They said consoles and wide range of radeon gpus whatever that means. It would be nice to have toggle button inside the software instead of waiting and hoping developers will add it to your game like with nvidia.

I know it's easy to say but this should have been ready for the RDNA 2.0 start. So now those owners still have to wait a lot to get something similar to dlss.

Looking forward to this tech. If they can make it available across the board, that will be fantastic.

Maybe it could even be open source like they did with FreeSync.

RDNA3 will be the complete upgrade. Nvidia will follow Intel's fate. Rdna was the stepping stone, rdna2 is the beta testing of success. You all saw what happened with zen, zen2 and zen3 ryzedominance.

This should have really been launched with the vapour launch of the 6800/6900XT. AMD really dropped the ball. And for the record I don't believe a single word Herkleman says, I don't know what he thinks his job is supposed to be, but I've seen him do nothing but spread lies to the point that he's hurt AMD's reputation.

itpro:

RDNA3 will be the complete upgrade. Nvidia will follow Intel's fate. Rdna was the stepping stone, rdna2 is the beta testing of success. You all saw what happened with zen, zen2 and zen3 ryzedominance.

I am a big supporter of AMD but I believe that Nvidia has both the money and the minds to give the world a new architecture, the competition is good for buyers personally I do not want AMD in any case monopoly I hope and wish Intel on the graphics cards is going well (for the prices to fall) 🙂

Can't wait to give my honest review of FidelityFX Super Resolution. You know, like AMD sycophants are continually doing in DLSS threads - never seeing it in action or even owning the required hardware. Unlike them I'll be using objective IQ comparison methods. See ya in a year or so 🙂

itpro:

RDNA3 will be the complete upgrade. Nvidia will follow Intel's fate. Rdna was the stepping stone, rdna2 is the beta testing of success. You all saw what happened with zen, zen2 and zen3 ryzedominance.

Intel stagnated. They are still releasing CPUs on a decades old process. Nvidia is not Intel.

You don’t need machine learning to do it, you can do this many different ways and we are evaluating many different ways.

If they are still evaluating the approach, it's probably not coming this year.. especially given that it's been 6 months or so since they announced it.

Denial:

If they are still evaluating the approach, it's probably not coming this year.. especially given that it's been 6 months or so since they announced it.

Worse. If they are giving up / are uncertain about ML approach, means they are running into walls. And using anything else than ML would truly equate to just another upscaler with no hope of achieving DLSS-like IQ. But how do yo go around the fact that they are already using their general cores for RT (already lagging behind Nvidia's dedicated hw) and putting ML upscaling on top of that would reduce performance even further. Its like this: IQ wise: ML or DIE Perf wise: ML and DIE

I hope this tech slips down to my Vega 64. I would at least like to fiddle with it. I am not optimistic on either count. Still, I interested to see what they end up with.

I doubt it will be anything meaningful

itpro:

rdna2 is the beta testing

true

Neo Cyrus:

This should have really been launched with the vapour launch of the 6800/6900XT. AMD really dropped the ball.

it'll have been a year after the launch that 6800/6900 owners get to use RT at playable framerates

Devid:

I know it's easy to say but this should have been ready for the RDNA 2.0 start. So now those owners still have to wait a lot to get something similar to dlss.

By that logic, that's like saying we should have 5nm GPUs and quantum processors in every PC by now. Complex things don't just magically happen overnight. How simple do you think it is to develop an AI that yields compelling results in realtime? Nvidia had years to work on DLSS without being pressured to release it, and their v1.0 was so bad that it was used as a reason to dismiss the high price of the RTX 2000 series. AMD's Windows driver team isn't exactly known to be fast and proficient. I'd say they're making decent time for a product that they weren't prepared to make. All that being said, I have low hopes that this feature is going to what we're all expecting. Let's hope it isn't as limited in what it can work on, though.

schmidtbag:

By that logic, that's like saying we should have 5nm GPUs and quantum processors in every PC by now. Complex things don't just magically happen overnight. How simple do you think it is to develop an AI that yields compelling results in realtime? Nvidia had years to work on DLSS without being pressured to release it, and their v1.0 was so bad that it was used as a reason to dismiss the high price of the RTX 2000 series. AMD's Windows driver team isn't exactly known to be fast and proficient. I'd say they're making decent time for a product that they weren't prepared to make. All that being said, I have low hopes that this feature is going to what we're all expecting. Let's hope it isn't as limited in what it can work on, though.

I don't think that analogy works and I'm not sure that's the reason for AMD not having a DLSS competitor. If Intel was publically working on quantum computers since 2014, had a breakthrough that showed one working decently well in 2017, and in 2019 shipped a quantum computer.. all while in 2019 AMD is saying "We also can build a quantum computer using our current technology" then two years later said "We aren't sure we are even going to build quantum computers" that would be the analogy. I think at least a few people would question whether AMD invested in the wrong thing, or invested too late in the wrong thing, etc. Either way I don't buy the idea that they couldn't have had a solution by now. They had plenty of time - they just invested in other places, which is fine but it might be a mistake depending on what a potential DLSS 3.0 and/or other AI enhancements bring.

sounds like the real DLSS answer is not coming this year and this is just a stopgap solution for pc but hey I'm sure forza horizon 5 and gears 6 and halo infinite will run twice as fast on AMD while nvidia cant even toggle RT

Denial:

I don't think that analogy works and I'm not sure that's the reason for AMD not having a DLSS competitor. If Intel was publically working on quantum computers since 2014, had a breakthrough that showed one working decently well in 2017, and in 2019 shipped a quantum computer.. all while in 2019 AMD is saying "We also can build a quantum computer using our current technology" then two years later said "We aren't sure we are even going to build quantum computers" that would be the analogy. I think at least a few people would question whether AMD invested in the wrong thing, or invested too late in the wrong thing, etc.

You're reading too far into what I said. The point of my analogy was to say that it's ignorant to suggest something so complex can be achieved in such a short amount of time. A DLSS competitor may be kindergarten complexity compared to quantum computers, but we're also only talking a relatively short period of development time. A good competitor isn't going to suddenly be available just because Nvidia already has one ready-to-go.

Either way I don't buy the idea that they couldn't have had a solution by now. They had plenty of time - they just invested in other places, which is fine but it might be a mistake depending on what a potential DLSS 3.0 and/or other AI enhancements bring.

Really now? The first version was released in early 2019, and was so limited and unimpressive that it just simply wasn't a threat to AMD at all. I forget what it was called but AMD had a competing feature that basically accomplished the same thing but worked on more games. So, for most of 2019, AMD didn't have a compelling reason to make something else. The v2.0 as we know it today isn't even 1 year old yet. DLSS has been in development since 2017, and only in early 2020 was it actually worth using. This implies that AMD would have got started working on their competing product no earlier than Q2 of 2020, and that's assuming all the management and other overhead didn't slow down development from beginning. Sure, Nvidia basically exemplified to AMD that it can be done, but obviously they're not going to share how exactly they did it. And as you said, Nvidia isn't done making improvements. So you mean to tell me it is perfectly reasonable to expect AMD (y'know, the company that is slow to release optimized Windows drivers) to release a product worth using in a little over 1 year, compared to Nvidia's 4? Let's not forget they have to release this for consoles too, which aren't identical to desktop GPUs.

Denial:

Intel stagnated. They are still releasing CPUs on a decades old process. Nvidia is not Intel. If they are still evaluating the approach, it's probably not coming this year.. especially given that it's been 6 months or so since they announced it.

Nvidia plays all money at specialized cores for extra work and features. Imagine being bland like Radeon. They would lose, since ages. If Radeon improves and they're willing to take the initiative, we will see an unprecedented change.

I doubt AMD will be able to do anything non AI accelerated on par with DLSS, which is hardly perfect and has lots of problems. And the last thing we need is more artifacts from bad upscaling mixed with a TAA hybrid and it's artifacts, not to mention excusing poorly optimized games with "Oh just run it with DLSS/AMD equivalent." Which is already becoming a problem with current games supporting DLSS. (FFXV,Watch Dogs,Cyberpunk,Control,System Shock Remake which the previously released demos already performed very badly for how basic the game looked,etc.) Last I checked AMD's DSR equivalent is still highly restrictive and you can't add custom resolutions unlike with DSR. AMD has always been behind in AA (Aside from the one DX11 demo they made with their own equivalent of SGSSAA that looked fantastic. Yet it was never used in any games, nor was it put into the drivers with any method to inject it into games). And I doubt that will ever change.

schmidtbag:

You're reading too far into what I said. The point of my analogy was to say that it's ignorant to suggest something so complex can be achieved in such a short amount of time. A DLSS competitor may be kindergarten complexity compared to quantum computers, but we're also only talking a relatively short period of development time. A good competitor isn't going to suddenly be available just because Nvidia already has one ready-to-go. Really now? The first version was released in early 2019, and was so limited and unimpressive that it just simply wasn't a threat to AMD at all. I forget what it was called but AMD had a competing feature that basically accomplished the same thing but worked on more games. So, for most of 2019, AMD didn't have a compelling reason to make something else. The v2.0 as we know it today isn't even 1 year old yet. DLSS has been in development since 2017, and only in early 2020 was it actually worth using. This implies that AMD would have got started working on their competing product no earlier than Q2 of 2020, and that's assuming all the management and other overhead didn't slow down development from beginning. Sure, Nvidia basically exemplified to AMD that it can be done, but obviously they're not going to share how exactly they did it. And as you said, Nvidia isn't done making improvements. So you mean to tell me it is perfectly reasonable to expect AMD (y'know, the company that is slow to release optimized Windows drivers) to release a product worth using in a little over 1 year, compared to Nvidia's 4? Let's not forget they have to release this for consoles too, which aren't identical to desktop GPUs.

All I'm saying is that this was telegraphed for half a decade - but especially in 2017 when Nvidia literally showed the foundation of DLSS at Siggraph... but now, it's 4 years later and AMD still hasn't decided if it's even going to use AI in the graphics pipeline. Also FWIW, AMD itself said they were already looking into a DLSS alternative in January 2019 utilizing DirectML. So I don't know how you can say they would have started in Q2 2020.