-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
Fidelity Super Resolution in 2021
- Thread starter Grim5
- Start date
Soldato
- Joined
- 22 Nov 2018
- Posts
- 2,744
Yeah, and it's obvious they did so to show how good the 20 series performs. It made the benchmarks look good.
Well I played part of Metro Exodus with RT enabled on my 1080Ti FTW3. I also tried Quake 2 RTX and Minecraft RTX. The 20 series was the first to offer hardware RT support, but like many I chose to skip due to its poor performance. I don't remember site benchmarking Pascal with RT enabled.
The fact remains. Nvidia added software support for RT with Pascal, which debunks the idea that Nvidia never adds support for new features to older cards.
Except we have public statements by AMD, full resolution uncompressed images released by AMD, and facts about DLSS snd the state of the art in image reconstruction and temporal super resolution.
No one has commented about anything we don't know, at least not without big caveats.
Anything you think is being discussed without evidence or facts?
The problem is you need AA or you get jaggied and ppwpixel flickering. For most modern gsme engines, the only way to do AA is TAA (not by choice, but this is literally all you can do).
do your question doesn't make sense
And this is the real point. You will have to have TAA in any spatial-only super resolution technique. This id why DLSS 2 woekd so well, because it is better than TAA.
But as a quick rebuttal. DLSS can run st native resolution and upscale from there, before a final downample before display. This only makes sense on old games when you sre not looking for performance but extra visual fidelity. But non one bothers backporting DLSS to old games
Well let's hope it does a better job than TAA and is widely adopted because Days Gone has TAA because it is a console port. TAA is used as a crutch and I cringe when people say Days gone is well optimised because TAA is the spawn of the devil and needs to die
The problem is , if FSR is spatial only then it used the output of the underlying TAA and magnies the TAA image with its flaws. So FSR will require a really good TAA implementation to work well. DLSs replaces TAA early in the rendering pipeline
You quoted my post with a lol emoji. Is that because you believe that everyone should upgrade their GPUs everytime a new one comes out?
.
It was due to your suggestion that 5 years wasn't as long as 5 years because.
Soldato
and twitter comments and facebook posts and you tube videos. Intel release comments about its hardware as does nvidia - which *stunningly* turn out not to be overly true in methodology either. Actual documentation about process ratehr than simply guessing
Associate
maybe you should also explain the difference between theory and law.
of topic but the new dark matter maps are suggesting again that Einstein might be wrong.
in any case theory is the wrong word its hypothetically that he meant
and yes you are right about the hypothesis.. im just a bit lazy
Last edited:
And i was accused of assigning a different meaning to what people write.
Allow me to break it down for you, since you are clearly struggling. Nowhere in my post did i claim that "5 years wasn't as long as 5 years". That's just some nonsense you made up in your head.
What i did do, was bring context to that 5 year time frame in terms of GPU generations. I pointed out that there has only been 2 GPU generation released in that 5 year time frame.
But I think you do know what i was trying to say, you just didn't expect to get challenged on your response and made up some nonsense to cover your arse.
Don't roll your eyes too hard they might pop out.
of topic but the new dark matter maps are suggesting again that Einstein might be wrong.
in any case theory is the wrong word its hypothetically that he meant
Was Newton wrong? Or was his theory of gravity just incomplete and relatively basic, as we simply gained much more understanding over time? We already know there's much more to discover and learn about gravity and quantum physics, but if our current equations work (and they do) they aren't wrong now and won't be wrong in the future (as they'll work just as well). However, they might be at worst incomplete and describe only part of something bigger. It's a big difference.
In general (I assume you know but many people do not) science isn't about ditching/deleting past knowledge, as that would be very silly. Science discards what doesn't work and keeps only what works (scientific theory). And then, over time, science builds upon it, adding more and more understanding, refining math, expanding theories, joining smaller ones into bigger ones etc. In other words, it's building on top of older knowledge instead of replacing it fully, unlike what common person imagines. Hence, Newton was right in what he knew, his equations work just fine still, on Earth, as they always have. And the same will always be with Einstein's equations.
Last edited:
Ah, so 5 years sounds like a long time until it doesn't
Well we had Pascal, Turing and now Ampere. Sounds like a long time, 5 years Ah, so 5 years sounds like a long time until it doesn't. Well we had Pascal, Turing and now Ampere. Sounds like a long time, 5 years
All of that comes down to one point - time means NOTHING if there are no meaningful changes that come with it. Why would people with a relatively new Intel CPU (few generations old) need to upgrade at all? They would pretty much get no significant FPS increase in games if they just upgraded few gen old Core i5 to newest one, or Core i7 to newest one. Big cost to do it, for minimal gains. Same with Pascal->Turing. Ampere is better but it's also an unobtanium for most. Hence, my question is - what's your point again?
Associate
- Joined
- 26 Dec 2020
- Posts
- 156
But that's the point, they are not pointing all strengths and weaknesses of a brand in proportionality, or in other words, they are not taking the whole product into account. If they were just doing a review of Ray Tracing as technology then that would be one thing but they exclusively focus on one point to the detriment of others and that does make them a biased outlet. Add to that their exclusive promotion for Nvidia and you know what you are getting into. If GN and HW were promoting AMD Tressfx or CAS without any context around it they would rightly be called biased.
Also, I did say that DF's analysis of things in isolation is fine so your point about GN saying DLSS is better and therefore they are also biased is a red herring or perhaps you failed you understand my point. You just have to look at the thumbnails of 3080ti reviews from these three outlets to understand the difference between them. Remember that these days people cannot fudge data and stay hidden for a long time. So even a biased outlet has to adhere to some standards (which DF does) and have to work around it to promote something. It's a shame, they were one the best and trusted GPU technology review outlets at one time.
Associate
- Joined
- 26 Dec 2020
- Posts
- 156
I think you are defending things that people are not attacking. DLSS is good for what it is. For mid-range and entry-level cards, it's great (although 1080p DLSS is not that great according to many). But to claim DLSS as the Holy Grail of GPU technology in terms of performance and quality is wrong, not that you are doing it but some are.
Your example of Death Stranding is perhaps closer to what Minecraft RTX is i.e. too little scene complexity to show the difference between DLSS and native. Checkerboard rendering is also a tech that has proven itself for years on consoles yet it is still understood by most of the enthusiast PC community as a low-quality performance boost trick. Also, the Consensus in this thread seems to be that DLSS can be better than native with AA. Without AA, native is better.