Will this become available in older models like FSR and Framegen or will it be exclusive to RDNA4. It would be a boon to RDNA3 mobile chips if the performance gain is comparable to FSR3.

  • Viri4thus@feddit.orgOP
    link
    fedilink
    English
    arrow-up
    1
    ·
    23 hours ago

    XESS is hardware agnostic, it’s just faster and slightly better with XMX vs DP4A. I’m hoping that will be a similar situation with RDNA4 and earlier. Granted, it’s wishful thinking, it would be a shame if that was not the case. There’s an entire gen of strix halo APUs with RDNA 3.5 and that would greatly benefit from a better upscaling algo to save battery life/improve performance. Hell, Mendocino, Dragon Range, Rembrandt and Phoenix Point getting RDNA 4 would be absolutely amazing.

    • bazsy@lemmy.world
      link
      fedilink
      English
      arrow-up
      1
      ·
      22 hours ago

      https://youtu.be/2p7UxldYYZM?t=508

      • Yeah. So I guess the next question for me is, does that mean that FSR four is going to be exclusive to,
      • look, I would say it’s right now it has to be because of what I just said. It requires a lot of compute to be able to generate the pixels.

      Just confirmed that at least on launch it will be RDNA4 exclusive. I just hope it can run on other hardware even if it is a dedicated NPU in mobile chips.

      • Viri4thus@feddit.orgOP
        link
        fedilink
        English
        arrow-up
        1
        ·
        22 hours ago

        Ye, I’m hoping they will bring it down the line like what happened with FSR2 and Frame Gen.