Nvidia DLSS 1 and 2 antialiasing discussion *spawn*

I know it was briefly touched upon in the console forum, but I wouldn't be surprised if Nvida pushed for tensor cores in the upcoming switch specifically to increase developer use of DLSS.
DLSS is a very minor thing from developer perspective, it doesn't need to be pushed into any console to be widely used.
Other ML based stuff though could be really interesting. The only problem is that being a handheld device Switch 2 (or whatever will be its name) won't have much tensor flops anyway.
 
Not sure the point you're trying to make here. The PS5 objectively outperforms the 2060, as I demonstrated.

The "/s" in the post you're replying to was signifying sarcasm.

Well on paper PS5 outperforms a 2060 easily but practically it's far from it.

My 2060 laptop runs Control at way higher settings than PS5 and also has enabled RT reflections. It does so at a steady 60 FPS at 1440p thanks to DLSS Performance. From a raw performance perspective, you are certainly right that a PS5 will outperform the 2060 as it is running at native 1440p upscaled to 4K as it of course has a more powerful GPU in general. But I don't think it matters when the reconstruction of DLSS is so good. DLSS and the RT cores give PCs a big advantage here.

controlll5ejul.png
 
Last edited:
Impossible to do "ground truth" comparisons between consoles/PC when dynamic resolution og checkerboard upscaling are on by "default" on consoles...but seems like DLSS ruffled some feathers /shrug
 
Critics here couldnt guess what was dlss vs native in stills, let alone in motion.
I see not many reasons to not use it if possible, in special seeing the returns.

HZD uses reconstruction on the pro to achieve 4k, and that was never a problem whilst not even up to par with dlss2.0.
 
I sort of wonder if it's just the nvidia sharpening that's available in the control panel. The sharpening factor seems the same.
I'm sort of wondering that too since if it is then it's likely not ready for prime time here which is why it wasn't exposed in the UI.
 
Last edited:
Dynamic resolution in action:


Brilliant find! So as suspect by some it holds a fixed output resolution and simply varies the internal resolution and DLSS quality factor to achieve the desired framerate. Much the same as many modern games with VRS and upscaling do to be fair, just with much better upscaling. It's the future of resolution!
 
My understanding of DLSS is there's a big chunk of ML time spent analyzing tons of mega-resolution scene details from the game to generate the proper "upscaling / detail filling" strategy. This was the whole reason why games had to be designed specifically for DLSS versus it being some generic toggle for all/nothing.

So how does a new plug-in "work"? Where is all that ML time doing scene analysis spent?
 
My understanding of DLSS is there's a big chunk of ML time spent analyzing tons of mega-resolution scene details from the game to generate the proper "upscaling / detail filling" strategy. This was the whole reason why games had to be designed specifically for DLSS versus it being some generic toggle for all/nothing.

So how does a new plug-in "work"? Where is all that ML time doing scene analysis spent?
It's my understanding that all the work spent in DLSS 1 was to gradually train the systems until eventually Nvidia was able to provide enough inputs across many games to allow a general catch-all NN capable of dealing with almost all situations. Thus we now have DLSS 2.
 
Hmm, ok. Sounds like I missed some reading during the DLSS v2 transition. I know it got "better" however I hadn't heard about the general use case possibilities.

Which then makes me ask: Why isn't it now just a general tickbox, not unlike DSR?
 
Hmm, ok. Sounds like I missed some reading during the DLSS v2 transition. I know it got "better" however I hadn't heard about the general use case possibilities.

Which then makes me ask: Why isn't it now just a general tickbox, not unlike DSR?
It still has to be integrated into the game engine by the developer, it's not just a straight post process. There are motion vectors that must be passed to DLSS, the game's own post-process exclusions, overlay UIs that need to occur post DLSS etc. I'm sure there's a lot more to it than my understanding.
 
Back
Top