should be fairly interesting ....

DLSS 3.0 is someone's made up fantasy. We have no idea if/when the next version will show up or what it will do.
 
DLSS 3.0 is someone's made up fantasy. We have no idea if/when the next version will show up or what it will do.
There’s a slew of things that DLSS needs to be accommodated for to work as per the engine requirements slide deck. So I definitely expect more DLSS models to address short comings.
 
Exactly, and at this point most of what we know regarding Ampere RTX is rumor. That's why the next few weeks should be interesting, though my bet is there will be another iteration of DLSS.

Even if it's not completely game agnostic, just getting it supported in a major engine like UE5 would be a very significant step.

I'm not really expecting that to happen though. Unfortunately a game agnostic DLSS 3.0 also seems a little too good to be true so I'm not expecting that either. Although if it where real it'd be a complete game changer.

The next few weeks and months are going to be the most interesting we've had in the hardware space for years. Ampere, RDNA2, Zen 3, Rocketlake, both new consoles, DirectStorage, next gen SSD's... what a time to be alive!
 
That's somewhat surprising, this is maybe the first DLSS 2.0 comparison I've seen where TAA actually looks sharper.
This is not DLSS 2, the developer of F1 2020 only implemented DLSS 1, there is no mention of DLSS 2 ever in any of the press releases for F1 2020. NVIDIA didn't even care to highlight any of the image and quality comparisons they usually do when they use DLSS 2.

https://videocardz.com/press-release/f1-2020-gets-nvidia-dlss-support

The developer just used the defunct DLSS 1 pathway that was present in F1 2019, this is just a lazy port to F1 2020.
 
Last edited:
This is not DLSS 2, the developer of F1 2020 only implemented DLSS 1, there is no mention of DLSS 2 ever in any of the press releases for F1 2020. NVIDIA didn't even care to highlight any of the image and quality comparisons they usually do when they use DLSS 2.

https://videocardz.com/press-release/f1-2020-gets-nvidia-dlss-support

The developer just used the defunct DLSS 1 pathway that was present in F1 2019, this is just a lazy port to F1 2020.

lol well that explains it. :)
 
I'm fairly certain Malo isn't the one misunderstanding things.
Agreed.
Long time lurker, love the info/debates on this forum but I had to register (in order to use ignore function) because the amount of "noise" has gone up lately here.

Now I can enjoy the good posts without the "noise" mixed in.
 
Agreed.
Long time lurker, love the info/debates on this forum but I had to register (in order to use ignore function) because the amount of "noise" has gone up lately here.

Now I can enjoy the good posts without the "noise" mixed in.

I did the same around the time of the PS4/XB1 release (new consoles seem to be the common thread). This forum is amazing because people with legitimate insight post regularly. The rest of us are better off just learning.
 
  • Like
Reactions: HLJ
While the mildly uninformed posts are indeed annoying, the occasional blatantly ignorant or fanboyish post kinda brightens up my day.

The 2 posters in question did not settle for a single post here and there...they have gone full Dunning-Kruger and it was getting an annoying amount of "noise" that I could not mentally ignore anymore.

I see it happening on other forums too, but the first time I have experiened it here.

I will stop the offtopic posting now and enjoy the noise free informative posts :)
 
Back
Top