• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

DLSS 5 - Generative AI

stop the slop

You won't, you haven't, you never will - especially with the disingenuous methods you lot have all tried so far.

Don't buy games that integrate it. While you still can.

Then just turn the toggle off when you see it in games that have it. While you still can.

Destiny.gif


You own a 5090.

You helped do this to yourself.
 
What was the lie?
One of them was "changed geometry!!!"

Edit:
Also "DLSS5 adds fake objects!!!" & "DLSS5 breaks animation!!!" and now misleading videos like the ones Erek is peddling.

DLSS(1,2,3,4,4.5,5) is all A.I. and they make a small minority cry while hte majority os users just use and game.
 
Last edited:
One of them was "changed geometry!!!"

Edit:
Also "DLSS5 adds fake objects!!!" & "DLSS5 breaks animation!!!" and now misleading videos like the ones Erek is peddling.

DLSS(1,2,3,4,4.5,5) is all A.I. and they make a small minority cry while hte majority os users just use and game.

Don't forget: "NOTICE HOW THEY ARE NEVER SHOWING IT IN MOTION, BECAUSE IT DOESN'T WORK IN MOTION!"

Edit: Ah, might be what you mean by 'the animations'

forgot-about-that-willem-dafoe.gif
 
One of them was "changed geometry!!!"
While that is technically false, the final image does look different than the original geometry.
Edit:
Also "DLSS5 adds fake objects!!!" & "DLSS5 breaks animation!!!" and now misleading videos like the ones Erek is peddling.
IDK about that, we need to see more to definitely confirm or deny that. I don't think nvidia is any more reliable of a narrator than the detractors in this.
DLSS(1,2,3,4,4.5,5) is all A.I. and they make a small minority cry while hte majority os users just use and game.
Yes they are all AI but only DLSS 5 is generative AI, that is not trying to reconstruct what's there but make up new details.
 
So many hate posts clearly have the intensity slider cranked up to max to purposely make the image look bad so they can push a negative narrative.

1788038163400.png


It's like Sharpening, a little bit of it definitely enhances the picture, but if you crank it up to 200% then of course it's going to look like shit.
 
Ive been playing with it today. It makes Oblivion look amazing, but adds flicker with frame gen on my 5080 laptop getting around 38fps maxed out, 60fps with frame gen.
FF7 Rebirth seems flawless and still runs well and looks better.

There are alot of fake pictures of the output with cleft chins... take that with a grain of salt.
I think the image of Tifa is washed out due to HDR.
ff7rebirth_ 2026-08-29 16-47-35_2.png
ff7rebirth_ 2026-08-29 17-02-57_13.png
 
Last edited:
Ive been playing with it today. It makes Oblivion look amazing, but adds flicker with frame gen on my 5080 laptop getting around 38fps maxed out, 60fps with frame gen.
FF7 Rebirth seems flawless and still runs well and looks better.

There are alot of fake pictures of the output with cleft chins... take that with a grain of salt.
I think the image of Tifa is washed out due to HDR.
View attachment 822776
Wow, I just found out that is a guy. Wow.
 
Tried it out on my 4090. About double the fps of my 5080 laptop in oblivion. That’s with my desktop at 4k and my laptop at 2560x1600.

Tried KCD2 it looks outstanding, but is limited to DLAA for DLSS 5 to hook, so 40ish fps in 4k on the 4090.
 
Blame square enix. It didn’t change his face geometry. The main difference is shading and skin texture and the materials of his clothes.
couldn’t a screen filter change the frame to appear to have more geometry?

been wondering what’s the difference

not really sure how dlss5 works but if it’s not being done per textures i imagine it already is morphing the apparent geometry
 
couldn’t a screen filter change the frame to appear to have more geometry?

been wondering what’s the difference

not really sure how dlss5 works but if it’s not being done per textures i imagine it already is morphing the apparent geometry
Shading can definately do that. Reference the asian chick makeup videos for evidence.
Pulled this off the Reno DX discord. The leaked verson doesnt change as much as they showed in the reveal videos.

RE4Leon.jpgRE4LeonDLSS5.jpg
 
So many hate posts clearly have the intensity slider cranked up to max to purposely make the image look bad so they can push a negative narrative.
Of course. This was one of those things that people decided was bad and thus with minds already made up they go looking for proof of their preconceived notion, not looking for evidence with an open mind. I mean hell, we saw all kinds of shit straight out made with other AIs to shit on it, which is hilarious to me: "I hate AI generated slop, so I'm going to use an AI to make AI generated slop to make a strawman to prove my point in this argument!" It's never done in good faith, always as a "my preconceived notion is right, and thus anything that supports it is correct, anything that challenges it is wrong."

To me, it looks like it can be very useful, when used right. I'm sure some games will use it wrong, they'll turn it all the way up and slap it on and call it good. Those will either get turned off, or modded (same as bad HDR gets modded now). However it is real good at improving some things. So far depth and shadow detail are the two that I notice and appreciate the most. I've just been looking at stuff online, I haven't gotten the modded DLLs to play with on my system since they aren't readily available on the RenoDX site.

It sounds like the real issue may end up being performance: If it halves your FPS, that is going to be a heavy cost that may not be worth paying in most cases. While it does improve the image, it isn't transformational like path tracing. That tanks FPS, but it transforms a game so much I'll pay that premium. This improves things, but doesn't transform them for the most part. So we'll have to see how performance goes in various games, but I can see that being something that for me makes it a thing I don't usually turn on.


However the performance thing, and also the fact that nVidia wants to officially gate it to 50 series, also tells me that people need to STFU and quit acting like it is going to fuck games: Clearly it is NOT going to be something mandatory in games. No dev with an ounce of sense is going to require a technology that only runs on the newest cards, and not on the consoles, and even then is so performance heavy that you realistically need a high end card. It'll be an option, not a requirement. So it isn't something to worry about. If you don't like it, just don't use it. If your card doesn't support it, don't worry games won't be requiring it any time soon, if ever.

I think part of it is just the knee-jerk "hate anything with the name AI in it" thing that so many people have. nVidia shouldn't have called it AI, they should have used other terms (like "a vision transformer based processing network that is able to improve lighting and shadow detail"). However I think part of it is this misguided idea that if there was just more "optimization" games could look photoreal without any new ways of doing things. If the engineers just engineered harder, GPUs could be so much faster and use less power. If the coder just coded harder, a game could run on a potato at high FPS. If the artists just arted harder the graphics could look amazing, without any new technology. None of this, of course, costs time or money, it is all about people being lazy and unwilling to just do the magic thing that would make it better and it wouldn't cost any more or take any more time.
 
There is a lot of chronic hate for anything "nvidia" and/or "AI". DLSS 5 happens to contain both, therefore it is lost cause convincing upset individual(s) of any benefits.
It a far to vocal minoity.
AMD is down to 8% sales, the noise is for to loud ofr ht numbers.

But Youtube shows this, the flooding of DLSS5 vidoes show the majoirty WANTS this.

Just ignore the cryybabies, it is coming and they cannot stop it.
 
After seeing a lot of videos I think it looks shitty in most cases. It might be useful for random NPCs that have low detail, but for main characters it changes the look and style too much. And people look shiny and like they have a lot of makeup on, even when they are supposed to be dirty.
 
  • Like
Reactions: M76
like this
DLSS upscaling pre-5 is literally generative AI - it generates more/missing pixels/data to fill in the complete higher res image
So literally not generative AI. There is no need to fanboy so hard that you misrepresent things. No one would normally call a machine learning based upscaler generative AI. DLSS pre 5 is an upscaler / frame interpolator nothing more, it does not try to make up details that were never there originally, it tries to fill the missing data between two existing pixels / frames. A wholly different concept than DLSS5 which is putting a completely new facade on things. Whether you like the results or not there a significant difference. I'm not against it, but not acknowledging this fundamental difference just gives ammo to the detractors.

The way nvidia coaches reviewers how they are allowed to talk about it just makes it look like they have something to hide. It makes haters hate it even more. The only way they would convince people is by letting the product speak for itsellf. But currently it is a mixed bag.
 
After seeing a lot of videos I think it looks shitty in most cases. It might be useful for random NPCs that have low detail, but for main characters it changes the look and style too much. And people look shiny and like they have a lot of makeup on, even when they are supposed to be dirty.
This is my assessment as well, it is great for random NPCs, but for named main characters if you dial it up it fundamentally changes how they look, not to mention that it seems to add 15-20 years to everyone's age. And if it is dialed down enough so it doesn't make the character look like a completely different person then the effects are so minimal that it hardly justifies using it at all.
 
So many hate posts clearly have the intensity slider cranked up to max to purposely make the image look bad so they can push a negative narrative.
But if you dial it back low enough to the point where it is not uncanny then there is barely any effect and it does not make it anywhere near photorealistic, so it does not justify the performance hit.

This looks like a photo of a real person.
Except it looks nothing like the person it supposed to be.
Blame square enix. It didn’t change his face geometry. The main difference is shading and skin texture and the materials of his clothes.
Sorry, why blame square enix? Everyone can tell it's a twink in the original, so if nvidia's magic AI turns it into a girl that's clearly their fault.
 
When available I will try it. If in a particular game it improves the experience, great, if not I won’t use it for that game. The sample with Control looked promising. Plus I would expect this method will improve over time.
 
Without Googling I wonder if DLSS5 can work on much older (15, 20 or more) games and how would they look if it did?
 
From many comparison photos posted here, it looks like dlss5 takes ambient occlusion to the next level, creating more realistic subtle shadows on the face, which in turn, makes the face appear like an actual photo.

Faces in game without dlss5, in comparison, look like flashlight is pointing right at their face, eliminating all shadowing, thus making the character look artificial. Unless there is a very obvious light source above, for example, that creates heavy shadow.

(Talking about Leon and that medieval man photos)

Tangle:
I've read somewhere someone tried modding dlss5 to work on 4 series and it took appx 50% performance hit. That leads me to believe anything below won't work, apart from taking screen grabs due to incredibly low fps.
 
animations feel weird with that graphics filter enabled
I am assuming that is due to uncanny valley effect. To your brain(?), characters look almost human-like in appearance, but your brain notices that's not how actual humans animate.

(Kinda like replying to forum AI)
 
So literally not generative AI. There is no need to fanboy so hard that you misrepresent things. No one would normally call a machine learning based upscaler generative AI. DLSS pre 5 is an upscaler / frame interpolator nothing more, it does not try to make up details that were never there originally, it tries to fill the missing data between two existing pixels / frames. A wholly different concept than DLSS5 which is putting a completely new facade on things. Whether you like the results or not there a significant difference. I'm not against it, but not acknowledging this fundamental difference just gives ammo to the detractors.

The way nvidia coaches reviewers how they are allowed to talk about it just makes it look like they have something to hide. It makes haters hate it even more. The only way they would convince people is by letting the product speak for itsellf. But currently it is a mixed bag.

Still literally yes generative AI - same with framegen - generative AI - the pixels/missing details/information/frames/etc were not there before - they are/were generated by the AI

animations feel weird with that graphics filter enabled

The animations are the most jarring aspect - time to get AI to make animations then too IMO 😁
 
Last edited:
From many comparison photos posted here, it looks like dlss5 takes ambient occlusion to the next level, creating more realistic subtle shadows on the face, which in turn, makes the face appear like an actual photo.

Faces in game without dlss5, in comparison, look like flashlight is pointing right at their face, eliminating all shadowing, thus making the character look artificial. Unless there is a very obvious light source above, for example, that creates heavy shadow.

(Talking about Leon and that medieval man photos)

Tangle:
I've read somewhere someone tried modding dlss5 to work on 4 series and it took appx 50% performance hit. That leads me to believe anything below won't work, apart from taking screen grabs due to incredibly low fps.
30 series doesn’t do FP8. Someone made it work at single digit fps. 40 and 50 series perform similarly, but 40 series other than the 4090 is vram constrained.
 
Last edited:
Wow, I just found out that is a guy. Wow.
I have been flipping through the pages of this thread and was genuinely perplexed as to why this AI crap keeps drawing Cloud as a female? Yes, he has long hair, but that should quantify him as being a female to an AI right? Whatever. Guess its back to pixel games until they decide to invade the sacred realm. Yes, I know about Lossless Scaling. :)
 
  • Like
Reactions: erek
like this
Still literally yes generative AI - same with framegen - generative AI - the pixels/missing details/information/frames/etc were not there before - they are/were generated by the AI
Filling in the gap between two pixels = interpolation. Completely making up (Guessing) the details is generative AI. They are fundamentally different. It's neither a good nor a bad thing in of itself, so IDK why deny it.
 
Back
Top