Didn't you just get DLSS confused for native 9 posts back?
Nvidia DLSS 1 and 2 antialiasing discussion *spawn*
- Thread starter DavidGraham
- Start date
-
- Tags
- nvidia
Didn't you just get DLSS confused for native 9 posts back?
Yes, and? Worst oversharpening ever on TAA being mixed with DLSS doesn't change anything regarding that.
It's only about objective graphs with same settings for everything, not subjective "it looks about the same" random settings per card.
There is no point debating subjective matters. This discussion always goes same route. I hate dlss, I rather lower settings, I rather this and that. Let those people and reviewers recommend dlss who find it useful and pleasurable. There is no perfect realtime graphics. Trying to argue from theoretical pov just doesn't match real life experience when playing real life games. Some games have really good dlss, some don't. Some people find dlss artifacts off putting, some people find dlss graphics better than native. It doesn't help that discussion focuses on zoomed in stills with carefully chosen locations showing error or better than native quality. The 99% of gaming experience where there is no trailing birds or moire is lost when focusing on the trees instead of the forest.
It's as silly as forcing everyone to listen britney spears. Some people love it, some don't. And even those who hate might love one of her songs and some who love her don't like all of her songs. Don't know if britney is dlss or not.
It's as silly as forcing everyone to listen britney spears. Some people love it, some don't. And even those who hate might love one of her songs and some who love her don't like all of her songs. Don't know if britney is dlss or not.
Your comparison is flawed. It would be same as putting Britney on top of sales charts because you happen to prefer it, despite it not selling as well as other artists. Subjective vs Objective. I like it > Sales vs Sales = Sales. I think it looks the same > Same res/settings vs Same settings for everyone
Good job!I rather have native over dlss , I rather dlss over bad frame rates , I rather up scaling over bad frame rates.
The driving force of me buying video cards for a good decade was FSAA.
I have a 3080 coming in so i will be able to comment on dlss based on my experiences soon
The driving force of me buying video cards for a good decade was FSAA.
I have a 3080 coming in so i will be able to comment on dlss based on my experiences soon
So not always worse than native then?
Realistically we aren't getting any huge fps boost anymore from process technology. We need smartness over brute force or graphics will stagnate. DLSS like algorithms is one way to combat this. Chip in 1080ti is about 4 years old and 3080 is barely 2x faster. Like it or not, brute force is not going take us forward. If someone wants native pixels they are in a world of hurt if they go look how much compromises and hacks and less than full native resolution rendering game engines already do. Native doesn't really even exist if you break it down to rendered elements.
Do we really want to be in place where 4 years from now we play same games with double the framerate or do we want smarter things to happen to get better scaling? That better scaling can happen with existing hw already,... Ray tracing is one of those smarter things to increase quality by allowing realtime gi, better shadows, better reflections in essence allow accurate simulation of light. Ray tracing being new technology can see good scaling as there likely is more room to optimize hw&sw implementation than there is fat in rasterization/compute side. DLSS like algorithms is another avenue for better scaling.
edit. Hopefully unreal5 is one of those smart things that allows scaling beyond what flops suggest. DirectStorage also might help by removing memory limitations.
Do we really want to be in place where 4 years from now we play same games with double the framerate or do we want smarter things to happen to get better scaling? That better scaling can happen with existing hw already,... Ray tracing is one of those smarter things to increase quality by allowing realtime gi, better shadows, better reflections in essence allow accurate simulation of light. Ray tracing being new technology can see good scaling as there likely is more room to optimize hw&sw implementation than there is fat in rasterization/compute side. DLSS like algorithms is another avenue for better scaling.
edit. Hopefully unreal5 is one of those smart things that allows scaling beyond what flops suggest. DirectStorage also might help by removing memory limitations.
Last edited:
You actually managed to buy one? Lucky you.
zotac drops them every weekday morning if you want the specifics of how to grab one pm me
We can hope RT scales well. we will have to see how they do in future generations because the scaling doesn't seem great between the two nvidia generations
Objectively always worse than native, subjectively it can be better, just like I said already several times?
Reviews are supposed to be objective, even playing field for everyone tested with same settings. Not subjective where some random person decides what settings look similar enough to be put in same graphs.
How do you define objectively better? What's the basis for concluding this?
So does supersampling "objectively always better" than native then?
It's what the devs actually intended the game to output at given settings.
SS is essentially just higher resolution native scaled down, it can be considered better (as downsampling doesn't suffer from same issues as upsampling) and obviously shouldn't be mixed in same graph with results not using Supersampling.
So it doesn't matter at all how it is scaled down? Point sampling would be better than native too?
And what if it's somehow resulting in a higher performance?
You’ve mentioned this a few times. Are there examples of reviewers comparing DLSS 4K vs native 4K and pretending they’re the same thing? I’ve never seen that.
Of course results can vary based on method, "can be" != "is". One needs to accept whatever method the dev chooses to be their method of choice to be the optimal.
But regardless of anything else, it like any other such feature should only be used when it's used on everything. Same settings for everyone, not different subjectively worse/better/similar settings for anyone.
Not sure if I have seen any reviewers do it, but I've seen several people suggest they should (not in this particular branch of discussion or even forum in all cases).
I'm sure any dev would agree that their game will look better with 64x supersampling which is exactly the target DLSS is aiming for. The algorithm may not achieve that but depending on how well it performs in a given title, it's perfectly feasible that it may achieve closer to that target that the native resolution does. Therefore I don't see how you can claim native is objectively always better, when objectively it's possible for the DLSS image to be both more detailed and less aliased than the native image. And certainly Digital Foundry in some have their analysis have highlighted instances where this is indeed the case.
So supersampled image can be worse than native? "Objectively" or not.
