I think you rather missed my point, the fact that it even runs on my unsupported last gen mid range card basically using hack upon hack and it's "acceptable" is quite impressive.
Yes, but that's not what huge majority will see - they don't "hack" things, they press a button in menu. And so far, for such people, DLSS5 is in one game, supported only on 5k series GPU, and the game frankly has ugly textures etc. with and without that tech. So, not a good use case and won't conquer anything.
PC gaming has always been about that compromise between fidelity and FPS and, as you say, it is subjective. I certainly wouldn't want to use it in a fast paced online shooter but it's perfectly fine for a slower paced RPG.
Modders can do wonders with existing tech. But majority of PC gamers will never see that - they either get mainstream stuff or it might as well not exist.
If you don't like it or are not willing to make those compromises that's fine don't, there are plenty of people who are quite happy with it though.
I am commenting about the official, real use of those things, not modding. I never stand against modding or personal preferences - it seems this gets easily conflated by many people.
Kinda contradicting yourself here, why would you want to use it in photo mode if it's not an improvement in IQ?
Photo mode =/= gaming. No contradiction.
I've used it in quite a few games and it deffo improves the IQ, is it an AI filter? Personally I don't give a **** what it is, it makes my games look better
Very subjective, good for you. I don't see it as an improvement in GAMING, I see it as a detraction. You know what's improvement in comparison? For example, what AMD just shown recently, with tetrahedral BVH "cheat" to speed up RT rendering from 300ms for very poly-heavy scenes (like with Nanite) to 3.3ms per frame. THAT is an improvement, as 3.3ms with half a trillion of polygons on the screen is an amazing achievement on current hardware, leaving plenty of ms to use for other graphical improvements. And then you get AI filter, 10ms+ eating of frametime with up-close improvements of faces and then a lot of issues in movement and with camera being further away (as HUB shown in details), along with not really improving textures and other things much, but hallucinating lighting (which people consider somehow more realistic, even though it's just screen-space hallucination). The main thing I disliked RT for in games is that it's just slow one existing hardware and often just not worth it. However, with methods and improvements like AMD's mentioned one, it could become very viable and actually provide proper lighting, shadows, work well with high-poly scenes, faces, textures etc. An actual proper rendering, not some AI filter rubbish.
If the same effect was reproduced without it using AI would you be less against it?
It's not about the effect; it's about the performance hit. If it drops to 3ms per frame, all power to them. With current official state, it's unusable IMHO.
There would be many if you engage with Tinek tbf. Walls of text
As I said a few times - I am really holding back
Maybe I could ask AI to summarise (and hallucinate) stuff for me.
I'm becoming more and more convinced this isn't anything specific to Nvidia but rather a new DirectX standard for all vendors.
Well, yes, it's just a simple AI filter, a post-processing, it should run on anything. The main issue is just the performance, but now vendors could come up with their own custom models, perhaps devs could even train some on THEIR specific game and use that instead, with much better result than the generic NVIDIA's model. There are possibilities, just again - performance hit is... meh. It's a show-stopper for majority of gamers IMHO.
And its off the back of Microsoft's work with AMD on Project Helix, a couple of years ago AMD made some noise about wanting to make the hardware for all these different platforms the same thing, currently despite Microsoft and Sony running AMD SoC's they are not the same, the APU in the PS5 is different to the APU in the Xbox which is different to PC GPU's, (Its how the PS5 with less cores than the Xbox is faster) they want everything to be the same so game developers are working with the same tools across all platforms.
Ye, well, this is mostly because each vendor orders different spec designed for their needs and plans, not what AMD would like them to be. And AMD is just a contractor here - they made things per the console vendor spec, custom. I don't see that changing anytime soon. On the other hand, AMD has been working MUCH closer with Sony than with MS in designing new chips and pushing forth improvements in such small home devices. MS mostly grabs what comes out of that and makes some changes to cut cost down, it seems.
What specifically did you say? if you repeated what the other poster said that DX12 is an AMD derived API then a 3 second ask of chatgpt says its false...
I wonder when will people finally realise ChatGPT and other models were trained (amongst other things) on Reddit posts. As in, whatever idiocy one can find on Reddit, those LLM will spit it out at you. Even for work I tend to specifically ask LLM to skip ALL the Reddit posts in research, just all of them. And only then I finally get answers, instead of wild speculations and pure idiocy. And to be clear, I am not talking about any specific thing in your response, just in general - people should not take what LLM tell you as a gospel, it regurgitates the Internet as a whole, and if one has been online for longer than an hour in their life, one should realise it's mostly rubbish with just rare diamond here and there - it requires a lot of sifting around to get proper info. Exactly same thing with LLM responses, most are sadly just rubbish.
An article where the author uses sensationalist headlines takes the position that AMD pretty much was responsible for DX12 of which some of the main article seems to be behind a sub so some of the context is missing and another that seems more balanced where AMD played a major part in DX12 and will fully support and collab with it proving its commitment with mantle development.
Pressing the AI on my previous post with a few more questions and asking for its sources and some of them are amd.com arstechnica.com, anandtech.com, microsoft.com etc the AI gave a pretty broad overview on the question asked and seems to me a reasonable answer given the time dedicated to it.
The main thing isn't who did what here, it's that DX12 as such turned out to be sadly a failure. And not AMD or Microsoft fault here - most of the best parts of DX12 require good programmer/developer to use them well, knowing how hardware works and how to properly write for said hardware. And it turned out, modern devs have NO clue most of the time, they just switch a function on/off in engine like UE5 and that's it. They can do high-level stuff, but to know anything about how CPU or GPU actually work on lower level to be able to write their code properly is.. just not happening. And then they argue back online that these days nobody needs to know such things as "compiler takes care of it, I don't need to know it!" (hundreds of such posts by devs online these days!). This is why MS is working now on redoing parts of DX12 to make it more... idiot-proof, easier to use, but likely less powerful in what one can actually achieve.
videocardz.com
