• 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 5.0

MikeTrike said:
Or they had time to pivot from the backlash, reforming their strategy.
Click to expand...
https://investor.nvidia.com/news/press-release-details/2026/NVIDIA-DLSS-5-Delivers-AI-Powered-Breakthrough-in-Visual-Fidelity-for-Games/default.aspx#:~:text=DLSS 5 then uses its deep understanding,applied to maintain each game's unique aesthetic.
DLSS 5 provides game developers with detailed controls for intensity, color grading and masking, so artists can determine where and how enhancements are applied to maintain each game’s unique aesthetic.

that was on the first march release, nothing I saw in siggraph seem new versus what the API was rumoured to be announcement week, it was not fully public but it leaked a bit from the people that had access

https://www.tomsguide.com/computing...pc-gaming-breakthrough-or-nvidias-ai-slop-era (closed to public backroom q&a and extra demo)
, but this is all completely in the hands of devs to use however they wish.
There’s changes being made to the cinematic lighting and material depth too, which all contribute to another breakthrough, but one that developers are going to have to figure out the right way to use. Nvidia confirms this is an early build of DLSS 5, and probably with sliders turned all the way to max. So between now and the fall, devs will tune it however they wish.
 
Last edited:
That was fast.

when-you-criticizes-nvidia-v0-4vca2udlvrag1.jpeg
 
MikeTrike said:
That was fast.

View attachment 816200
Click to expand...
No point in commenting anything negative about nvidia on here, we're talking about 10+ year loyalists who've had maybe one or two criticisms over that period. I just stick to other topics or ignore them.

The implementation they're showing here is interesting, but it's also more dev overhead and not creating a simpler solution. It also won't work on AMD which means it wont work on consoles, so there's that. RT was a time savings or at least that's the claim. This seems like it's another tool you have to layer on top of an existing techniques. It's something you need to manually adjust and tweak on a asset basis. It's interesting, but it seems like belongs in the RTX Remix Suite and doesn't have anything to do with DLSS.
 
Last edited:
kram182 said:
AMD will have their own version - and I'm sure you'll all love it then - as we saw with upscaling, framegen, etc etc

View attachment 816247
Click to expand...

I wont, don't care for it at all. Why I play at 1440P, no need to upscale or use frame gen. Don't particularly care for either ones implementation of it and I would rather turn off ray tracing than use it, if the performance hit is too high. Very few games make excellent use of Ray Tracing for it to be worth the performance sacrifice.
 
Gideon said:
I wont, don't care for it at all. Why I play at 1440P, no need to upscale or use frame gen. Don't particularly care for either ones implementation of it and I would rather turn off ray tracing than use it, if the performance hit is too high. Very few games make excellent use of Ray Tracing for it to be worth the performance sacrifice.
Click to expand...

Still won't stop the tech from coming into existence from either side nor stop it from being implemented in games for those who do want it - for all the people like yourself who said they'll never use ray tracing/framegen/upscaling etc those still persist while still becoming ever more prevalent to even required in games. No reason to think this will be or go any different. Just turn the toggle off to all those who don't like it - while you still can at least.

Destiny.gif
 
kram182 said:
AMD will have their own version - and I'm sure you'll all love it then - as we saw with upscaling, framegen, etc etc

View attachment 816247
Click to expand...
I have no problem with new tech, it's just not upscaling tech like DLSS. The initial showcase I hated. What i just watched was much more interesting and useful. FSR and DLSS i like, i even liked FSR 3.1 quality at 4k on my couch. It was okay enough in Ghost of Tsushima when i was sitting back a few feet. DLSS is better. I'm kind of a double hater at this point though, because i know they don't want us as customers anymore. AMD makes it more obvious we're an irritation to their AI focus though.
 
are closed said:
To me, DLSS is simply a tool. If it lets me have playable framerate at desired settings, it gets enabled.
Click to expand...
That's the correct way to look at all of the DLSS tech. Each part of it (I really wish they didn't call everything DLSS) does something and if you find it useful in a given game, use it. DLSS super resolution, the AA/upscaling I use in basically every game. It is free frames at quality more or less. DLSS frame generation, that depends. I've had both good and bad experiences with it. In Cyberpunk 2077, it is great. In Satisfactory, it is not. Also some games just flat out run fast enough to not need it, I'd rather run at 150fps native than 220fps frame gen'd.

This new tech? Well we'll just have to see. I'm very interested, what they've showed off look like some real nice improvements to me. More than the characters I'm impressed by the lighting and shadows in the world. To me it is making things look more real, more grounded. But it'll depend on how it looks in operation. I'm sure there will be per-game variance. As I said: The easy thing for a developer to do to "support" it would be pick a model, crank it the fuck up, and just apply it to everything. Low effort, but also might not look great. Other games will probably spend the time to use it carefully and artfully and those I expect will look great.

MikeTrike said:
That was fast.
Click to expand...
This is, by far, the silliest take. A straw man/genetic fallacy and the idea that if someone criticizes what a big company does, you aren't allowed to disagree with that criticism because it is a multi-trillion dollar company therefore you are some kind of chud for "defending" them.
 
MikeTrike said:

View: https://youtu.be/drQVF9eNVx8
Click to expand...

I rather developers, artist work on the game vice moving sliders for trivial differences which has a cost. A cherry-picked demonstration, no rapid movement, no performance cost info. Pointless. As for FG and Nvidia, use it when it does not degrade image quality significantly and increases smoothness enough to give a better overall gaming experience. For Doom TDA, for me it is pure crap, motion with severe fringing on objects, textures, add in Path tracing and I get inconsistent frame times making timing for fast actions out of sync. Not a better gaming experience. Gaming experience should be the overall goal not a number of FPS. Anyways in the video I see zero gaming experience improvement.
 
Sycraft said:
This is, by far, the silliest take. A straw man/genetic fallacy and the idea that if someone criticizes what a big company does, you aren't allowed to disagree with that criticism because it is a multi-trillion dollar company therefore you are some kind of chud for "defending" them.
Click to expand...
You are allowed. But I'm still amused because it's powerfully magnetic for white knighting.
 
I'm reserving judgement on DLSS 5 until I see some Indy games using it. What I'd really like to see from this tech is small studios being able to get a lot closer to AAA games on graphics. AAA game development has gotten hugely expensive to the point that companies that make them are playing it safe. I'd love it if a low budget indy game could look almost as good.
 
zandor said:
I'm reserving judgement on DLSS 5 until I see some Indy games using it. What I'd really like to see from this tech is small studios being able to get a lot closer to AAA games on graphics. AAA game development has gotten hugely expensive to the point that companies that make them are playing it safe. I'd love it if a low budget indy game could look almost as good.
Click to expand...
It'll be even nicer if it can be done with reasonable performance on a $500-$600 GPU as opposed to the original demonstration with $8,000 worth of GPUs. I think that 10x~ is the key on this one to make it useful to any reasonable degree.

I still hate that it's proprietary.
 
Sycraft said:
And either way, it is more logical fallacies (ad hominem this case) when you run around saying people are "white knighting" because we say "Many this looks cool and I think it could be good."
Click to expand...
I'm gonna stop before I get in trouble for leading the blind off a cliff.
 
MikeTrike said:
I'm gonna stop before I get in trouble for leading the blind off a cliff.
Click to expand...
By all means, if you think you are so clever "lead" on. To me it seems like you just like to obliquely insult people.

learners permit said:
I often think how boring life would be if we all had the same way of seeing things
Click to expand...
I don't mind disagreements, heck that's part of the reason for a forum like this is to discuss different views. If you look at the Steam Machine thread you can see that I feel it isn't a good system, ands tate why. However I also don't shit on the people buying it, I'm not going to tell someone not to buy a toy if they want it.

I just get annoyed when people go to personal attacks because they don't like your opinion, but lack a good counter argument.

zandor said:
I'm reserving judgement on DLSS 5 until I see some Indy games using it. What I'd really like to see from this tech is small studios being able to get a lot closer to AAA games on graphics. AAA game development has gotten hugely expensive to the point that companies that make them are playing it safe. I'd love it if a low budget indy game could look almost as good.
Click to expand...
Unfortunately, I don't know how much this, or any, technology will help with that. I mean one thing that better tech does, particularly things that deal with improving realistic lights and shadows is that an artist doesn't have to work as hard on a scene, things will look "right" if they are placed in the scene. That can help a lot but you still need artists to make things look good. The issue a lot of indies have is just not having the budget to hire enough artists.

So while this, like path tracing, looks to help make things look more realistic with less artist effort, it isn't magic. A beautiful scene is still going to require good assets and a lot of time to assemble them, and that is what it expensive.
 
The ignore list is still functional if I recall correctly. I use it occasionally as it allows me to state things the way I see them openly and pay attention to only reasonable comments. Just saying I do not frequent this forum to be annoyed by combative adolescence.
 
My problem is that developers would be overly reliant on the tech and become even more lazy like with DLSS and frame Gen. Why put extra work in making the game engine more efficient when people can just use Nvidia tech to make up the short comings in performance. Games will just look even more the same then they already do using UE 5.
 
vegeta535 said:
My problem is that developers would be overly reliant on the tech and become even more lazy like with DLSS and frame Gen. Why put extra work in making the game engine more efficient when people can just use Nvidia tech to make up the short comings in performance. Games will just look even more the same then they already do using UE 5.
Click to expand...
I mean, I'd day that is not a good way of looking at things for two reasons:

  1. You say they'll be "more lazy" as though you get better art through just wanting it bad enough. Good art takes time and effort, and thus money. Games are already getting stupidly expensive. We should want things that make them cheaper. It isn't like if programmers just "coded harder" or artists just "art'd harder" we'd have games that looked photo real and did so without requiring new technology.
  2. Many of the things they are doing, like simulating subsurface scattering and such, are just expensive to do computationally. To do it "correctly" is slow and there's just no way around that. This *IS* an optimization. One of the ways we optimize rendering is figuring out how we can take shortcuts that give us something that looks good without doing the full calculations to do it right.
I mean, if you know of a way that can make games look realistic and perform well and not require new hardware and techniques then by all means, please reveal it. However I think it is pretty unfair to call game developers lazy when it is one of the areas of tech where people work the hardest and have the biggest crunch times.
 
vegeta535 said:
My problem is that developers would be overly reliant on the tech and become even more lazy like with DLSS and frame Gen. Why put extra work in making the game engine more efficient when people can just use Nvidia tech to make up the short comings in performance. Games will just look even more the same then they already do using UE 5.
Click to expand...
The incentives for the producers, get game out earlier, get costs down and layoff the programmers/team earlier vice spending another 6 months to a year. Get income earlier, save costs $$$$. Too bad a potential great game comes out sucking instead for most part due to performance issues and game tanks because of it.
 
Sycraft said:
I mean, I'd day that is not a good way of looking at things for two reasons:

  1. You say they'll be "more lazy" as though you get better art through just wanting it bad enough. Good art takes time and effort, and thus money. Games are already getting stupidly expensive. We should want things that make them cheaper. It isn't like if programmers just "coded harder" or artists just "art'd harder" we'd have games that looked photo real and did so without requiring new technology.
  2. Many of the things they are doing, like simulating subsurface scattering and such, are just expensive to do computationally. To do it "correctly" is slow and there's just no way around that. This *IS* an optimization. One of the ways we optimize rendering is figuring out how we can take shortcuts that give us something that looks good without doing the full calculations to do it right.
I mean, if you know of a way that can make games look realistic and perform well and not require new hardware and techniques then by all means, please reveal it. However I think it is pretty unfair to call game developers lazy when it is one of the areas of tech where people work the hardest and have the biggest crunch times.
Click to expand...
This is tech we have to see in action and not a restricted demo. How does that subsurface scattering look while in motion? Fast motion? How does the lighting which is not being calculated per pixel in the rendering pipeline but by AI after the rendering? Performance impact? Plus it would be better if the game was developed from the start to have features desired vice tacking on outside sources to make up for short comings. Modern game engines already can do these types of things, not sure who this tech is for other than experiments. It is like Nvidia's another attempt at black box code for developers to get trapped by.
 
noko said:
This is tech we have to see in action and not a restricted demo. How does that subsurface scattering look while in motion? Fast motion? How does the lighting which is not being calculated per pixel in the rendering pipeline but by AI after the rendering? Performance impact? Plus it would be better if the game was developed from the start to have features desired vice tacking on outside sources to make up for short comings. Modern game engines already can do these types of things, not sure who this tech is for other than experiments. It is like Nvidia's another attempt at black box code for developers to get trapped by.
Click to expand...

Showing things like subsurface scattering/light shining through earlobes etc is hard to see when/if working even perfectly, in motion - pixel peeping/freeze framing is the right way to show stuff like that off - showing 'in motion' would be good just to show no 'shimmering'/motion vector artificing/etc is introduced like you can see sometimes with upscalers etc - but you wouldn't really be able to see 'what it's doing' unless you froze the frame then zoomed in and pixel peeped - same goes for a majority and crap ton of graphics settings/tech whether they're rasterized/AI generated/hand crafted etc - hence why in 'optimized/comparison settings' pages/videos you're usually looking at/shown screenshots or stationary screens when actually shown the settings being discussed/comparisons between graphics levels.
 
noko said:
This is tech we have to see in action and not a restricted demo. How does that subsurface scattering look while in motion? Fast motion? How does the lighting which is not being calculated per pixel in the rendering pipeline but by AI after the rendering? Performance impact? Plus it would be better if the game was developed from the start to have features desired vice tacking on outside sources to make up for short comings. Modern game engines already can do these types of things, not sure who this tech is for other than experiments. It is like Nvidia's another attempt at black box code for developers to get trapped by.
Click to expand...
For sure. I said I am interested in it, not that I am sure it'll be useful/transformative. I'll need to see it in action, on my system, before making any final judgements. However I am very interested at this point. It looks good, and it looks like something game designers can use to improve the quality of their games, if they choose. We'll have to see how it hold up in game, is it temporally stable (they claimed it was in the keynote), and what is the performance cost? If it tanks performance then it may be a non-starter, though path tracing tanks performance and I love that.

We'll see but I'm optimistic based on their track record. DLSS super sampling and frame gen have been great. They aren't perfect, and frame gen in particular is one that works better in some titles than others, but they have been very useful technologies. They also were things, particularly frame gen, that people hated on as soon as it was announced just because of what it was, without actually trying it. Heck I still see people angrily decrying "fake frames" as something bad that nobody should be allowed to like.
 
vegeta535 said:
My problem is that developers would be overly reliant on the tech and become even more lazy like with DLSS and frame Gen. Why put extra work in making the game engine more efficient when people can just use Nvidia tech to make up the short comings in performance. Games will just look even more the same then they already do using UE 5.
Click to expand...

Maybe but upscaling in video game became the norm with the ps3/xbox 360 generation and the effort/budget into making game exploded after upscaling became the norm versus before, not sure why AI super sampling would be different than every previous game engine performance tricks, they never lead to a reduced effort which is determined by your competitor effort more than anything else.

From anti aliasing instead of pushing resolution... too better culling algorithm, they usually do not use trick to put less effort and money into game.... but making bigger game instead.
 
LukeTbk said:
Maybe but upscaling in video game became the norm with the ps3/xbox 360 generation and the effort/budget into making game exploded after upscaling became the norm versus before, not sure why AI super sampling would be different than every previous game engine performance tricks, they never lead to a reduced effort which is determined by your competitor effort more than anything else.
Click to expand...
I think it may come from the mislead belief that developers are leaving a lot of performance on the table and games could be much faster, higher rez, etc if devs just spent more time "optimizing" and that literally the only reason they don't do that is they are lazy. That's not the case, of course, While there ARE games that have poor optimization, it isn't some unlimited well you can draw from and just make things better to an unlimited extent. Graphical effects cost performance, and there's only so much a system has.

As you say, upscaling had long been a thing, or more realistically, down rezing. Games just ran at a lower than native resolution and the "scaling" was just a bilinear filter. Rather blurry results. Things like DLSS just give us a better option than "turn down the resolution and deal with the blurry and/or pixelated image." It gives a better result to something we were already doing: Running games below the native resolution of the display.
 
I watched the entire DLSS5 segment of the SIGGRAPH presentation so I sort of get some of its inner workings and how it's different from generative AI because rather than just using text prompts, it uses graphical inputs from the game like motion vectors, RT/PT, geometry, etc. in order to have consistency and to stay in line with what the game is supposed to look like. It doesn't use a general model but rather a highly specialized/streamlined model that should, if Nvidia is to be believed, not have a crippling performance impact and memory footprint making it viable for real-time interactive graphics.

My takeaway is that if the artistic intent for the game is photo realism, and photo realism is hard and computationally expensive to do dynamically in real-time graphics, then DLSS 5 is a tool to help game developers achieve that without having to brute force or wait until the hardware is fast enough. The more I think about it, if photo realism in computer game graphics was so easy to do with raster graphics and existing graphics hardware, then games should be photo realistic by now. I mean even with film and tv shows, which uses off line rendering, it's still pretty easy to tell if something is CG.

I will reserve my opinion about it until I get to try it for myself but if it will be anything like DLSS super resolution, it may be crap at first but it probably has the potential to improve over time.
 
What they showed is something that has been done with generative models since early 2023, in terms of maintaining intent with img2img like ControlNet has done (eg: being able to feed it normal maps, depth maps, albedo, etc, to inform the model how to produce an output that adheres to them). Tbh that's always been where the more interesting gen AI output has been.

The feat is how they've managed to tune models to handle this in real-time (with temporal coherence). In terms of artistic intent they've already shown their DLSS 5 models taking some liberties with details (especially faces) since it has the same downsides of existing img2img where the model is still ultimately making choices about determining what the object/surface is and interpreting it. I could see it being used tastefully to enhance certain environment details in some games though.
 
Last edited:
Captain Newmackwa said:
if photo realism in computer game graphics was so easy to do with raster graphics and existing graphics hardware, then games should be photo realistic by now. I mean even with film and tv shows, which uses off line rendering, it's still pretty easy to tell if something is CG.
Click to expand...
I think you could overrate how much it is easy for most people, a post 2010 Scorsese, Fincher or Nolan movie is full of CGI that go totally unnoticed to me.

Say Wolf of Wall street cgi, which is old by now:

View: https://vimeo.com/83523133?fl=pl&fe=cm

View: https://www.youtube.com/watch?v=QChWIFi8fOY

what percentage of the audience watching this even on movie theater screen had the first clue ? it is 15 years old tech by now. (not just talking about the computer aided composition of real shots, which by then was near perfect but that actual added CGI in them has well).

Has for photo realism that what the 3 models have been trained on, but I imagine you could use something very similar to build a model with a different intent-astheatic than photo-realism would you want, easy to imagine an "anime" or cell shaded, oil painting astheatic end result, they have 3 models now but it could quickly become 12 and big studio making their own
 
noko said:
This is tech we have to see in action and not a restricted demo. How does that subsurface scattering look while in motion? Fast motion? How does the lighting which is not being calculated per pixel in the rendering pipeline but by AI after the rendering? Performance impact? Plus it would be better if the game was developed from the start to have features desired vice tacking on outside sources to make up for short comings. Modern game engines already can do these types of things, not sure who this tech is for other than experiments. It is like Nvidia's another attempt at black box code for developers to get trapped by.
Click to expand...
Agreed. It would be generous to call it mediocre going off what they've shown up to now. In this most recent video, the butterflies (or whatever) in the first scene blinked out of existence at some points and the faces barely moved while constantly being obscured by highlighting the affected area.
 
You must log in or register to reply here.
Back
Top