Skip to main content

AMD Pinky Promises FidelityFX Super Resolution Is Still on Track for 2021 Debut

AMD Booth
(Image credit: Shutterstock)

AMD Radeon CVP and GM Scott Herkelman said in a video interview with PCWorld that FidelityFX Super Resolution (FSR), the company's response to Nvidia DLSS, is "progressing very well internally" and that he believes it could debut later this year.

Nvidia introduced DLSS in 2018 and released DLSS 2.0 in 2020, so in some ways, AMD's response to the technology is coming later than some might have expected. But that can partly be attributed to AMD's plans to make FSR a cross-platform tool.

DLSS is currently limited to Nvidia graphics cards. FSR is supposed to operate across AMD's GPUs, including those used in the Xbox Series X|S and PlayStation 5, as well as graphics products made by Intel and Nvidia. That's a much bigger undertaking.

"Our commitment to the gaming community is [FSR] needs to be open, it needs to work across all things, and our game developers need to adopt it and feel like it's a good thing," Herkelman said in the PCWorld interview, which you can watch here:

Herkelman also said that FSR is "probably one of the biggest software initiatives we have internally, because we know how important it is that if you want to turn on ray tracing, you don't want to [...] have your GPU get hit so hard." 

Unfortunately, it seems like foundational aspects of FSR still have to be figured out. Herkelman told PCWorld the tool wouldn't have to be based on machine learning, and that it's working with game developers to find the best way to improve performance.

In the meantime, Nvidia said Wednesday that nearly 40 titles currently support DLSS and that "there are many more implementations of these technologies waiting in the wings to be announced and released in the coming weeks and months."

This could turn out to be a tortoise-and-the-hare situation. An open source, cross-platform solution like FSR could easily appeal to developers more than a proprietary technology like DLSS. The problem is FSR hasn't even shown up to the race track.

  • hannibal
    Interesting...
    This could be good!

    Options are, we have Nvidia sponsored games with Nvidia dlls, we would have AMD sponsored games with super resolution and Intel sponsored games having some other upscaling option...

    Or we could have common api to all upscale solutions...
    We need this, or we will have very segmented market... again!
    Then it would be just about who makes that api work better, and faster than two other competitors. Instead of having three separate systems that only work with one manufacturer...
    Reply
  • spongiemaster
    I'd prefer a better solution to a universally available inferior technique. DLSS 2.0 is supposed to be relatively simple for developers to implement, and Nvidia undoubtedly helps them through the process. If FidelityFX is a completely software driven hardware agnostic technique, I don't see how it will perform as well as DLSS 2.0 or any upcoming newer version.
    Reply
  • hotaru.hino
    hannibal said:
    Or we could have common api to all upscale solutions...
    We need this, or we will have very segmented market... again!
    Then it would be just about who makes that api work better, and faster than two other competitors. Instead of having three separate systems that only work with one manufacturer...
    There isn't going to be an API for upscaling solutions because said solutions are implementation details. API does not define implementation details.

    spongiemaster said:
    I'd prefer a better solution to a universally available inferior technique. DLSS 2.0 is supposed to be relatively simple for developers to implement, and Nvidia undoubtedly helps them through the process. If FidelityFX is a completely software driven hardware agnostic technique, I don't see how it will perform as well as DLSS 2.0 or any upcoming newer version.
    The name of the feature itself to me implies it's using super resolution imaging. Though honestly, the only thing I'm taking issue with now is AMD appears to be branding things that are essentially open or have existed like they came up with it or something. Adaptive syncing was already a thing. PCIe resizable BAR was already a thing. And if it's more or less true that FidelityFX Super Resolution is simply a super resolution imaging technique, then well, time to add that to the list.
    Reply
  • husker
    "AMD Radeon CVP and GM Scott Herkelman said in a video interview with PCWorld that FidelityFX Super Resolution (FSR), the company's response to Nvidia DLSS, is "progressing very well internally" and that he believes it could debut later this year. "

    Hmmm... not really that positive of a comment. It's only March and AMD is saying it could debut this year. That's the same as saying it possibly could not debut this year. But since I won't be able to find a RX6800 or 6900 till next year anyway, I guess it's a mute point.
    Reply
  • salgado18
    hotaru.hino said:
    There isn't going to be an API for upscaling solutions because said solutions are implementation details. API does not define implementation details.

    The name of the feature itself to me implies it's using super resolution imaging. Though honestly, the only thing I'm taking issue with now is AMD appears to be branding things that are essentially open or have existed like they came up with it or something. Adaptive syncing was already a thing. PCIe resizable BAR was already a thing. And if it's more or less true that FidelityFX Super Resolution is simply a super resolution imaging technique, then well, time to add that to the list.
    By "was already a thing" you mean "the spec existed, but AMD made the first practical implementation"? Because there were no products with Adaptive sync or PCIe resizable BAR before AMD made them. Even Nvidia is now supporting Adaptive Sync.

    I agree that proprietary solutions give better results, but they are not good for consumers in general. I won't even mention PhysX again, just consider that you needed a GeForce card and a G-sync monitor to make it work, and now you can get any card and almost any monitor.

    Proprietary is only good in the company's perspective, as it profits from the monopoly of technology. And, as a monopoly, it can price whatever it wants. It is never good for the consumer.
    Reply
  • hotaru.hino
    salgado18 said:
    By "was already a thing" you mean "the spec existed, but AMD made the first practical implementation"? Because there were no products with Adaptive sync or PCIe resizable BAR before AMD made them. Even Nvidia is now supporting Adaptive Sync.
    Well here's a counterpoint to what I feel AMD is doing. NVIDIA funded the research for FXAA. I don't see NVIDIA's branding on it anywhere nor did NVIDIA try make it seem like the own the algorithm.

    I mean it's cool and all that AMD pushed for something to be done with it. But I feel it's a slap in the face to call it "open" when you repackage it and put your own branding on it.

    I agree that proprietary solutions give better results, but they are not good for consumers in general. I won't even mention PhysX again, just consider that you needed a GeForce card and a G-sync monitor to make it work, and now you can get any card and almost any monitor.

    Proprietary is only good in the company's perspective, as it profits from the monopoly of technology. And, as a monopoly, it can price whatever it wants. It is never good for the consumer.
    And I'm not saying it is or isn't. However, I'm not a fan of a company putting their mark on something they didn't actually invent in order to win points.
    Reply
  • mitch074
    hotaru.hino said:
    Well here's a counterpoint to what I feel AMD is doing. NVIDIA funded the research for FXAA. I don't see NVIDIA's branding on it anywhere nor did NVIDIA try make it seem like the own the algorithm.

    I mean it's cool and all that AMD pushed for something to be done with it. But I feel it's a slap in the face to call it "open" when you repackage it and put your own branding on it.


    And I'm not saying it is or isn't. However, I'm not a fan of a company putting their mark on something they didn't actually invent in order to win points.
    They may not have invented it, but they were the first to actually make it work. Intel and Nvidia never bothered until AMD rubbed their nose in it.
    Reply
  • spongiemaster
    salgado18 said:
    I agree that proprietary solutions give better results, but they are not good for consumers in general. I won't even mention PhysX again, just consider that you needed a GeForce card and a G-sync monitor to make it work, and now you can get any card and almost any monitor.

    Proprietary is only good in the company's perspective, as it profits from the monopoly of technology. And, as a monopoly, it can price whatever it wants. It is never good for the consumer.
    Why do you need a G-sync monitor for PhysX? G-sync didn't even exist when Nvidia acquired PhysX.

    Nvidia isn't making money directly from DLSS. The development costs are rolled into the cost of the video card just like the development cost for AMD's FidelityFX will be rolled into the cost of their GPU's. Don't think because FidelityFX is supposed to be open that that makes it free to anyone buying an AMD card. You're still paying for their development costs. Their is no licensing fee for developers to use DLSS. The only money Nvidia makes off of it is from customers buying their GPU's because they want the feature.
    Reply
  • salgado18
    spongiemaster said:
    Why do you need a G-sync monitor for PhysX? G-sync didn't even exist when Nvidia acquired PhysX.
    I'm not linking the two like that. I mean PhysX is a great physics simulation tech that was always locked to Nvidia GPUs, and because of that it never got used as a game tool, only as a special effects gimmick. Consumers lost a lot all those years because of it (no game could be heavily based on PhysX, risking locking the game to one GPU manufacturer, because CPUs can't handle the load properly).

    spongiemaster said:
    Nvidia isn't making money directly from DLSS. The development costs are rolled into the cost of the video card just like the development cost for AMD's FidelityFX will be rolled into the cost of their GPU's. Don't think because FidelityFX is supposed to be open that that makes it free to anyone buying an AMD card. You're still paying for their development costs. Their is no licensing fee for developers to use DLSS. The only money Nvidia makes off of it is from customers buying their GPU's because they want the feature.
    They are making money out of DLSS. Their cards double performance with it, AMD cards don't because they don't support it. This is fine from the company's perspective, and I think it is natural, but it's not good for consumers, because only one vendor does it, which is a monopoly, which lets the company price it however it likes.

    Want a pro-consumer idea? Make DLSS open. It still needs hardware that AMD doesn't have, but it is not locked to Nvidia anymore. RX 6000 could have the hardware to do it, and then it would be able to compete on the implementation front (like it is one generation behind in RT performance), but I could get an RX 6800 and use some sort of DLSS with it. Nvidia would be better? Sure, but then devs wouldn't need to implement two solutions, consumers wouldn't be locked out of inovation, every game could have it (imagine only AMD cards had anti-aliasing, how would that feel?).

    Again, it makes sense from the company perspective, but it's bad for consumers. Open technology, proprietary implementation, that's good.
    Reply
  • spongiemaster
    salgado18 said:
    I'm not linking the two like that. I mean PhysX is a great physics simulation tech that was always locked to Nvidia GPUs, and because of that it never got used as a game tool, only as a special effects gimmick. Consumers lost a lot all those years because of it (no game could be heavily based on PhysX, risking locking the game to one GPU manufacturer, because CPUs can't handle the load properly).

    PhysX has been open source for a few years now. It's never caught on because it doesn't bring much to games. The technology is 20 years old now, it's not like Nvidia bought a universally loved tech and then ruined it. It was never popular. The implementation required too much work by developers for something that was never going to sell games. Advance physics modeling isn't something you can demonstrate in static pictures in a magazine. Originally requiring an add in PCI card only for physics calculations pretty much doomed the tech from the start.


    They are making money out of DLSS. Their cards double performance with it, AMD cards don't because they don't support it. This is fine from the company's perspective, and I think it is natural, but it's not good for consumers, because only one vendor does it, which is a monopoly, which lets the company price it however it likes.

    Want a pro-consumer idea? Make DLSS open. It still needs hardware that AMD doesn't have, but it is not locked to Nvidia anymore. RX 6000 could have the hardware to do it, and then it would be able to compete on the implementation front (like it is one generation behind in RT performance), but I could get an RX 6800 and use some sort of DLSS with it. Nvidia would be better? Sure, but then devs wouldn't need to implement two solutions, consumers wouldn't be locked out of inovation, every game could have it (imagine only AMD cards had anti-aliasing, how would that feel?).

    Again, it makes sense from the company perspective, but it's bad for consumers. Open technology, proprietary implementation, that's good.
    Nvidia controls 80% of the dGPU market. Anything that Nvidia produces for their cards isn't exactly fragmenting the market. They pretty much are the market. This isn't like the late 90's when there were 5 or more companies each trying to develop their own technologies. Nvidia didn't develop DLSS then give it to devs and tell them good luck. Nvidia works with the developers to utilize their technologies. Especially with DLSS 1.0, Nvidia had to train every resolution for every game that used it on their own supercomputers. Why on earth would Nvidia then turn around and just let it work on AMD's hardware? Nvidia shoulders all the work and costs and then let's AMD benefit from it for nothing? That's not consumer friendly, because companies aren't going to develop technologies with financial returns that benefit their competitors more than it does themselves.
    Reply