Thursday, January 11th 2024
AMD Believes NVIDIA is Behind in Driver-Based Upscaler Development
AMD is readying its Fluid Motion Frames (AFMF) technology for public release later this month (January 24 to be exact). Aaron Steinman, a Senior Radeon Manager, believes that arch rival NVIDIA will need to take some drastic steps once AFMF arrives due to its more open nature. He stated in a short interaction with PC Gamer: "I would be curious to know if NVIDIA feels now they have to match what we've done in making some of these solutions driver-based." His software engineering buddies have already released the Radeon Super Resolution (RSR) technology, which functions via in-driver operation.
Unlike Team Red's heavily marketed FidelityFX Super Resolution (FSR) system, AFMF and RSR are not reliant on official support from games developers. The driver-based solutions will be packaged within an upcoming version of AMD's HYPR-RX feature set. Steinman continued with his statement: "I think what we're gonna start seeing, DLSS is only available on certain solutions, so either NVIDIA is going to have to benefit from our solution because we did make it open-source and cross-vendor, or they're probably going to need to do something similar." The publication points out that Team Green has something in the same ballpark—NVIDIA Image Scaling—but its nowhere near as advanced as their headlining "AI-infused" DLSS tech. Steinman conceded to PC Gamer that his main opponent will inevitably pull ahead in the future: "I mean, the competition will never end, right? We'll have new technologies, they (NVIDIA) will have new technologies."
Sources:
PC Gamer, VideoCardz
Unlike Team Red's heavily marketed FidelityFX Super Resolution (FSR) system, AFMF and RSR are not reliant on official support from games developers. The driver-based solutions will be packaged within an upcoming version of AMD's HYPR-RX feature set. Steinman continued with his statement: "I think what we're gonna start seeing, DLSS is only available on certain solutions, so either NVIDIA is going to have to benefit from our solution because we did make it open-source and cross-vendor, or they're probably going to need to do something similar." The publication points out that Team Green has something in the same ballpark—NVIDIA Image Scaling—but its nowhere near as advanced as their headlining "AI-infused" DLSS tech. Steinman conceded to PC Gamer that his main opponent will inevitably pull ahead in the future: "I mean, the competition will never end, right? We'll have new technologies, they (NVIDIA) will have new technologies."
84 Comments on AMD Believes NVIDIA is Behind in Driver-Based Upscaler Development
Also FWIW, Native is not the best image quality, people really need to forget that antiquated notion. Supersampling for example easily exceeds rendering at a given panels native resolution. Sure you need to downsample to that exact native pixel count, but the pixels will have better colour information for a cleaner, more detailed, and more stable image. Native rendering is a good reference point, but it's not where image quality just stops improving.
As for Steinmans amusing jab, yeah somehow I don't think Nvidia are particularly worried by AFMF, and if they are, we'd hear about it in pretty short order But hey the competition is nice and it's refreshing to see AMD's turnaround time to compete in this space improving.
I hope AMD will be able to improve FSR, AFMF etc. to match Nvidia features or close. (Why is it called AFMF tho haha, should just be called FMF.......) FSR is nowhere near DLSS and DLAA in my testing, way more shimmering and artifacts with FSR.
These features are important for buyers and this is why Nvidia sells way more GPUs than AMD. Steam Hardware Survey shows that AMD dGPU marketshare keeps dropping quarter for quarter. Intel eats marketshare slowly too. If AMD don't react they will drop below 10% marketshare soon.
AMD have one entry in the top 25 which is "AMD Radeon Graphics" their mobile and APU GPU like Intels "Iris Xe Graphics" and "Intel(R) UHD Graphics"
First real AMD GPU is at spot 29, which is the 7 year old RX 580 which sold for around 200 bucks.
They need some good value/perf cards, like RX480/470/580/470 and 5700XT. And this is exactly what Radeon 8000 is going to be, hopefully. No high-end, just good mid-end value cards. Marketshare-eaters.
Just about every of their features are worse than the Nvidia counterpart. You don't get ahead by copying others but AMD should prioritize FSR and AFMF way higher than Ray Tracing performance because even most Nvidia people don't use RT anyway -> We use DLSS, DLAA, DLDSR, FG, Reflex etc. These are the killers features. Yes RT can be used just fine but the fps hit is still way too huge for most people to accept, especially people with high refresh rate monitors. We are simply not going to accept 40-80 fps gameplay with RT when we can get 120-240+ without.
AMD always like to act like the good guy and make their features open source, but RTX users are not going to use FSR if DLSS is an option.
AMD should start inventing some groundbreaking features themselves maybe. Again, you don't get ahead by copying and MCM for GPUs was somewhat of a failure so far.
We won't see a high-end Radeon 8000 offering, but they need to have something big to counter 5000 series, which launch in less than a year from now.
If AMD actually improves FSR to match DLSS/DLAA I could see many people considering AMD GPU again, myself included. However, right now, AMDs features are just too much behind. Raster performance and nothing else in 2024 is not enough. Upscaling is here to stay. Every single developer embraced it and how many AAA games in the last few years released without at least one upscaler? Pretty much none.
With Nvidia you can use DLAA to improve heavily on native. Use DLSS on the higher presets to pretty much match - or even slightly improve - on visuals while boosting performance by 50-75% or use the lower presets if you struggle with performance and just want performance. I use DLAA or DLSS in pretty much every new game at this point.
But the truth is, its all completely irrelevant what AMD or Nvidia marketing says. The product makes the reality. So while AMD just dug itself a hole, its not like it will influence anything.
If AMD gets Actual Flashy Motherfkng Frames to go big, it will sell itself. If they don't, they'll release another set of consoles and RDNA cards and keep doing as they do. Whatever.
In the meantime, Nvidia is selling what AMD wants to sell on discrete GPUs. Gosh, its almost like its 2016.It shouldn't, because the vast majority doesn't give a rat's ass, they just want to game rather than compare pixels. It needs to look palatable, that's enough, and upscale shouldn't noticeably cost image quality. DLAA and all the other four-to-eighteen-letter turbo features are completely irrelevant except to the niche that is anal about pixels.
Anyone really just playing games for gaming rather than watching graphical improvements meticulously, won't stop to even think about all of this. That is why the only worthwhile thing to talk about in all these technologies is the package that gets offered with every game as a standard thing. That is: a dropdown box with your FSR or DLSS ingame options, saying Quality or Performance or perhaps even some Ultra Quality thing that is actually DLAASSSGGAATURBOMODEWITHSPECIALSAUCETM. Everything that is not elevated to that level of market saturation, is dead in the water and nothing more than a marketing tool for gullible consumers that want to believe.
Most of the PCs are using their GPU, they don't have any competition on the top tier where really money is made with high margins, and they want to push people to upgrade to their newer GPUs.
But again, this is an engineer point of view, thinking GPUs as devices and not as products.
AMD is moving as many gaming GPUs as Nvidia does, in case you missed it :) And that's without the booming handheld market.
-_-
In 5 years time or so, most of the fps work will be done by AI upscaling software and not classic raster. Be prepared to evolve along or leave gaming all together.
Just like with cars. If you had told me, 10 years ago, i will have to pay a fee to have my seat heated i would laugh my ass out. Now, it is happening in some cases. You know what sucks the most here? In cold countries where you want to use this feature, the price is higher :)
15 years ago, almost to the day.
1600p60 FSAA was also a thing already on many games. And by the end of the year some cards had same performance level as the best of this graph but in 1600p.
You can't work on you physique, work on your mind. :)
And copy past this to ‘gasoline cult’ of old school and human driven cars vs electric/other energy sources self driving “cars”.
So many many things will become ‘nostalgic’ in so little time…
Want a standard? buy a console. I want options, competition, innovation. I use a PC and am here to talk about it. I 100% understand and appreciate your love and affinity for open standards, believe it or not I love them too, perhaps even ones that come would with every(?) game, but the way you broadly paint gamers with a given brush, and speak in generalizations, not for me at all. I am certain some people are like you describe, good for them I suppose, I speak on behalf of many others too, and I will make my best effort not to claim what anyone/everyone should or should do or care/talk about.
It's not about how many units you sell, or how much revenue you get, it's all about how much profit you get from each sold unit, AMD is not getting back much profit on those, in fact MS and Sony chose them because NVidia didn't want involved in that market.
You're talking about the booming handheld market, but you forgot Nintendo is a handheld device and they sell more units than PS5, this year with the new Nintendo console/handheld it will be a clear winner in units sold.
NVidia is ok in selling their GPUs to Nintendo because of their special negotiated agreement and because they can sell them at high margins, since they are much reduced versions compared to current generation, based on some info they are even based on last generation.