• Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.

DF: Control vs DLSS 2.0: Can 540p Match 1080p Image Quality? Full Ray Tracing On RTX 2060?

Armorian

Banned
This is the real gamechanger. I lways felt that native 4K rendering was a big waste of power and DLSS sixes that. With this shit even "8K" will be possible on consoles :pie_roffles:

Question is: will AMD, MS or Sony be able to replicate Nv solution.
 
Last edited:

TGO

Hype Train conductor. Works harder than it steams.
Ain't seen the Video but
"Can 540p Match 1080p Image Quality"
No.
But I believe a 1080p can get close to 4k with this technology
How did I do?
 

Evilms

Banned
DLSS 2.0 is simply crazy !

py7xfhH.png

EUwzGNYWAAILTcQ

uTKfy6N.png
 
Last edited:

Fbh

Member
See, this is the sort of cool "secret sauce" tech that I want to see next gen. Cool "before/after" comparisons where the difference is very clear.
And not these "theoretically maybe if games are made a certain way we could potentially see advantages under the right circumstances" sort of things.
 

Portugeezer

Member
Really impressive, from where I am sitting they look 99% the same.

I expect next gen games to use similar type of techniques. Maybe 4K won't be such a waste.
 

Shmunter

Member
See, this is the sort of cool "secret sauce" tech that I want to see next gen. Cool "before/after" comparisons where the difference is very clear.
And not these "theoretically maybe if games are made a certain way we could potentially see advantages under the right circumstances" sort of things.
Wonder if that Sony sound chip could be used for something like this. They say it’s like a cell processor, so a computer unit great at real-time image Processing.

They already have quite a good reality engine on the TVs that does a pretty convincing job of increased detail. .
 
  • Like
Reactions: Fbh

INC

Member
said it before, and ill say again, VR needs this tech now, and it needs a universal nvidia control, not per game basis
 

TaySan

Banned
I game at 1440p 120hz instead of 4k 60hz because i think 4K is a waste of resources, but this is truly great stuff that will make 4k gaming more accessible. I might not even need to upgrade my 2080 Super this year if more games get 2.0 support.
 

GymWolf

Member
good stuff.

does metro exodus use dlss 2.0?
i kinda wanna try my 2070super at 4kdlss with all the rtx activated to see if 60 frame are possible...
 

888

Member
Damn!

Im getting some nvidia shares monday. Their shares are already holding up better than Sp500.

This is a big selling point for RTX3K.
I can feel safe buying one and knowing they will run nextgen RT games really well.

They have been good to invest in lately.
 

TaySan

Banned
good stuff.

does metro exodus use dlss 2.0?
i kinda wanna try my 2070super at 4kdlss with all the rtx activated to see if 60 frame are possible...
No. Only Control, Wolfenstein Youngblood and another game i forgot the name of uses 2.0.
 

DaGwaphics

Member
See, this is the sort of cool "secret sauce" tech that I want to see next gen. Cool "before/after" comparisons where the difference is very clear.
And not these "theoretically maybe if games are made a certain way we could potentially see advantages under the right circumstances" sort of things.

I think AI/ML is the way of the future, without question. It would be foolish for devs not to take advantage of these techniques, the bandwidth savings alone would be monumental. All aboard the GPU compute train.
 

pawel86ck

Banned
Agree, I still have many DVD movies (not all movies have BD transfer) and even BD movies would look better as well.
 

Jigga117

Member
r-kelly-usher-shocked-music-video-reaction-gif-1363306207k.gif

Damn, I knew this tech was impressive but you have to have alteast something of detail in there or it turns to shit.
But that shit is impressive.
They need that crap built into TVs



They already do in 4k and 8K televisions. Been around a couple of years now.
 

StreetsofBeige

Gold Member
If this DLSS shit works this well in actual gaming (and not just images), it looks fantastic. SOmewhere else someone posted 540p pics DLSS'ed to 1080p and it looked great like it really was a native 1080p game.
 

dxdt

Member
r-kelly-usher-shocked-music-video-reaction-gif-1363306207k.gif

Damn, I knew this tech was impressive but you have to have alteast something of detail in there or it turns to shit.
But that shit is impressive.
They need that crap built into TVs
The latest NVIDIA Shield has AI upscale that's pretty cool. But I would love to have this on my LG OLED.
 

Myths

Member
Basically, a rack and a half should be the upper limit for RTX3K. I will definitely be upgrading, this is looking insanely good considering that’s on 2080ti alone.
 

Skyr

Member
I'VE SEEN FUCKING LIGHT

DLSS (60+ FPS, 30% GPU utilisation )

utGeHkX.jpg


pK4nIG2.png


NATIVE (cca 40FPS)

qDDxD6h.jpg



Just HOW?
I do actually prefer the sharper look with dlss here. I'm sure you can achieve it by using a sharpness filter with native 4k but still.. this is incredibly impressive.

Fuck me I'm so excited to upgrade from my 1080ti to a 3080ti now. Not only probably double the performance but DLSS on top. I'm gonna coom!
 
Last edited:

M1chl

Currently Gif and Meme Champion
I do actually prefer the sharper look with dlss here. I'm sure you can achieve it by using a sharpness filter with native 4k but still.. this is incredibly impressive.
I actually think, that DLSS result is better. Which is crazy. Also no MSAA for DLSS, so another compute saving.
 

Bogey

Banned
Does anyone know how this works on the development/implementation side?

I'm curios if the model needs to be trained beforehand. Eg devs doing some playthroughs in 4k and possibly lower res at the same time, and thereby train the model of how to best reconstruct the image depending on current scene/environment etc.

Or is it literally a one size fits all solution that needs no game specific training before it can be applied?
 

Skyr

Member
Does anyone know how this works on the development/implementation side?

I'm curios if the model needs to be trained beforehand. Eg devs doing some playthroughs in 4k and possibly lower res at the same time, and thereby train the model of how to best reconstruct the image depending on current scene/environment etc.

Or is it literally a one size fits all solution that needs no game specific training before it can be applied?
I think it used to be like that. With DLSS 2.0 it became a generalized model.

Put the timestamp in for you:
 
Top Bottom