OK DLSS 2.0 - Promising technology that ACC could benefit from

One of my biggest gripes with ACC has been clarity. It's been an issue from day one, and although improvements have been made across the board, I still find it as an issue that you have to really trade a lot of FPS, for image clarity.

So - Issues with specifically TAA are widely documented and discussed. And from this there have been plenty of improvements.

Unfortunally it seems that TAA is just not suited, and never has been, especially for something like a racing game / sim with fast moving objects and scenes.

Recently, Nvidia have started rolling out DLSS 2.0, and the results are very impressive.
Forget all about the original implementation which was absolutely garbage, this seemingly actually works, and from testing it myself, it works well in the few titles it has been implemented in.


Even though you (Kunos) cancelled the implementation of Ray-tracing, I hope that this technology is on the horizon for DLSS. Promising better image clarity, at high frame rates, almost sound too good to be true, but it really is true.

I'd love to hear from the devs, what their own thoughts are on the current state of sharpness in ACC, and if DLSS is something that they're currently considering to add to the 'todo'.




For anyone interessted in learning more about it, there're plenty of articles out there talking about the new version of DLSS.

A good place to start, with a longer video format is right here:



Cheers

 

ATAA looks very promising.. Hopefully this is just around the corner.

https://research.nvidia.com/publication/2018-08_Adaptive-Temporal-Antialiasing

 

Well, hopefully it will actually make TAA not actually useless. But I don't think it'll be relevant once DLSS is more widely available. The difference in quality is neglagable (visible, but arguably a very minor improvement) and the performance cost is a lot. Introducing Ray-tracing as a method of cleaning up the ****-stains that TAA leaves behind, sounds like a terrible and not very inefficient way of dealing with it.

EDIT:
Ontop of that, claiming that it'll stay within "33 ms frame budget" is useless. You know what 33ms is? 30 fps.. 16ms = 60 fps.

And arguably, we actually want way higher framerates than 60...


EDIT 2:
Moved on a little bit since the publication of the article you linked:
https://news.developer.nvidia.com/understanding-the-need-for-adaptive-temporal-antialiasing/

Seemingly it can introduce much higher fidelity than what TAA currently can, while staying with a 16ms frame budget.
Interessting for those PC's that can power through that and give a proper experience, I can see someone running this over DLSS. But personally, I'd take the FPS gain of 20-100% (depending on the resolution) of DLSS.

 

DELETE

 

... a little more info on ATAA. Sorry I didn't mean to hijack your thread.. either way it looks like a solution is on the horizon..

 

I don't think this is a good idea once you think about it, TAA runs on every video card, DLSS 2.0 will require an RTX card, if I understand how it works correctly.

Therefore, there is little improvement from Kunos point of view because most of their customers won't be able to access any new work carried out implementing DLSS into ACC, even if they did implement it, which is questionable because it may require them using a different branch and hence possibly break stuff to do all that for probably ~5% of the customer base isn't worth it.

Note: I'm not a developer or have any experience with the Unreal engine so my guess on the branch stuff might be wildly inaccurate

 

This is a really good point. I believe this type of technology which requires a specific patented card type from a specific manufacturer. will not be implemented until its available on all cards

 

DLSS was part of the reason I chose a 2080 over a 1080Ti, I would love to see it implemented, it’s not present in a single game I play

 

They proved they were willing to go there when ACC was first announced.

But if you're only implementing technology that everyone can run, you would not have ACC.
It's running Unreal Engine and even though ACC has seen many improvements, it's still a pretty far cry from running on any kind of system.

And to be fair, we're talking about implementing an additional Anti-Aliasing method. It is not uncommon for developers to implement technologies that are unique to one brand/series.

At the end of the day you have two card suppliers(maybe three in a few years if we're lucky), and RTX is only gonna get bigger.

 

https://news.developer.nvidia.com/dlss-2-0-ai-rendering/


good news can be implemented in acc ?

 

Please Kunos, as you know ACC VR performance isn't good enough,

Could you implement this feature?

Thanks

 

Why are you directing it to VR performance only? ACC lacks performance even on a single screen.
And why implement a thing for Nvidia users only?

 

ACC VR performance is even worse than single screen, need more help of development...

Because its a Nvidia technology...

 

Would love to see this implemented to help VR, but most of the users I know run TAA in VR to get the better clarity so i'm not sure if this helps. It looks to me like it only works with TAA which isn't ideal for VR at the moment.

 

Maybe I wasn't clear enough.
Covering symptoms with a bandage does not help to eliminate the cause.

And of course it is a nvidia technology only for nvidia users... you missed the point: Why implementing a new technology into a game where not all of the gamers will benefit from. It takes a huge amount of time and there will be loads of gamers still complaining because they do not run the game with a nvidia card or maybe a graphics card (i.E. 9xx/10xx) which is not capable of running DLSS.
If it were planed right from the beginning it would make sense to have a new technology to advertise this game but afterwards it's not worth the effort.

 

You right, but... Kunos couldn't improve performance (in VR) since ACC came out, if there is a way...

 

...all players should benefit, shouldn't they?

 

The argument that is for all or none is completely invalid, to say the very least.

Just keep the following in mind as a mere example: I don't have an HDR monitor, so, by your reason of thought, I should oppose to HDR implementation (since v1.2, not from begining) and Kunos should not have made the effort to add it.
Obviously this would be a preposterous position about HDR and the same can be said about any other added technology support that Kunos may conceive to implement.

Finally, Kunos didn't made any announcement, or said a word of any kind, about DLSS.

 

I think that any new technology should be implemented, you can use it or not
we must add not subtract

 

I think it's more of a question of resources, do you pull dev time from implementing other features or fixes or do you use that limited resource to implement a hardware specific technology.

If they add other features we all get to use them.
If they fix stuff we all benefit.
If they implement hardware specific features only those players who have the hardware will benefit.

Add it to the list of stuff to do but priority will always dictate what comes first.