Nvidia starts to support DLSS in VR

[Fri13]

 

If the technology is not only for one vendor (Nividia or AMD) then it should be implemented if just possible. 

If it is just for one vendor but not for other, then it shouldn't be implemented if it is not very simple and it is causing extra performance loss to another. 

 

And that is it. 

Edited May 20, 2021 by Fri13

[Lurker]

32 minutes ago, Fri13 said:

 

If the technology is not only for one vendor (Nividia or AMD) then it should be implemented if just possible. 

If it is just for one vendor but not for other, then it shouldn't be implemented if it is not very simple and it is causing extra performance loss to another. 

 

And that is it. 

 

 

That's not how this works. There is no performance loss if you use AMD, however there would also be no performance gains. If Eagle Dynamics have stated that they will not use proprietary technology (such as DLSS) then that is their prerogative, but this is decision has nothing to do with the technology itself, and more to do with costs\time\resources used as well as simply a company wide stance on these types of technologies. 

 

If this is indeed the case, then this thread should probably be closed. 

[Fri13]

Just now, Lurker said:

That's not how this works.

 

You just said that it works as I said it does...

 

Just now, Lurker said:

There is no performance loss if you use AMD, however there would also be no performance gains.

 

Hence no reason not to implement it as it is not causing problems for AMD users.

 

Just now, Lurker said:

If Eagle Dynamics have stated that they will not use proprietary technology (such as DLSS) then that is their prerogative, but this is decision has nothing to do with the technology itself, and more to do with costs\time\resources used as well as simply a company wide stance on these types of technologies. 

 

Yes I know that, why if you read carefully my post, I wrote: "It shouldn't be implemented if it is not very simple".

If it takes a 50 hours from one to implement it, giving a Nvidia users XX % performance increase while AMD users receive 0%, then it is worth it.

If it takes 5000 hours from five to implement it, then it is not sensible thing to do. 

 

Understand the point?

 

Just now, Lurker said:

If this is indeed the case, then this thread should probably be closed. 

 

That is the case. ED has stated that they do not implement anything that is benefitical for one but not to other. So they want to walk the "middle path" that leads that both sides are suffering as neither one will get the features they would benefit from. 

 

These things should be case-by-case considerations by ED instead hardline policies.

 

Hypothetical example, if we would be talking here that Nvidia that has 80% of the market share, would benefit from this feature in VR by 300% performance increase, but nothing for AMD. And it would be just a few hours from couple programmers to implement it. Why shouldn't they?

 

It is a feature and capability in the hardware that someone decided to buy. It is not about does a DCS World work with only Nvidia or AMD.

Otherwise we can stop right here every development because Nvidia 7xx series card owners can't anymore play DCS World in VR. The Windows 7 lost its support. The Facebook account is required for Rift Owners etc etc...

 

If one's hardware offers performance benefits that doesn't affect others, and it is acceptable implementation (it is provided by the Microsoft, Nvidia, AMD or someone else for easy work) then it should be considered is it to be added.  

[ShaunOfTheFuzz]

DLSS 2.0 would be a complete game changer for Nvidia 20XX+ series owners, including future cards as this is a core Nvidia technology. 

Currently 74% of the gaming market is on Nvidia, and my experience in the DCS community bares that out, DCS players are overwhelming Nvidia. There is only upside to using DLSS 2.0, to the tune of 50%+ performance gain. The gains should be especially drastic in VR, where resolution is such a limiting factor. 

IMO, ED's stance on proprietary tech is cutting their nose off to spite their face. If the tech can help a large number of your community (and more since the quest 2 is driving VR adoption) and the barrier to implementation is relatively low, then at least look at it. 

DLSS 2.0 has to potential to be more transformative for VR users than the core engine rework as the vast majority of VR users will be resolution limited.

[ShaunX]

23 hours ago, 5ephir0th said:

 

 

Nah, it´s not even close, without ReBar your CPU can access to only 256MB of VRAM already, with ReBar the CPU can access to all VRAM so in some scenaries, without even optimize to it, there can be a slight enhance.

On the other hand for DLSS ED needs to work with nVidia, as nVidia has to run some game code over his machine learning systems and, once done, the nVidia driver must has support for it as far i remember so i think it´s impossible if ED don´t make it happen

 

According to this article, DLSS 2.0 (and it is version 2.1 that is required for VR) https://www.nvidia.com/en-gb/geforce/news/nvidia-dlss-2-0-a-big-leap-in-ai-rendering/

 

"

One Network For All Games - The original DLSS required training the AI network for each new game. DLSS 2.0 trains using non-game-specific content, delivering a generalised network that works across games. This means faster game integrations, and ultimately more DLSS games

"

 

..suggests maybe not such a big effort for DCS VR.

 

[ShaunOfTheFuzz]

Just now, ShaunX said:

 

According to this article, DLSS 2.0 (and it is version 2.1 that is required for VR) https://www.nvidia.com/en-gb/geforce/news/nvidia-dlss-2-0-a-big-leap-in-ai-rendering/

 

"

One Network For All Games - The original DLSS required training the AI network for each new game. DLSS 2.0 trains using non-game-specific content, delivering a generalised network that works across games. This means faster game integrations, and ultimately more DLSS games

"

 

..suggests maybe not such a big effort for DCS VR.

 

 

Correct, DLSS 2.0 does not require you to send code to Nvidia. Implementation is far easier than 1.0.

[Fri13]

Just now, ShaunOfTheFuzz said:

 If the tech can help a large number of your community (and more since the quest 2 is driving VR adoption) and the barrier to implementation is relatively low, then at least look at it. 

 

Only as long that such tech does not cause performance loss to a other brand users.

 

Like if this would be 150% increase to Nvidia users and -20% loss for AMD, then it is No-go. 

If this is 50% for Nvidia users and 0% loss for AMD, then it should be a green light. 

 

 

[ShaunOfTheFuzz]

1 minute ago, Fri13 said:

 

Only as long that such tech does not cause performance loss to a other brand users.

 

Like if this would be 150% increase to Nvidia users and -20% loss for AMD, then it is No-go. 

If this is 50% for Nvidia users and 0% loss for AMD, then it should be a green light. 

 

 

 

DLSS has no direct impact on AMD users as can be seen in titles like Cyberpunk and Metro Exodus. DLSS is an upscaling algorithm that runs on Nvidia hardware, it has to be actively switched on in the game menu and is not passive.

 

The only potential impact is the side effect of a game developer neglecting one GPU manufacturer for another on the basis of assuming people will buy the most suitable hardware for the tech they've implemented, but that's on the developer. 

 

[5ephir0th]

29 minutes ago, ShaunX said:

 

According to this article, DLSS 2.0 (and it is version 2.1 that is required for VR) https://www.nvidia.com/en-gb/geforce/news/nvidia-dlss-2-0-a-big-leap-in-ai-rendering/

 

"

One Network For All Games - The original DLSS required training the AI network for each new game. DLSS 2.0 trains using non-game-specific content, delivering a generalised network that works across games. This means faster game integrations, and ultimately more DLSS games

"

 

..suggests maybe not such a big effort for DCS VR.

 

The point of my reply was to explain that while a game that hasn’t been specifically optimized for ReBar still can see a some performance increase depending on how the engine works for DLSS has to be implemented on game engine

[Headwarp]

DLSS 2.0 is amazing and any developer whose game has performance issues by other means of anti-aliasing should do themselves and their community a favor and implement it.   We've been suffering with MSAA performance since they switched to deferred rendering 3 years ago.  DLSS would be an awesome solution for nvidia users.   Ignoring it and not implementing it is a bad idea, especially when so many of us are always looking for performance increases.  

 

I don't care if AMD users can't utilize it.  That's on them for buying an AMD card.  The end user made the choice not to have that advantage.   I'm pretty sure the majority of us are rocking an nvidia GPU. 

 

Here's the truth regarding hardware.   It's only as good as the developers that will utilize it.   When game devs opt not to support new cool toys those toys die out.  Bottom line though, this technology would allow for better performance and a clearer image than we currently have.  When this forum has the amount of threads it does complaining about performance, jaggies, and shimmering not jumping on that is lazy and kind of dumb imo.  Frankly the benefits of deferred rendering are kind of defeated when it leaves you with a washed out shimmering image.. which may improve with higher resolutions, but still.. not quite as good as msaa 2x, which introduces a significant performance impact. .  DLSS was invented as a solution for that.   It was even advertised as a solution to anti-aliasing in titles that utilize deferred rendering. 

 

It certainly would bring those of us using nvidia 20 and above GPU's more performance and increased image quality.    I can't say I agree with holding AMD's hand while they catch up.  If DLSS were to cause DCS users choose nvidia cards over AMD cards, that's on AMD, not Eagle Dynamics.   As it stands the whole "we don't support proprietary tech" excuse when it comes to a game changer like DLSS is re re talk.  It just means we all have to deal with the same crappy performance rather than some of us getting a massive fps increase and improvement to image quality.   

 

DLSS would cause a 2070's value to jump way up for DCS.   As it stands.. people are buying the most expensive offerings from both AMD and Nvidia to try and keep up with DCS VR, and that's the joke.   Not supporting this technology benefits no one other than saving developers the amount of work it would take to implement.  let's not pretend the decision to ignore such a technology is saving a single customer any money, where its implementation could make a lower priced 3070 offer better performance than the 6800XT, IF anybody could even get their freaking hands on one right now. 

Edited May 20, 2021 by Headwarp

[SaimOn67]

I bought nvidia not only for there hardware, but also for there software (in this case, DLSS!).

I didn't knew about EDs aproach to not support proprietary solutions.

As already stated above, to not support such an groundbraking technology while needing every frame in VR is a joke and I can't remember seeing this before.

I'm gaming/siming since 1995, so let's remember the first 3dfx voodoo and sound blaster cards, mmx or dual cores etc.

Where would we stand today if nobody would have taken advantage of new technologies?

[C3PO]

Here’s Steam’s hardware survey for April. Not necessarily representative of DCS user-base, but relevant I think.

 

https://store.steampowered.com/hwsurvey/videocard/?sort=pct

 

My two-cents: from what I have seen over the past 30 years is that proprietary technology delivers change for the better across everywhere … so good to get in there.

Edited May 20, 2021 by C3PO

[RealDCSpilot]

AI driven image enhancement with raytracing on top is definitely the next gen step like pixel shaders was over a decade ago:

 

Edited May 21, 2021 by RealDCSpilot

[GunSlingerAUS]

Given the way DLSS basically upsamples a low-res image, I don't think it's compatible with games that have a long draw distance, such as DCS, Read Dead Redemption, etc. It's very obvious even in titles like Cyberpunk 2077, which has a relatively middling draw distance, that far-off details are for the worse overall. 

 

I'd love to be proven wrong, as I have an RTX 3090 that gets a huge performance boost with DLSS in compatible titles, but sadly I don't think DCS is going to get such an easy performance increase for minimal coding effort.

[mimikiwi]

About the long draw distance drawback, we might get a good idea when No Man's Sky releases its DLSS VR update. That surely has some long draw distances. I for one am a big fan of DLSS and hope we do get it at some time in the future.

[Lurker]

At this point Eagle Dynamics implementing DLSS is probably very close to wishful thinking. Don't get me wrong, I would love to see it implemented, but ED are very slow to implement even new graphical APIs let alone proprietary technology. I wouldn't get my hopes up. 

[SharpeXB]

DLSS is a really over-hyped feature. Look at what Nvidia themselves say:

“DLSS 2.0 offers image quality comparable to native resolution while rendering only one quarter to one half of the pixels”

Note comparable does not mean equal. It would be impossible to produce equal image quality from 1/4 the number of pixels. That just can’t be done. To claim this is like claiming you could upscale the image from a DVD at 480p quality and get an output equal to 1080p.

There’s “better” upscaling and “acceptable” upscaling but it’s not possible to get equal upscaling. (They didn’t say equal…)

And in VR or on a very large 4K display you’ll be more able to perceive the difference in quality because especially in VR you’ve got your eyes right up close to the pixels. I’m sure many reviewers are looking at smallish 4K monitors where the difference might not be appreciable. 
 

Also DLSS isn’t really super sampling, it’s upscaling so they should have called it DLUS. But that doesn’t have such a nice ring to it huh?

 

What would be odd about using DLSS in VR is most people have super sampling or pixel density set to increase and then downsample the image. So you’d be artificially upscaling and then downsampling the image. Why bother? Just run native it would probably end up looking better. 
 

DLSS wouldn’t help in CPU-limited scenarios which is what you mostly deal with in DCS anyways. 

Edited May 21, 2021 by SharpeXB

[C3PO]

2 hours ago, SharpeXB said:

DLSS is a really over-hyped feature. Look at what Nvidia themselves say:

“DLSS 2.0 offers image quality comparable to native resolution while rendering only one quarter to one half of the pixels”

Note comparable does not mean equal. It would be impossible to produce equal image quality from 1/4 the number of pixels. That just can’t be done. To claim this is like claiming you could upscale the image from a DVD at 480p quality and get an output equal to 1080p.

There’s “better” upscaling and “acceptable” upscaling but it’s not possible to get equal upscaling. (They didn’t say equal…)

And in VR or on a very large 4K display you’ll be more able to perceive the difference in quality because especially in VR you’ve got your eyes right up close to the pixels. I’m sure many reviewers are looking at smallish 4K monitors where the difference might not be appreciable. 
 

Also DLSS isn’t really super sampling, it’s upscaling so they should have called it DLUS. But that doesn’t have such a nice ring to it huh?

 

What would be odd about using DLSS in VR is most people have super sampling or pixel density set to increase and then downsample the image. So you’d be artificially upscaling and then downsampling the image. Why bother? Just run native it would probably end up looking better. 
 

DLSS wouldn’t help in CPU-limited scenarios which is what you mostly deal with in DCS anyways. 

 

 

 

 

 

 

 

 

Edited May 21, 2021 by C3PO

[SharpeXB]

24 minutes ago, C3PO said:

 

 

 

 

 

 

 

 

Screenshots are never the same as looking at the real thing. What’s the point? And those are from Nvidia promo material?

Edited May 21, 2021 by SharpeXB

[speed-of-heat]

probably worthwhile you updating your world view on DLSS Cyberpunk 2077 DLSS Quality Comparison vs. Native, Benchmarks, & Blind Test - YouTube

and its associated performance testing

 

[SharpeXB]

25 minutes ago, speed-of-heat said:

probably worthwhile you updating your world view on DLSS Cyberpunk 2077 DLSS Quality Comparison vs. Native, Benchmarks, & Blind Test - YouTube

and its associated performance testing

 

TL:DW. That guy mumbles too much… whatever. And YouTube captures don’t look like the real thing. No way anything makes 1080p look like 4K via upscaling. It could be worth some other benefit like being able to run higher settings which as a trade off for higher res, sure that’s fair. But it won’t be truly equal. If DLSS is such magic why does Nvidia make $2,000 3090s? Just flip on DLSS and poof! Magic performance!
If I ever end up with a game that supports it I’ll see for myself. DLSS 1.0 was horrible. 

[speed-of-heat]

no further argument is needed .. <sarcasm> you are of course all knowing </sarcasm>... good luck with HDR... 

Edited May 22, 2021 by speed-of-heat

[SharpeXB]

11 minutes ago, speed-of-heat said:

no further argument is needed .. <sarcasm> you are of course all knowing </sarcasm>... good luck with HDR... 

 

I’d rather see them do HDR. DCS is losing the gorgeous graphics crown to that other big flight sim. Although we got pretty clouds now. 

[C3PO]

17 minutes ago, SharpeXB said:

I’d rather see them do HDR. DCS is losing the gorgeous graphics crown to that other big flight sim. Although we got pretty clouds now. 

DCS looks gorgeous to me. Maybe a hardware upgrade to get the quality up? What ED achieve on one core - pretty much - is incredible. Multi core will arrive later this year, freeing up FPS for quality bumps. It sure feels odd to suggest DCS isn’t one of the amazing looking sims out there. On my previous rig had it running well at 4K in pancake mode.

Edited May 22, 2021 by C3PO

[Oceandar]

I agree DCS should support DLSS but thats because I have NVIDIA card lol. Why dont one of you guys make a poll. It could probably get ED attention.