Nvidia Streamline (open source upscaling from Nvidia)

[CuriousTomCat]

Has Nvidia just become the good guy in the whole DLSS vs FSR vs XeSS debate?

Nvidia has introduced an open-source framework to make it easier for developers to plug anyone's upscaling algorithm into the rendering pipelines of their games. Intel's on board, and AMD's FSR 2.0 looks like a prime candidate, too. None of which seems like something I would have expected to be writing today. Or ever really.

https://www.pcgamer.com/has-nvidia-just-become-the-good-guy-in-the-whole-dlss-vs-fsr-vs-xess-debate/

https://wccftech.com/nvidia-announc...ution-technologies-updated-rtx-sdks-and-more/

https://www.tomshardware.com/uk/new...fy-developer-support-for-upscaling-algorithms

https://developer.nvidia.com/rtx/streamline

 

[Grim5]

It's helps everyone but it's also good for Nvidia because it will get DLSS into more games

 

[Wrinkly]

Nvidia, Intel & Hardware Vendor #3

As I said in another thread, we have needed this for some time already, but not sure Nvidia is the right player to control this. Nor would I like to see Microsoft take control due to Vulkan. Perhaps the Khronos Group could step in here.

 

[Rroff]

Seems about controlling the narrative to me as you can bet the menu option/feature mention will be nVidia Streamline regardless of what options are in the sub-menu and/or used via it.

 

[shankly1985]

How I see it is Nvidia has no choice! They either try or see DLSS die out. If Intel and FSR 2.0 can be just as good or better then DLSS is dead due to the hardware requirement.

This is just like history repeating itself with Nvidia Physx went open before that died out a last ditch effort from Nvidia.

Gsync for the most part has also gone the same way! Does the monitor still ship with the module?

 

[Nexus18]

Seems about controlling the narrative to me as you can bet the menu option/feature mention will be nVidia Streamline regardless of what options are in the sub-menu and/or used via it.

This.

I also think it is partly down to them not having to worry about FSR 2.0 or TSR showing up DLSS as from DF recent video on ghostwire, looks like dlss is still better on the whole (they don't rate TAAU at all it seems... so hopefully FSR is more akin to TSR) including for motion. Unless amd has added some of their own tweaks on top.....

How I see it is Nvidia has no choice! They either try or see DLSS die out. If Intel and FSR 2.0 can be just as good or better then DLSS is dead due to the hardware requirement.

This is just like history repeating itself with Nvidia Physx went open before that died out a last ditch effort from Nvidia.

Gsync for the most part has also gone the same way! Does the monitor still ship with the module?

I don't think nvidia are too worried about that as DLSS is in a lot of games, which aren't nvidia sponsored and also see above video on dlss vs TSR (with references to TAAU). Since getting dlss integrated into the engines directly, it is literally a couple of clicks for developers to incorporate. Where as dlss is not allowed in any AMD sponsored games it seems.

Gsync was a completely different scenario, it was a physical item, which cost monitor manufacturers money and had to be incorporated into the monitor housing, which required monitor manufacturers to take this into consideration for their design/manufacture process. DLSS is free and easy + quick to add and like FSR 2.0, can be added with ease if the game already has some kind of TAA method.

Physx was a bag of spanners (especially to work with according to developers comments) and I think nvidia knew that too hence all their efforts/changes with RTX in general.

As for this topic at hand, this is probably the best scenario for us but ideally it would be better if the industry just adopted one standard and all worked to improve it but obviously that'll never happen.

 

[Grim5]

This.

I also think it is partly down to them not having to worry about FSR 2.0 or TSR showing up DLSS as from DF recent video on ghostwire, looks like dlss is still better on the whole (they don't rate TAAU at all it seems... so hopefully FSR is more akin to TSR) including for motion. Unless amd has added some of their own tweaks on top.....

I don't think nvidia are too worried about that as DLSS is in a lot of games, which aren't nvidia sponsored and also see above video on dlss vs TSR (with references to TAAU). Since getting dlss integrated into the engines directly, it is literally a couple of clicks for developers to incorporate. Where as dlss is not allowed in any AMD sponsored games it seems.

Gsync was a completely different scenario, it was a physical item, which cost monitor manufacturers money and had to be incorporated into the monitor housing, which required monitor manufacturers to take this into consideration for their design/manufacture process. DLSS is free and easy + quick to add and like FSR 2.0, can be added with ease if the game already has some kind of TAA method.

Physx was a bag of spanners (especially to work with according to developers comments) and I think nvidia knew that too hence all their efforts/changes with RTX in general.

As for this topic at hand, this is probably the best scenario for us but ideally it would be better if the industry just adopted one standard and all worked to improve it but obviously that'll never happen.

A single standard means no innovation

it would also be easier if there was only one GPU manufacturer and only one GPU model - just cut down to one for everything it's makes development so easy but completely destroys innovation

 

[shankly1985]

This.

I also think it is partly down to them not having to worry about FSR 2.0 or TSR showing up DLSS as from DF recent video on ghostwire, looks like dlss is still better on the whole (they don't rate TAAU at all it seems... so hopefully FSR is more akin to TSR) including for motion. Unless amd has added some of their own tweaks on top.....

I don't think nvidia are too worried about that as DLSS is in a lot of games, which aren't nvidia sponsored and also see above video on dlss vs TSR (with references to TAAU). Since getting dlss integrated into the engines directly, it is literally a couple of clicks for developers to incorporate. Where as dlss is not allowed in any AMD sponsored games it seems.

Gsync was a completely different scenario, it was a physical item, which cost monitor manufacturers money and had to be incorporated into the monitor housing, which required monitor manufacturers to take this into consideration for their design/manufacture process. DLSS is free and easy + quick to add and like FSR 2.0, can be added with ease if the game already has some kind of TAA method.

Physx was a bag of spanners (especially to work with according to developers comments) and I think nvidia knew that too hence all their efforts/changes with RTX in general.

As for this topic at hand, this is probably the best scenario for us but ideally it would be better if the industry just adopted one standard and all worked to improve it but obviously that'll never happen.

Sure its doing well for the most part but it could be better! Look at how meny FSR titles are already vs DLSS. Keep in mind DLSS had like a year ahead start so on Nvidia I see 161 games vs FSR 81 in less than a year June 2021 I believe.

The reason is broad adoption, ease to get the source code, and ease to add to the games for a bigger player base.

End of the day this is all guessing so time will after tell.

 

[Nexus18]

A single standard means no innovation

it would also be easier if there was only one GPU manufacturer and only one GPU model - just cut down to one for everything it's makes development so easy but completely destroys innovation

True that.

Sure its doing well for the most part but it could be better! Look at how meny FSR titles are already vs DLSS. Keep in mind DLSS had like a year ahead start so on Nvidia I see 161 games vs FSR 81 in less than a year June 2021 I believe.

The reason is broad adoption, ease to get the source code, and ease to add to the games for a bigger player base.

End of the day this is all guessing so time will after tell.

DLSS was pretty **** for the first 1-2 years, both in terms of how it looked and how much of a pita it was to incorporate, I would say it has only really gained proper traction in the last year and a bit. No doubt FSR has had a good adoption rate though but sadly not for the games where it is really needed imo, although that has changed recently with them getting into some of those games i.e. cp 2077

 

[tommybhoy]

Keep in mind DLSS had like a year ahead start so on Nvidia I see 161 games vs FSR 81 in less than a year June 2021 I believe.

DLSS(Feb 2019) has had a 2 yr+ start over FSR(June 2021)

161 DLSS >3 yr
V
81 FSR<1 yr

Unsurprisingly-the adoption rate is very
impresive in comparison to DLSS due to having a very limited user base with a so far high buy in cost.

Imo expecting both of them to disolve over time when FSR gets standardised and gets integrated into a future DX12/13 update.

I can see why there's been a lot of comparison to G-Sync(introducing tech but slowly dying out) but Nvidia will no doubt replace DLSS with some other must have if you want to pay for it feature.

 

[Nexus18]

DLSS(Feb 2019) has had a 2 yr+ start over FSR(June 2021)

161 DLSS >3 yr
V
81 FSR<1 yr

Unsurprisingly-the adoption rate is very
impresive in comparison to DLSS due to having a very limited user base with a so far high buy in cost.

Imo expecting both of them to disolve over time when FSR gets standardised and gets integrated into a future DX12/13 update.

I can see why there's been a lot of comparison to G-Sync(introducing tech but slowly dying out) but Nvidia will no doubt replace DLSS with some other must have if you want to pay for it feature.

Not sure nvidia users will be losing much sleep over having a feature set for the past 1-2 years, which has proved to been of extreme value especially in big triple a games where the extra perf. has been much needed for RT visuals or/and 4k60 gaming especially when they will have paid the same or less compared to competitors equivalent gpus.... And even if people did overpay, I think a lot can see the worth in paying extra (within reason....) for that feature, especially when it has been shown that FSR 1 has not come anywhere close to dlss >2.0 versions and we're still waiting on FSR 2.0 to be usable. I've tried FSR 1 in several games now and have found it incredibly underwhelming for anything but 4k and the ultra quality preset and even then, it still hasn't impressed me.

I can't see FSR becoming the "standard" either, TAAU or/and TSR, yes though (which early insights suggest FSR 2.0 is based on TAAU and from DF brief insight as per the above video, TAAU does not seem to be as good as dlss or TSR potentially unless AMD have added their own tuning to it.....). Same way freesync is not an "amd" thing but it is based on the industry standard called adaptive sync (as per what nvidia does with their gsync "compatible" brand name to utilise "adaptive sync") and same way their "SAM" feature is resizable bar.

AMD do great in the end with their solutions but they have 2 problems:

- the time it takes for them to catch up, maybe people are happy to wait 1-2/3 years before starting a game? Or happy to run lesser settings and or/ resolution or/and less FPS?
- they have an over the fence approach, as in, once done, they don't push to have their features in games and basically leave it up to the developers to do as they please: see tressfx, true audio, the flames/fire effects in alien isolation

 

[Wrinkly]

The problem is that FSR is based on 30+ year old ideas. Nothing wrong with that if you are on a budget as that is ideal for low end, consoles and mobile. Both Intel and Nvidia realise the benefits of AI based image reconstruction, something that we weren't able to do 30+ years ago, or for that matter even today at speed. DLSS and XeSS do however pave the way for true image reconstruction in the future, something that FSR in it's current state, 2.0, just can't do.

 

[Grim5]

^^ That's incredibly impressive

I wonder if this could actually simplify game development if the scene and assets that the AI creates can be imported into a game engine - cause let's say I'm working on making a new skateboarding game; I can just use my phone to recorder doing some tricks myself and use AI to instantly turn it into a 3D world with assets and animations

 

[Chuk_Chuk]

^^ That's incredibly impressive

I wonder if this could actually simplify game development if the scene and assets that the AI creates can be imported into a game engine - cause let's say I'm working on making a new skateboarding game; I can just use my phone to recorder doing some tricks myself and use AI to instantly turn it into a 3D world with assets and animations

It's application will probably be quite niche and maybe only useful for indie devs. It could be useful for look dev but going back to the current methods for the final product.
The large devs already use 3d scanning and motion capture. so the trade off will be quality vs time spent.
The unknown is if the AI will miss the smaller imperfections that are present in the real world as that stuff adds realism.