Rage3D Discussion Area

Rage3D Discussion Area (http://www.rage3d.com/board/index.php)
-   Other Graphics Cards and 3D Technologies (http://www.rage3d.com/board/forumdisplay.php?f=65)
-   -   Nvidia RTX DLSS/Ray Tracing Discussion (http://www.rage3d.com/board/showthread.php?t=34049038)

Ziklitschli Apr 5, 2020 03:05 PM

The screenshot challenge is nonsense. Most upscaling techniques can hold much better on stills than in motion and DLSS 2.0 is good enough to hold for sure. The screenshots also seem low on transparencies and fine detail (haven't seen all though just this page), even FXAA holds on fat geometry when still.

Not to mention, @Exposed, that it's hard to take your championing of DLSS 2.0 seriously when even with pre 2.0 version, you claimed that it's so good, 90% (or even 90 sth) of native res etc etc. Now, in the Digital Foundry Control DLSS 2.0 video, they show the older DLSS breaking left and right and how the new one is better 1080p to 4k than the old one 2560 to 4k.

Anyway, I have nothing against it existing, just the way it's promoted as a feature. Just like that crap hiding details on edge of the screen etc because you don't look there that much. Soon console games will look like a light LSD trip, morphing and switching and going in and out of focus (some already do), it's all "lower your IQ" options and not features worth to put on a box. NVIDIA 1280x720, NEW FEATURE TO KEEP YOUR 60! DLSS is good now for an upscaling tech but you need native res for a truely pristine image and that's not going to change.

Exposed Apr 5, 2020 04:10 PM

Quote:

Originally Posted by Ziklitschli (Post 1338186059)
The screenshot challenge is nonsense. Most upscaling techniques can hold much better on stills than in motion and DLSS 2.0 is good enough to hold for sure. The screenshots also seem low on transparencies and fine detail (haven't seen all though just this page), even FXAA holds on fat geometry when still.

You know what else is nonsense? Coming in here making blanket claims without even taking the time to actually do some research. Because if you had, you'd see there were not one, but three different sources with video showing just how indistinguishable DLSS 2.0 is from native 4k. Not to mention, using zoom magnifications from 200 to 800%. So you didn't have to use my screenshots, the data is out there for itself.

Sorry, but my screenshots are 100% untouched in full 3840x2160. You seem to be making excuses without even looking at them, or looking at them and frustrated you can't tell the difference because you are one of those people that look for reasons to **** on a specific brand feature? It would be very clear which screenshot would be FXAA, in fact I'll include them in my next set of screenshots to show just how foolish your statement was.

Quote:

Not to mention, @Exposed, that it's hard to take your championing of DLSS 2.0 seriously when even with pre 2.0 version, you claimed that it's so good, 90% (or even 90 sth) of native res etc etc. Now, in the Digital Foundry Control DLSS 2.0 video, they show the older DLSS breaking left and right and how the new one is better 1080p to 4k than the old one 2560 to 4k.
I've always said DLSS 1.0 was 90% as good as native 4k in Metro Exodus. This was after their updated DLSS patch. I've never said the same about any other game. Now, you wanna go back and reread those posts (and those screenshots)?

Also, the new Digital Foundry video just proves my point even further (and yours even less). They basically state with DLSS 2.0 Performance mode (1080p upscaled to 4k), there's only minor differences from Native 4k. With 1440p as the base reconstructor, it's damn near indistinguishable, and they have video and screenshots that shows this.

So thanks for bringing that new video up proving my point further. Heck you don;'t even have to take theirs, or my word for it, just take a look at the comments in the PC gaming section of this forum on Control. Others saying it looks as good and even better than native, surely they must have eye issues?

Quote:

Anyway, I have nothing against it existing, just the way it's promoted as a feature. Just like that crap hiding details on edge of the screen etc because you don't look there that much. Soon console games will look like a light LSD trip, morphing and switching and going in and out of focus (some already do), it's all "lower your IQ" options and not features worth to put on a box. NVIDIA 1280x720, NEW FEATURE TO KEEP YOUR 60! DLSS is good now for an upscaling tech but you need native res for a truely pristine image and that's not going to change.
Of course nothing would truly beat native resolution. But what happens when native 4k gives **** performance? You would rather play at 4k 25-30 fps than DLSS 2.0 which would give indistinguishable (and sometimes better) quality at 60 fps?

bill dennison Apr 5, 2020 04:32 PM

:lol:

DLSS 1.0 was 70% complete and utter garbage at 4k .

it made RT faster and that was it's only use, IQ was crap



i have looked at DSLL 2.0 A bit so far and it is better at 4k but it is not like 100% like 4k .


I will only use DSLL if I can not get a game to run any other way

...........


with the amount of games with RT and the problems with DLSS

my 2080 ti strix is a great 4k card, but nvidia's whole RTX is a fail at 18 months .

Exposed Apr 5, 2020 05:28 PM

Which is why you played Metro Exodus without ray tracing/DLSS at native 4k, right? :lol:

Despite what you say Bill, you still find reasons to give money to Nvidia somehow. :lol:


Digital Foundry also has their newest video on the recent DLSS 2.0 patch for Control, though it focuses more on the performance mode rather than quality mode.


Ziklitschli Apr 5, 2020 05:41 PM

Quote:

Originally Posted by Exposed (Post 1338186061)
You know what else is nonsense? Coming in here making blanket claims without even taking the time to actually do some research. Because if you had, you'd see there were not one, but three different sources with video showing just how indistinguishable DLSS 2.0 is from native 4k. Not to mention, using zoom magnifications from 200 to 800%. So you didn't have to use my screenshots, the data is out there for itself.

That claims were about the challenge itself.

Quote:

Originally Posted by Exposed (Post 1338186061)
Sorry, but my screenshots are 100% untouched in full 3840x2160. You seem to be making excuses without even looking at them, or looking at them and frustrated you can't tell the difference because you are one of those people that look for reasons to **** on a specific brand feature? It would be very clear which screenshot would be FXAA, in fact I'll include them in my next set of screenshots to show just how foolish your statement was.

Another sentence you didn't understand. Calm down, what I meant is that upscale artifacts are visible on some things more than the others and I saw a lot of geometry but little transparencies. Didn't say that your screens are touched at all, said that even fxaa would look good on those pics, read again.

On the other hand, I can bet that you didn't even turn off TAA or variants of it for the screenshots, when it is known to hide upscale artifacts.


Quote:

Originally Posted by Exposed (Post 1338186061)
I've always said DLSS 1.0 was 90% as good as native 4k in Metro Exodus. This was after their updated DLSS patch. I've never said the same about any other game. Now, you wanna go back and reread those posts (and those screenshots)?

Ok then and no, don't care. Still, a pre 2.0 DLSS so not that far off.


Quote:

Originally Posted by Exposed (Post 1338186061)
Also, the new Digital Foundry video just proves my point even further (and yours even less). They basically state with DLSS 2.0 Performance mode (1080p upscaled to 4k), there's only minor differences from Native 4k. With 1440p as the base reconstructor, it's damn near indistinguishable, and they have video and screenshots that shows this.

It's not indistinguishable. I respect Digital Foundry but they tend to go for positivity wherever possible, and are fans of dynamic resolution solutions and push those as a feature, where I consider it total crap. Not to mention, I already said DLSS is good for an upscale, not sure what you're on about.

Quote:

Originally Posted by Exposed (Post 1338186061)
thanks for bringing that new video up proving my point further. Heck you don;'t even have to take theirs, or my word for it, just take a look at the comments in the PC gaming section of this forum on Control. Others saying it looks as good and even better than native, surely they must have eye issues?

My post wasn't as much about DLSS as it was about your challenge.

I don't care for people claiming it looks better than native.

Quote:

Originally Posted by Exposed (Post 1338186061)
Of course nothing would truly beat native resolution. But what happens when native 4k gives **** performance? You would rather play at 4k 25-30 fps than DLSS 2.0 which would give indistinguishable (and sometimes better) quality at 60 fps?

I'd go for worse framerate to play in native res every time, obviously. Don't have any problem with you choosing otherwise, just with all your arrogant yet uninformed pushing of it.

Exposed Apr 5, 2020 05:59 PM

Quote:

Originally Posted by Ziklitschli (Post 1338186081)
That claims were about the challenge itself.

You said stills don't show upscaling artifacts. In other words, motion artifacts? So what sources do you have that discusses DLSS motion artifacts? Seems to be non existent in the three outside sources previously mentioned who have done video comparisons. Or are you simply drumming up artifact arguments without bothering to check?


Quote:


Another sentence you didn't unerstand. Calm down, what I meant it's that upscale artifacts are visible on some things more than the other and I saw a lot of geometry but little transparencies. Didn't say that your screens are touched at all, said that even fxaa would look good on those pics, read again.

Again, making assumptions without checking? Can you show me a single review from the last 15 years that shows FXAA looking close to it's original native screenshot? Last I checked FXAA is also a blur fest and DLSS 2.0 has shown it does not blur compared to the native image (unlike DLSS 1.0 which did) while retaining it's upscaling algorithm.

Quote:


On the other hand, I can bet that you didn't even turn off TAA or variants of it for the screenshots, when it is known to hide upscale artifacts.

A clear example here of not understanding the technology. You can't turn "on" TAA or any other AA method when DLSS is running. DLSS IS the anti-aliasing method. All of your criticisms so far are based on wrong assumptions.

Quote:


Ok then and no, don't care. Still, a pre 2.0 DLSS so not that far off.
You cared enough to bring it up.

What's funny is that every major tech outlet that has done an overview on DLSS 2.0 are all in agreement, so you can completely ignore everything I state. The dissidence seems to be coming from people who for some reason cannot accept or make a partial judgement for whatever reason (though it's clear why for some people).

You look at this thread and who has provided all the videos, sources, screenshots, etc...? What has the "sour grapes" provided? Blanket statements with nothing to back it up.


Quote:


It's not indistinguishable. I respect Digital Foundry but they tend to go for positivity wherever possible, and are fans of dynamic resolution solutions and push those as a feature, where I consider it total crap. Not to mention, I already said DLSS is good for an upscale, not sure what you're on about.


You were the one that brought them up to show the differences between the "old" DLSS and the "new". You cannot accept their stance on one and disregard the other, the other being DLSS 2.0 quality is nearly indistinguishable and sometimes better than native (which let me say again, was also the finding of other tech sites like Hardware Unboxed). Also, they're fans of dynamic resolutions on consoles. On PC they're really in depth in their overviews and get into technical details most other sites don't. Of course, that means if you use them as a source, better be prepared to accept their findings as a whole, not pick and choose.




Quote:



My post wasn't as much about DLSS as it was about your challenge.

I don't care for people claiming it looks better than native.
In other words, you have a problem accepting that as a reality(or even just a possibility) despite that being the findings from other tech sites.



So you tell me, which one is native and which one is DLSS? Shouldn't be that hard, it's a pixel to pixel comparison. Where are these upscaling artifacts? Please point them out.,





Quote:



I'd go for worse framerate to play in native res every time, obviously. Don't have any problem with you choosing otherwise, just with all your arrogant yet uninformed pushing of it.
Ah, so this a "I hate that you're right" kind of debate? Is that why you can offer nothing of substance? Can you at least provide some form of counter evidence to what has been posted? Artifacts, etc..?



If you own a Turing then you are free to choose to play in whatever settings you want, with or without DLSS/RTX etc...

SIrPauly Apr 5, 2020 06:03 PM

If I had any say, I would be promoting Dlss 2.0 strongly. To receive that kind of performance boost for 4k, with impressive fidelity is almost magical and too good to be true. Great feature for flexibility.

SIrPauly Apr 5, 2020 06:17 PM

It also helps raytracing and brings much more value to the Rtx family even the Rtx 2060.

Exposed Apr 5, 2020 06:24 PM

Quote:

Originally Posted by SIrPauly (Post 1338186085)
If I had any say, I would be promoting Dlss 2.0 strongly. To receive that kind of performance boost for 4k, with impressive fidelity is almost magical and too good to be true. Great feature for flexibility.

https://www.eurogamer.net/articles/d...-zero-analysis

"There are a couple of interesting effects delivered by DLSS. On inner surfaces, textures appear to deliver more detail - sometimes even more than native resolution (and to be fair, sometimes less, though you need extreme magnification side-by-side shots to tell). It's not just a factor of contrast adjustment or sharpening either. On the rock wall ahead of Jesse in the very first playable scene in Control, single pixel detail on reflective elements of the rock wall shine with DLSS when they don't with native rendering. Remember that DLSS is a replacement for the temporal anti-aliasing found in many games - and TAA does tend to add some level of blur that DLSS does not."



It's not magic, it's simply evolution of graphics. Why would anyone hate on this is beyond belief. If you get a massive performance boost while retaining quality, why wouldn't that be a good option to have? Especially when 4k performance is suffering and not everyone can afford a 2080ti? If anything this video shows that RTX 2060 owners can get near 4k image quality with DLSS Performance mode without having to shell out bucks for a more expensive card.

bill dennison Apr 5, 2020 07:37 PM

Quote:

Originally Posted by Exposed (Post 1338186078)
Which is why you played Metro Exodus without ray tracing/DLSS at native 4k, right? :lol:

Despite what you say Bill, you still find reasons to give money to Nvidia somehow. :lol:


Digital Foundry also has their newest video on the recent DLSS 2.0 patch for Control, though it focuses more on the performance mode rather than quality mode.


after playing with it i did turn it off on Metro Exodus when it came out and i played

looking at for it keep getting me killed :bleh:

SIrPauly Apr 5, 2020 08:00 PM

Quote:

Originally Posted by Exposed (Post 1338186088)
https://www.eurogamer.net/articles/d...-zero-analysis

"There are a couple of interesting effects delivered by DLSS. On inner surfaces, textures appear to deliver more detail - sometimes even more than native resolution (and to be fair, sometimes less, though you need extreme magnification side-by-side shots to tell). It's not just a factor of contrast adjustment or sharpening either. On the rock wall ahead of Jesse in the very first playable scene in Control, single pixel detail on reflective elements of the rock wall shine with DLSS when they don't with native rendering. Remember that DLSS is a replacement for the temporal anti-aliasing found in many games - and TAA does tend to add some level of blur that DLSS does not."



It's not magic, it's simply evolution of graphics. Why would anyone hate on this is beyond belief. If you get a massive performance boost while retaining quality, why wouldn't that be a good option to have? Especially when 4k performance is suffering and not everyone can afford a 2080ti? If anything this video shows that RTX 2060 owners can get near 4k image quality with DLSS Performance mode without having to shell out bucks for a more expensive card.

I think Nvidia deserved constructive negative views on earlier iterations of Dlss because there were limitations and artifacts and didn't match their original marketing. Now, it's potentially a game changer if they can get the feature in more content.

jimjobob Apr 5, 2020 08:03 PM

Quote:

Originally Posted by Exposed (Post 1338186078)
Which is why you played Metro Exodus without ray tracing/DLSS at native 4k, right? :lol:

I played Metro w/o. I thought it looked and performed about the same as 80% resolution scale but it completely broke HDR. When it first released, HDR with DLSS gave odd color and later they just made it so you got the same image with and w/o HDR set to on if you had DLSS enabled.

The only game I have played with DLSS 2.0 is youngblood and I wound up leaving it on before I even knew about 2.0 because I thought it somehow looked better than native.

Pre 2.0 it really was not worth using in any game.

SIrPauly Apr 5, 2020 08:20 PM

And i didn't think your challenge was nonsense, but welcomed, Exposed. To judge fidelity, to me, is all about dynamics and movement primarily, but static images help as well.

Ziklitschli Apr 5, 2020 08:26 PM

Quote:

Originally Posted by Exposed
You said stills don't show upscaling artifacts. In other words, motion artifacts? So what sources do you have that discusses DLSS motion artifacts? Seems to be non existent in the three outside sources previously mentioned who have done video comparisons. Or are you simply drumming up artifact arguments without bothering to check?

I said stills can hold much better. Not only you play catching words, you're also bad at it.

Quote:

Originally Posted by Exposed
Again, making assumptions without checking? Can you show me a single review from the last 15 years that shows FXAA looking close to it's original native screenshot? Last I checked FXAA is also a blur fest and DLSS 2.0 has shown it does not blur compared to the native image (unlike DLSS 1.0 which did) while retaining it's upscaling algorithm.

I don't even know what you're answering here.

Quote:

Originally Posted by Exposed
A clear example here of not understanding the technology. You can't turn "on" TAA or any other AA method when DLSS is running. DLSS IS the anti-aliasing method. All of your criticisms so far are based on wrong assumptions.

I just didn't know you can't have both on because I don't use it. Is that so btw? Sound stupid, you can mix everything else.

Quote:

Originally Posted by Exposed
You cared enough to bring it up.

What's funny is that every major tech outlet that has done an overview on DLSS 2.0 are all in agreement, so you can completely ignore everything I state. The dissidence seems to be coming from people who for some reason cannot accept or make a partial judgement for whatever reason (though it's clear why for some people).

You look at this thread and who has provided all the videos, sources, screenshots, etc...? What has the "sour grapes" provided? Blanket statements with nothing to back it up.

Just noone cares as much as you.

My dissidence comes from logic. You'll see how great 3.0 will be, fixing all the flaws of 2.0.

And yes I don't care. Just posted what I think about IQ challenges with forum screenshots. That with you every DLSS is unrecognisable from native in real life is just an impression I had of your opinion, posted it, you explained, much shorter and ridiculous to expect me to actually research that.

Quote:

Originally Posted by Exposed
You were the one that brought them up to show the differences between the "old" DLSS and the "new". You cannot accept their stance on one and disregard the other

Really? How old are you?

Quote:

Originally Posted by Exposed
the other being DLSS 2.0 quality is nearly indistinguishable and sometimes better than native (which let me say again, was also the finding of other tech sites like Hardware Unboxed). Also, they're fans of dynamic resolutions on consoles. On PC they're really in depth in their overviews and get into technical details most other sites don't. Of course, that means if you use them as a source, better be prepared to accept their findings as a whole, not pick and choose.

No they were blabbering about how PC needs dynamic resolution in multiple videos. I like them and their technical analysis but disagree on some points ( imagine that).


Quote:

Originally Posted by Exposed
In other words, you have a problem accepting that as a reality(or even just a possibility) despite that being the findings from other tech sites.

Ok, you're right, upscaling from 1080p is better than 4k. Ridiculous, isn't it.

The fun fact is that if you read that article you quote, it's full of mentions of artifacts "reduced to a minimum", leftover halo effects sth and some things that supposedly better than 4k. You seem to read it extremly selectively.

Not to mention, there already are problems visible in their analysis, they just don't focus on them, probably happy with the improvement etc and showcasing their generaly positive attitude (which is ok btw).

Quote:

Originally Posted by Exposed
So you tell me, which one is native and which one is DLSS? Shouldn't be that hard, it's a pixel to pixel comparison. Where are these upscaling artifacts? Please point them out.,

...

Ah, so this a "I hate that you're right" kind of debate? Is that why you can offer nothing of substance? Can you at least provide some form of counter evidence to what has been posted? Artifacts, etc..?

No I don't care nearly enough about DLSS to waste time on proving anything, wasted enough on this ping pong quote match already. Why would I partake in a challenge I consider worthless, would kind of defeat the point of posting about it in 1st place, wouldn't it.

Btw one reason to hate it is the day when it's just implemented without choice, because "people can't see a difference", "just as good as native etc.

Trunks0 Apr 5, 2020 09:00 PM

It's a nice to have really, although its incredibly limited as is. If it could work pan across everything, then it would be a must have and exciting feature. As is it will likely fade away as RT acceleration gets faster.

DLSS 2.0 sure does look pretty great though. :up: to nVidia providing a solution to make RT more usable right now.

demo Apr 5, 2020 09:23 PM

I have to agree with Z-man in that it doesn't handle motion well (shimmering). Also, as I've complained about all along is a halo/sharpening artifact around everything that I can't stand. Digital Foundry discussed both of these negatives in their latest vid.

Exposed Apr 5, 2020 09:31 PM

Quote:

Originally Posted by Ziklitschli (Post 1338186109)
I said stills can hold much better. Not only you play catching words, you're also bad at it.

Enough with the frivolous 1 line empty statements. Show me where in the screenshots where you see upscaling artifacts. Or reference any video where DLSS motion artifacting is seen.

Quote:

I don't even know what you're answering here.
Enough with the frivolous 1 line empty statements. Show me where in the screenshots where you see upscaling artifacts. Or reference any video where DLSS motion artifacting is seen.

Quote:

I just didn't know you can't have both on because I don't use it. Is that so btw? Sound stupid, you can mix everything else.

You jumped to an invalid conclusion based on invalid logic. That sort of defines your entire argument here. You're essentially downplaying DLSS 2.0 without even understanding what it is, and refuse to acknowledge it's actually worked as originally intended.


Quote:


Just noone cares as much as you.
Enough with the frivolous 1 line empty statements. Show me where in the screenshots where you see upscaling artifacts. Or reference any video where DLSS motion artifacting is seen.


Quote:

My dissidence comes from logic. You'll see how great 3.0 will be, fixing all the flaws of 2.0.
Your logic is lacking evidence here. Show me where in the screenshots where you see upscaling artifacts. Or reference any video where DLSS motion artifacting is seen.

Quote:

And yes I don't care. Just posted what I think about IQ challenges with forum screenshots. That with you every DLSS is unrecognisable from native in real life is just an impression I had of your opinion, posted it, you explained, much shorter and ridiculous to expect me to actually research that.
Then you must think every review screenshot from every game/GPU/setting since the dawn of PC gaming is also ridiculous. Which means your logic is flawed, and you can't seem to recognize it.

Again, show me where in the screenshots where you see upscaling artifacts. Or reference any video where DLSS motion artifacting is seen.



Quote:

Really? How old are you?

Enough with the frivolous 1 line empty statements. Show me where in the screenshots where you see upscaling artifacts. Or reference any video where DLSS motion artifacting is seen.

Quote:


No they were blabbering about how PC needs dynamic resolution in multiple videos. I like them and their technical analysis but disagree on some points ( imagine that).

Oh look, moving goal posts, sidestepping, and strawman all in one.

Show me where in the screenshots where you see upscaling artifacts. Or reference any video where DLSS motion artifacting is seen.

Quote:

Ok, you're right, upscaling from 1080p is better than 4k. Ridiculous, isn't it.
https://www.eurogamer.net/articles/d...-zero-analysis

"There are a couple of interesting effects delivered by DLSS. On inner surfaces, textures appear to deliver more detail - sometimes even more than native resolution (and to be fair, sometimes less, though you need extreme magnification side-by-side shots to tell). It's not just a factor of contrast adjustment or sharpening either. On the rock wall ahead of Jesse in the very first playable scene in Control, single pixel detail on reflective elements of the rock wall shine with DLSS when they don't with native rendering. Remember that DLSS is a replacement for the temporal anti-aliasing found in many games - and TAA does tend to add some level of blur that DLSS does not."



Quote:


The fun fact is that if you read that article you quote, it's full of mentions of artifacts "reduced to a minimum", leftover halo effects sth and some things that supposedly better than 4k. You seem to read it extremly selectively.

The fun fact is those artifacts are well known because of the fact they were using DLSS performance mode (1080p). They made the point that despite using 1080p as the base, DLSS provided very good matching results to 4k. those "artifacts" were non existent when using 1440p as the base. You seem to read it extremly selectively.


Quote:

Not to mention, there already are problems visible in their analysis, they just don't focus on them, probably happy with the improvement etc and showcasing their generaly positive attitude (which is ok btw).
They were positive because DLSS 2.0 gave RTX 2060 users a credible option to run 4k DLSS with 1080p upscaling. Nothing more nothing less. Notice how you are strangling the argument down to DLSS performance mode while disregarding DLSS quality mode.

Quote:


No I don't care nearly enough about DLSS to waste time on proving anything, wasted enough on this ping pong quote match already. Why would I partake in a challenge I consider worthless, would kind of defeat the point of posting about it in 1st place, wouldn't it.

You clearly care enough to respond to me line by line. Now again, show me where in the screenshots where you see upscaling artifacts. Or reference any video where DLSS motion artifacting is seen.

Shirley there is something to support your logic?

What's amusing is others here see the tech for what it is. Maybe you should go to the PC gaming section and tell those players running DLSS quality mode (not DLSS performance) and noticing it looks as good as native, even better that obviously they're wrong? While your at it maybe write a letter to Digital Foundry and Hardware Unboxed as well?

Quote:


Btw one reason to hate it is the day when it's just implemented without choice, because "people can't see a difference", "just as good as native etc.
Well this I agree with, I'd rather the option be given. Consoles already suffer through this though with dynamic resolution and checkerboard rendering, both of which will give a noticeable image degradation to maintain framerate.

Exposed Apr 5, 2020 09:34 PM

Quote:

Originally Posted by demo (Post 1338186121)
I have to agree with Z-man in that it doesn't handle motion well (shimmering). Also, as I've complained about all along is a halo/sharpening artifact around everything that I can't stand. Digital Foundry discussed both of these negatives in their latest vid.

There's zero shimmering in the two DLSS 2.0 games I've played (Youngblood and Deliver Us the Moon). Which DLSS 2.0 games are you referring to? Running quality or performance mode?

SIrPauly Apr 5, 2020 09:50 PM

Movement with strong lighting and color contrasts may offer a halo effect that may be destracting for gamers with very keen eyes,

Edit: Digital Foundry raised this issue and is asking for a slider to help reduce the artifacts, for flexibility.

SIrPauly Apr 6, 2020 12:44 AM

Another limitation, doesn't really work well to enhance denoisers for raytracing, where taa does. I actually raised this question.

Ziklitschli Apr 6, 2020 04:43 AM

Quote:

Originally Posted by Exposed (Post 1338186124)
Enough with the frivolous 1 line empty statements.

I can frivolous one liners as much as I want thankfully, freedom babe.

I have 2,5 remote jobs and 2 kids small enough that I have to teach them myself atm (wife is a teacher but is teaching her classes remotly) and then keep them busy till evening so kindly stop the crazy nerd routine.

I have this crazy fix for Path of Exile but managed 20 minutes of it the last 3 days. I have a ****ing mustache. I'm not into a quote match (that, admitely, I shouldn't have started in the first place) with someone like you who will not shut up unless run to the ground (which I'm not saying would happen, to be clear), which I never enjoy in the end anyway, find it hard to be not nice to people.

Quote:

Originally Posted by Exposed (Post 1338186124)
Show me where in the screenshots where you see upscaling artifacts. Or reference any video where DLSS motion artifacting is seen.

https://hothardware.com/reviews/inve...-5-and-control

"Because DLSS 2.0 applies itself to objects in motion as the scene moves, there's a little bit of temporal distortion (no, this is not Star Trek). That means that things which move and animate, like Jessie, give off a very faint halo. Try as we might, we couldn't capture it in screenshots in a way that was obvious because the halo is only apparent in motion."

It's funny how you repeat it ad nauseam yet everyone out there mentions sth and mother logic herself says artifacts will be there.It's not me having nothing, it's just mind bending that you require smb to show you that, like you didn't understand basic limitations in the world. Also if there's sth when scaling from 1080p, then it's probably going to be there upscaling from 2560, no?

Quote:

Originally Posted by Exposed (Post 1338186124)
Then you must think every review screenshot from every game/GPU/setting since the dawn of PC gaming is also ridiculous.

No, only those that don't work.

World and its tech sites were full of MLAA/ FXAA initial analysis and conclusions based on screenshots, HardOCP (afair) even claimed it's better than msaa for edges. Now we know how stupid that was.

Quote:

Originally Posted by Exposed (Post 1338186124)
https://www.eurogamer.net/articles/d...-zero-analysis

"There are a couple of interesting effects delivered by DLSS. On inner surfaces, textures appear to deliver more detail - sometimes even more than native resolution (and to be fair, sometimes less, though you need extreme magnification side-by-side shots to tell). It's not just a factor of contrast adjustment or sharpening either. On the rock wall ahead of Jesse in the very first playable scene in Control, single pixel detail on reflective elements of the rock wall shine with DLSS when they don't with native rendering. Remember that DLSS is a replacement for the temporal anti-aliasing found in many games - and TAA does tend to add some level of blur that DLSS does not."

This quote that you warrior so much with kind of defeats itself - they are comparing to native with TAA. TAA ruins IQ so much that it's no wonder that sharpened AI upscaling can look better at places.

I was never talking about native as in with TAA. They are talking native in that particular game and Alex actually says "with TAA" all the time.

Quote:

Originally Posted by Exposed (Post 1338186124)
You clearly care enough to respond to me line by line.

That was a mistake and hope it's the last time I'm making it.

Quote:

Originally Posted by Exposed (Post 1338186124)
While your at it maybe write a letter to Digital Foundry and Hardware Unboxed as well?

Listen, ignoring that cringey quote, I don't really know what we are arguing about. I already told you that DLSS 2.0 is good for an upscale, this is like the best you can get from me about an upscaler, it's actualy also a huge improvement. Screenshots not enough for reconstructing techniques, that's common knowledge and they can have a supportive role at best. I even agree with you that for RTX 2060 users it's a good thing (except me heh) and that a standard user will be happy with it. It's just not a sth that should ever be a flagship feature and that's subjective, it's only your weird demands of admitting its unachievable perfection that fuels this bs. It's even general technique now, not per game, how can it be perfect.

demo Apr 6, 2020 05:27 AM

Quote:

Originally Posted by Exposed (Post 1338186125)
There's zero shimmering in the two DLSS 2.0 games I've played (Youngblood and Deliver Us the Moon). Which DLSS 2.0 games are you referring to? Running quality or performance mode?

Deliver us the Moon is fortunate in the sense it has a lot of plain geometry. In games with intricate geometry and many alpha/transparent textures (IE- foliage, grass, hair, fences, wire mesh etc), DLSS often shimmers in motion. I also notice some odd/blurry looking jaggies occasionally, that look like missed edges from the lower, pre-processed, resolution. They look really odd, like running non native res on an LCD screen - which I guess is what it's doing.

The sharpening artifacts are also a big issue for me, but I'm convinced Nvidia just added a sharpening filter when everyone complained how blurry DLSS was originally, and I bet this can easily be toned down (Digital Foundry alluded to this).

Just note, I've known Z-Man for about a decade and much like Sirpauly (and myself to a lesser extent), he is absolutely anal about AA quality. I think those two gentlemen know more about AA modes than the rest of the forum combined. They are the types of guys to use exotic mixed AA modes (IE- SGSSAA+trMSAA+CSAA plus texture LOD adjustment) that can only be used via Inspector and require correct compatibility bits. I trust Z-mans judgement, he is seeing what I'm seeing.

I suppose what they (and I) notice is that DLSS is very good for shader based AA, and a welcomed feature, but like all shader AA its real goal is performance and not quality. To that end it does very well, but it does not look anywhere near as clean, as sharp, or as shimmer free as 'real AA'. It doesn't seem to filter textures or shaders cleanly like SSAA either, instead just sharpening them (again, convinced there is simply a sharpening filter added). I'd like to know what happened to MFAA. I liked the idea of real MSAA with temporal jitter, for 4xMSAA quality at 2x performance hit etc. That seemed promising.
Shader AA scales extremely well with resolution - or to be more precise, pixel pitch and PPI. It looks horrid at lower PPI and pixel pitch, but does the job just fine with extreme PPI and pixel pitch. I don't think we're at the point where real AA can be discarded yet, though, as even at 4K 55" I notice jaggies just as bad as a 27" 1440p monitor (similar PPI). You know, a 60" 4k display has the same PPI as 30" 1080p. Meaning both need some serious filtering to eliminate jaggies and shimmering.

EDIT: Another thing of note, is that while those comparison screenshots look great, they are compared to TAA/FXAA etc. It would be interesting to see against native with MSAA/SSAA, and without JPG compression. I have always found it hard to compare AA via screenshots. For example, 4xMSAA looks pretty good vs SSAA in screenshots, but in person the difference is night and day.

SIrPauly Apr 6, 2020 05:59 AM

Personally desire more quality with the denoising. I think 2.0 is good quality overall, especially with the context of performance savings. Certainly, not perfect or ideal for all, but still maturing and evolving. I disagree with Z and think it should be a flagship feature.

demo Apr 6, 2020 06:08 AM

Quote:

Originally Posted by SIrPauly (Post 1338186162)
especially with the context of performance savings.

Yeh exactly. That's what it sets out to achieve, and to be fair, delivers great results.


Quote:

Originally Posted by SIrPauly (Post 1338186162)
Certainly, not perfect or ideal for all, but still maturing and evolving. I disagree with Z and think it should be a flagship feature.

Oh it's definitely a flagship feature. This is paving the way for the future of AA filtering IMHO.

Ziklitschli Apr 6, 2020 06:45 AM

Thank you demo for the kind words. That's exactly what I would say about you, trusting your opinion and judgment.

Quote:

Originally Posted by SIrPauly (Post 1338186162)
I disagree with Z and think it should be a flagship feature.

I guess you come at it from a flexibility angle. I just let perfection be the enemy of good in this, would prefer insane quality features up there (ie Tensor cores run, fixed TAA you mentioned somewhere) and DLSS next to the 3rd/ 4th dot, in a slightly thinner font. I escape from console with all its reconstructed, blurred, grained, vignetting dynamic shortloded mess to my PC and it kind of bugs me to see nVidia focusing so much on consolesque techniques.

There's an undeniable potential for suprise in the tech though, thanks to the Tensor Cores, and I see the appeal. Miracles don't come easy though.

Nunz Apr 6, 2020 06:55 AM

DLSS 2.0 is definitely a massive improvement, but can't say I see it getting much better. DLSS 1.0 was pretty awful, so this new version had a lot of room for improvement; with the updates fixing a lot of the blur and Halo issues with 1.0, I don't see them making another large quality jump without some new technology that will be marketed as something else.

SIrPauly Apr 6, 2020 07:22 AM

I find these discussions really interesting.


Here is Nvidia discussing some aspects of 2.0:


https://developer.nvidia.com/gtc/2020/video/s22698



Quote:

Originally Posted by Nvidia
In this talk, Edward Liu from NVIDIA Applied Deep Learning Research delves into the latest research progress on Deep Learning Super Sampling (DLSS), which uses deep learning and the NVIDIA Tensor Cores to reconstruct super sampled frames in real-time. He discusses and demonstrates why scaling and image reconstruction for real-time rendering is an extremely challenging problem, and examines the cause of common artifacts produced by traditional up-sampling techniques. With this background, Edward then shows how the DLSS, deep learning based reconstruction algorithm, achieves better image quality with less input samples than traditional techniques. The talk concludes with a walkthrough of DLSS integration in Unreal Engine 4 and a look at DLSS in the acclaimed sci-fi adventure “Deliver Us The Moon” from KeokeN Interactive


Destroy Apr 6, 2020 07:27 AM

Quote:

Originally Posted by Ziklitschli (Post 1338186109)
Btw one reason to hate it is the day when it's just implemented without choice, because "people can't see a difference", "just as good as native etc.

Exactly.

acroig Apr 6, 2020 08:37 AM

Quote:

Originally Posted by Exposed (Post 1338186125)
There's zero shimmering in the two DLSS 2.0 games I've played (Youngblood and Deliver Us the Moon).

Add Control to that list.

demo Apr 6, 2020 08:43 AM

Actually yes ^ I'm pretty annoyed with how TAA is the new standard and most games don't offer MSAA or SSAA at all anymore. At least it's better than FXAA I suppose (Ughhh).

I understand the possible issues with MSAA, but at least offer SSAA (thankfully some games natively support downscaling).
But yeh, I miss the old days when mixed AA modes were the height of IQ enhancements, allowing users to go above and beyond what the game offered or devs intended. It's the main reason I ran SLI/CFX all those years. Not just for the AA quality itself, but the impact those modes also had on textures, shaders, motion clarity, etc.
Now it just seems you set your game to max, enable TAA, and that's it. Doesn't matter if you upgrade hardware, the games graphics cannot be enhanced any further. Hell, we're lucky if the game has a sharpening filter slider to de-blur the shader AA... haha progress.

Now get off my lawn.


All times are GMT -5. The time now is 04:18 AM.

Powered by vBulletin® Version 3.6.5
Copyright ©2000 - 2022, Jelsoft Enterprises Ltd.
All trademarks used are properties of their respective owners. Copyright ©1998-2011 Rage3D.com