Nvidia "confirms" DLSS 5 relies on 2D frame data as testing reveals hallucinations

Polls in 2026 mean jack sht. It only takes a few douchebags to employ bots to swing the majority in their favor.

I haven't trusted public polling for a long time (on any subject).
Techpowerup filters bots.
While the results are different because of the poll questions. 22% are on the fence while 7.5% loved it, 54.1% didn't like it and 6.3 only if it improves performance. Reminder that techpowerup is mostly made up of enthusiast pool.
1000056480.jpg
 
Polls in 2026 mean jack sht. It only takes a few douchebags to employ bots to swing the majority in their favor.
Even ignoring the bot potential, opt-in polls are the very definition of cherry-picking sampling error. If 90% of people moderately support an issue, but 10% vehemently oppose it, the latter will generally overwhelm any opt-in mechanism.
 
Because it uses neural networks built on Tensor Cores just like super sampling and frame generation. What this new thing is really is an amalgamation of those two technologies put together and likely running on the same Transformer Neural Network as DLSS 4+ does
Because you're defending their talking points. You are both a fool in your understanding of AI and a complete *** when trying to discus technology. But you're just an AI bot, aren't you?
 
Even ignoring the bot potential, opt-in polls are the very definition of cherry-picking sampling error. If 90% of people moderately support an issue, but 10% vehemently oppose it, the latter will generally overwhelm any opt-in mechanism.
I can't wait for you to go broke when AI breaks the economy and it can't fund itself anymore. What tech bros don't understand about their infinite money glitch is that no one can buy anything if no one has money
 
In reality, Nvidia should had provided a development kit to devs for designing their models. It would have been their choice to implement the new renders or not, but passing over it after the game is complete with adding an additional GPU for the functionality is just ludicrous.

AI is a tool, and it will do crazy things in the industry, however this should have been a proof of concept and not a DLSS implementation.
 
Even ignoring the bot potential, opt-in polls are the very definition of cherry-picking sampling error. If 90% of people moderately support an issue, but 10% vehemently oppose it, the latter will generally overwhelm any opt-in mechanism.
Even cherry picked bot removed polls have data that can be extrapolated which reflects the majority comments in articles and yt videos discussing dlss even bias echo chambers if you want the full spectrum of subjective data objectively. Even if you average out all the polls in a meta analysis they would still show significant majority level of disproval. Not sure why some have a affinity to the status quo!

Please let us know why Digital foundries got ratiod and reneged on their original video!
 
I can't wait for you to go broke when AI breaks the economy
There are two types of people in the world: those happy to see the success of others, and those who respond with nothing but jealousy and hate. Two points:

- Most AI firms are recording astronomical margins (40%-100%+); many are already turning large profits; those which are not are doing so because they're prioritizing growth, not because their business model is flawed.

- Even as the Internet dot-com bubble was bursting, total Internet penetration and usage, and total spending on Internet-related products continued to increase monotonically. Certainly the AI bubble will burst and cause a few high-profile bankruptcies at some point. It won't slow down the industry one whit.

... What tech bros don't understand about their infinite money glitch is that no one can buy anything if no one has money
What you fail to understand is that OpenAI, Anthropic, and NVidia could all go bankrupt overnight and the total amount of M2 money supply in circulation wouldn't change by one penny. Market cap is a convenient fiction; meaningless until realized.
 
Even cherry picked bot removed polls have data that can be extrapolated....
Basic statistics: you can't "extrapolate" away sampling bias from opt-in polls. And the vast majority of such poll results -- particularly on narrowly-focused populations like enthusiast sites -- are utterly, absurdly out of touch with reality, and represent virtue signaling (AI bad! BAD!) more than reality. I won't even mention the fact that, in this particular case, 99.5% of those responding haven't even seen DLSS 5 in action, making their opinion useless.

Please let us know why Digital foundries got ratiod and reneged on their original video!
Glad you asked! For the same reason Dr. Frankenstein cowed before the mob, when they arrived at his castle with torches and pitchforks. For the same reason Guy Fawkes confessed to attempting to blow up Parliament, after being tortured on a rack for hours. For the same reason the Central Park Five all confessed to a brutal rape that none committed.
 
There are two types of people in the world: those happy to see the success of others, and those who respond with nothing but jealousy and hate. Two points:

- Most AI firms are recording astronomical margins (40%-100%+); many are already turning large profits; those which are not are doing so because they're prioritizing growth, not because their business model is flawed.

- Even as the Internet dot-com bubble was bursting, total Internet penetration and usage, and total spending on Internet-related products continued to increase monotonically. Certainly the AI bubble will burst and cause a few high-profile bankruptcies at some point. It won't slow down the industry one whit.


What you fail to understand is that OpenAI, Anthropic, and NVidia could all go bankrupt overnight and the total amount of M2 money supply in circulation wouldn't change by one penny. Market cap is a convenient fiction; meaningless until realized.
keep trading, brother. AI bros are just the new crypto bros. AI will crash and you'll disappear just like all the crypto bros did
 
keep trading, brother. AI bros are just the new crypto bros. AI will crash and you'll disappear just like all the crypto bros did
AI is worse. Crypto was an isolated resource hog that interrupted the GPU market for a short while. When it cooled, life was back to normal. With AI, when it crashes, its going to bring down the whole house. People are utilizing AI output in almost every facet of life.

You're already seeing IQs drop sharply among the masses because....why learn or remember anything anymore when AI does both for you? (imo this been happening before AI went mainstream, AI just accelerated the process).

Right now AI is deriving its resources from human-input over the decades (mostly verifiable factual data)...what happens when AI starts consuming recycled AI slop from the 2nd generation? and so on,etc.

The one constant that has been around since the dawn of the first micro chip is:
Garbage in. Garbage out.
 
In reality, Nvidia should had provided a development kit to devs for designing their models. It would have been their choice to implement the new renders or not, but passing over it after the game is complete with adding an additional GPU for the functionality is just ludicrous.

AI is a tool, and it will do crazy things in the industry, however this should have been a proof of concept and not a DLSS implementation.
It *IS* a proof of concept, and has not yet been released. <shrugs>
 
AI is worse. Crypto was an isolated resource hog that interrupted the GPU market for a short while. When it cooled, life was back to normal. With AI, when it crashes, its going to bring down the whole house. People are utilizing AI output in almost every facet of life.

You're already seeing IQs drop sharply among the masses because....why learn or remember anything anymore when AI does both for you? (imo this been happening before AI went mainstream, AI just accelerated the process).

Right now AI is deriving its resources from human-input over the decades (mostly verifiable factual data)...what happens when AI starts consuming recycled AI slop from the 2nd generation? and so on,etc.

The one constant that has been around since the dawn of the first micro chip is:
Garbage in. Garbage out.
I believe that the excessive use of LLM's doesn't just lead to a lower IQ, it is extensively used by low IQ people in the first place. One guy I know, who is heavily reliant on these LLM's, explained how a large thesis should be broken down into smaller chunks with the opening statements and conclusion be fed to the LLM first, with other parts of the thesis entred into the software later and separately. Thus, the user needs to read the thesis first prior to being able to use the LLM, defeating the entire purpose of using it as an analytical tool, as the reader will have come to a conclusion already. And his post was written by Ollama, which was quite clear due to wording and structure.

An electrical engineer complained endlessly yesterday in a technology group I'm in about how these tools are unreliable and constantly give erroneous data and analyses. He said that they have stopped using LLM's as even purpose built LLM's by Siemens and Bosch can't recognise their own products and create proper graphs.

I can't find a single use for these LLM's that can't be done better and more efficiently without these tools. It might be possible to write code but considering it's been reported that it takes more time to correct errors than it would had the code been written by hand, I believe that the only people who are 'more productive' through the use of these tools are not qualified to perform their tasks in the first place.
 
Last edited:
I believe that the excessive use of LLM's doesn't just lead to a lower IQ, it is extensively used by low IQ people in the first place. One guy I know, who is heavily reliant on these LLM's, explained how a large thesis should be broken down into smaller chunks with the opening statements and conclusion be fed to the LLM first, with other parts of the thesis entred into the software later and separately. Thus, the user needs to read the thesis first prior to being able to use the LLM, defeating the entire purpose of using it as an analytical tool, as the reader will have come to a conclusion already. And his post was written by Ollama, which was quite clear due to wording and structure.

An electrical engineer complained endlessly yesterday in a technology group I'm in about how these tools are unreliable and constantly give erroneous data and analyses. He said that they have stopped using LLM's as even purpose built LLM's by Siemens and Bosch can't recognise their own products and create proper graphs.

I can't find a single use for these LLM's that can't be done better and more efficiently without these tools. It might be possible to write code but considering it's been reported that it takes more time to correct errors than it would had the code been written by hand, I believe that the only people who are 'more productive' through the use of these tools are not qualified to perform their tasks in the first place.

Nailed it. These LLMs aren't AI, they're Artificial Stupid - at best. Scrapers for clankers, and the marketing is so pathetic. All of DLSS is the same nonsense - not one "intelligent" frame has ever been generated by this non-tech. It's simply Nvidia pushing their weaker "tensor cores" for mass marketing, creating an industry that doesn't exist, almost like a religion. And it's actually hilarious watching all the zealots flock to every buzz-word, jumping on every post to defend a technology that doesn't even exist because they just want to look SO COOL.

The tech is decades-old. Even Zork had better responses than most of the LLMs do.
 
That's the problem, Nvidia is changing the look of characters and art design without the original developers or characters artist's permission. It's like taking a photo of a painting and throwing an AI filter over it and saying it's "better". ...

The scary part is that Nvidia is in control. If some government contacts Nvidia with the question to make sure the visualisation of a character is altered in a certain way, Nvidia can make that happen regardless of the opinion of the creator of the game. So they can by default "mod" a game and it is then to the user who doesn't agree to find a way to undo that. How many % of PC gamers are into modding and know what to do? And what about console or Geforce NOW gamers?
 
Developers can adjust the level of enhancement applied by DLSS5, and even allow or disallow it for specific scenes or models within a scene.

Based on some screenshots shared in this specific thread, using current rendering technique (be it native or upscaling). If what you are saying is true on the sentence I quote, then what makes DLSS 5 very special vs the current technique?

The screenshots showed nothing wrong with the faces rendering that it needs special "AI filter" shown by DLSS 5. And then by your logic, it will add additional jobs for them to apply the "enhancement" one by one on specific scene or character.

The game development already took longer than it should and then the developers now face this "dilemma".
 
If what you are saying is true on the sentence I quote, then what makes DLSS 5 very special vs the current technique?
You failed to grasp the context of my point, which was simply to refute the earlier suggestion that DLSS5 exerts some draconian level of control over rendering, horrendously disrupting the "artistic integrity" of videogames: a global crisis as terrifying as a swarm of alien astrophages consuming our Sun. *

(* apologies, Mr. Gosling)

The screenshots showed nothing wrong with the faces rendering that it needs special "AI filter" shown by DLSS 5.
There was nothing wrong with the graphics rendered by my favorite game from 1987, either. Should we roll back all the graphics enhancements since then, since that was "good enough"?
 
Last edited:
yRaz actually asked the most important question in this entire thread and it got buried: why call it DLSS at all?

DLSS 1-4 earned goodwill because it solved a real problem – getting more fps without destroying image quality. Calling this completely different technology "DLSS 5" was a deliberate choice to borrow that trust. If they'd called it "Nvidia Neural Rendering" or whatever, people would've evaluated it on its own merits instead of comparing it to what DLSS has always been.

The branding decision alone caused half the backlash. The other half is the 2x 5090 requirement for a tech demo. Nothing says "this is ready for consumers" like needing $4000+ in GPUs to run it. lol
 
Calling this completely different technology "DLSS 5" was a deliberate choice to borrow that trust. If they'd called it "Nvidia Neural Rendering" or whatever, people would've evaluated it on its own merits I
I seriously don't know where you people get this stuff. It's been right there in the name since Day 1:

Deep Learning ( AI based enhancements ) Super Sampling (examining a large area around each pixel to determine how to modify them.

DLSS has been more than just simple antialiasing since version 1 ... there is literally nothing new here except the degree of enhancement the more sophisticated models are able to apply.
 
In a very bad way. And you've just looked at still shots... the motion makes it even worse.
Yep look at the Nvidia's with PCmag dlss 5 demos. All of the Videos show Nvidia toggle off dlss 5 when moving around.
Only gets turned on when still.
 
Because you're defending their talking points. You are both a fool in your understanding of AI and a complete *** when trying to discus technology. But you're just an AI bot, aren't you?
Actually I'm a retired Electronic Engineer with 33 years experience in my field that has been building PCs on the x86/x64 platform since 1988 and actually worked for Gateway Computers from 1988 - 1993 as a component compatibility tester back in the Wild West days of the IBM PC Compatible Era (A misnomer if there ever was one).

Back in the 80's it was "Computers are going to steal all our jobs" Back in the 90's it was "The Internet is going to steal all our jobs." Neither of which happened in fact both led to even more jobs. Today it's "AI is going to steal all our jobs" but more likely the results will be the same as with Computers and with the Internet, in the long run more jobs will be created than were lost. AI is the third "Disruptive Technology" I've seen in my lifetime so I have no irrational fear of it. If you think the memory shortage is bad now go look up the shortage in 1988 where a mere 2 MB of RAM went for $505 or the equivalent of $1,418 today .... for just 2 MB of RAM for your 386 CPU.
 
Last edited:
Back