Okay but I still don’t want a filter putting makeup on all the characters
Comparative performance claims like “faster” are completely meaningless without figures, or even a point of comparison.
Kien’s runtime also improved with AMDNR version 0.3.4. On an AMD Radeon RX 7800 XT at 1440p (FSR Quality), the update cut the neural network pass from 73 ms down to 52 ms, yielding a 29% drop in render time.
So, assuming the game’s own render pipeline is instant (ie. 0.0ms, entirely unreasonable, but whatever) - 1000/52= ~19.23, so at unrealistic ideal conditions the DLSS ai-nonsense-pass runs at ~19 fps with the metioned hardware & resolution
edit: 1000/73= ~13.7, so hooray for the 6 extra frames. It is massive uplift in terms of percentage, but still unusable for gaming, imo.
now add 90x frame gen to turn it into unplayable *high framerate* gaming
They are coming for your anime girls!
Who is desperate to use DLSS 5 even on an AMD card… ?
I imagine it’s the same crowd as every other DLSS version.
The new stuff looks great once the devs get the tools. It’s going to be a highly popular feature just like framegen and early versions of diss which also had gamers screeching after seeing early performance hits.
I’ve found XESS tends to look better but I haven’t tried the newest DLSS.
All of the upscalers suck with transparency effects, XESS just sucks slightly less
Nice AI hallucinations. The ongoing destruction of the planet is definitely worth a small but noticeable upgrade in video game graphics!
DLSS5 sucks for many reasons, but “the ongoing destruction of the planet” is not really one of them, as it runs locally on a regular GPU.
The models were still (and probably still are actively being) trained in giant data centres that devour energy and water.
Sure, you using it technically doesn’t hurt the environment any more but the technology as such definitely does.Training AI models takes quite a bit of energy. But probably less than you think. Without knowing how much you think I can’t say for sure, but I can say that AI is so far down the list of water consumers that any time it comes up I can be fairly confident.
Without knowing for sure we should not be giving data centers from giant corporations the benefit of the doubt.
30 million gallons of water stolen noticed by the entire towns water pressure dropping seems rationally low to not have triggered a review of their systems?
https://www.politico.com/news/2026/05/08/georgia-data-centers-water-00909988
They chose small towns and desert areas because they can be bullied and lobbied for cheaper for the investors, not because its for the good of the community.
DLSS models are not even in the same league as LLMs from a training perspective.
The comment section is full of people trying to use anti-LLM arguments against DLSS because they both use neural networks.
It’s like people are against matches because they’re hot and nuclear weapons are hot so matches are basically nuclear weapons.
Yeah. It’s beyond unnuanced thinking and is a bit brain-dead :(
See AI, neuron activates, “AI bad”
The models were still (and probably still are actively being) trained in giant data centres that devour energy and water.
I hate to tell you, but your Lemmy instance is in a datacenter too. Most everything on the Internet is in a datacenter.
The important detail is that an upscaling model and a large language model are vastly different things despite both containing the word model and using neural networks.
As to training costs. You can train an upscaling model on your phone in a few hours. It would take you millions of years to train an LLM on your phone (not to mention requiring about 12,000x as much RAM).
You’re just repeating talking points on topics without understanding them or how they apply.
I do not have a problem with upscaling models and DLSS in general (up to ver. 4.5) is perfectly fine.
However, DLSS 5 is not just upscaling anymore, it’s fully analyzing and generatively replacing parts of the image (lighting, material properties, …) And image based gen AI definitely cannot be trained on a phone, even less so than LLMs.So you’re either not fully informed about DLSS 5 or arguing in bad faith.
The point is addressing the cost of training the DLSS5 model.
The ‘models trained in giant data centers that devour energy and water’ are LLMs. DLSS5 isn’t an LLM, nor does it have anywhere near the training costs. DLSS5’s training costs are a rounding error when compared to LLMs.
However, DLSS 5 is not just upscaling anymore, it’s fully analyzing and generatively replacing parts of the image (lighting, material properties, …)
I know what Neural Rendering is. Calling it upscaling or generative upscaling doesn’t change my point about training.
And image based gen AI definitely cannot be trained on a phone, even less so than LLMs.
I know some CS professors that would be surprised to learn this.
I’m looking at the result of a lab where I completed training on 4x4 MLP denoiser on a TI-Nspire CX II… a graphing calculator with 64MB RAM and a 369MHZ processor. About 300k FLOPS/step, takes about an hour.
Yes, training a larger model on a phone would take longer but as long as a reasonable sized chunk of the model could fit into RAM it can be trained, eventually.
They’re not trained on phones as a rule because you can throw a lot more compute at them with dedicated hardware. You can train an LLM on an abacus if you’re patient enough.
That’s great marketing propaganda but you’re still supporting the shitty planet killing AI obsession. Keep bullshitting yourself if it makes you feel better but don’t try to bullshit me in the process.
That’s great marketing propaganda
No, it’s an easily verifiable fact. I’m not supporting anything. I’m saying that we should argue against it with arguments that actually apply to the situation, such as DLSS5 destroying artistic intent, having an unreasonable performance cost etc.
“Omg don’t try to bullshit me by stating objective facts that go against my opinions REEE”
Tbf the constant drive for more powerful pc components is one of the causes for planet destruction, though it’s not a leading cause and AI DLSS bs is not the only reason for this drive.
It’s not even an upgrade to me.
No. shut up. That’s not how this works. That’s not how any of this works.
There’s a million reasons to hate AI and environmental reasons are absolutely one of them. Doesn’t apply here.
Edit: I’m being hyperbolic, you don’t actually need to shut up. But Jesus Christ, I’m just really sick of really, really bad anti-AI arguments there are so many perfectly good ones.
I agree. I wish people who feel passionate enough to post also took the time to educate themselves. The conversation quality level around these subjects is dogwater.
No. shut up.
No u
edit: nice edit but you’re still an asshole. blocked.
My pleasure, truly.
who even wants this
Idiots
people looking for a new way tp experience old games
I don’t want a makeup Snapchat filter on my games.
For the 8 weirdos that want yassified leon? Shit sucks
