Announcement

Collapse
No announcement yet.

VKD3D-Proton 2.12 Released With Initial Support For NVIDIA Reflex

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • #11
    I remember looking at it some years ago. It didn't work with vsync, so useless to me. But apparently it's still not fixed. No vsync support. Other than that, it says "10ms latency reduction" in big letters, but there an "up to" in small letters above it. Meanwhile, Reflex easily does 25ms reduction:



    It's a good technology and works well for me. It might not be open, but it works well and it's very useful to me.

    What is required is not a vendor agnostic method, but a common API for this that ties into each vendor's implementation (Reflex on Nvidia, Anti-Lag on AMD, and whatever Intel is using.) In the Windows world, a new cross-vendor API is being introduced for upscaling (DirectSR, which maps to FSR/DLSS/XeSS.) This is the road that should be taken. The vendor should decide which implementation is best for their GPU, as long the API to access it is cross-vendor. Put it in Vulkan, for example.
    Last edited by RealNC; 15 March 2024, 03:20 PM.

    Comment


    • #12
      thanks you Nvidia, hope we see zero day release of API on Linux too, please add SDR to HDR video support in Linux too

      Comment


      • #13
        Originally posted by Kjell View Post

        How does this translate to Linux?
        What type of tuning is done?

        Couldn't this library be used with AMD GPUs to perform the same type tuning?
        It translates the DX12 nVidia Reflex calls to the Vulkan nVidia extension VK_NV_low_latency2 which is then handled by the nVidia driver. For this to work on AMD the AMD drivers have to expose similar functionality since this is all handled by the driver and not by VKD3D (it just handles the translation between DX12 and Vulkan).

        Comment


        • #14
          Originally posted by F.Ultra View Post

          It translates the DX12 nVidia Reflex calls to the Vulkan nVidia extension VK_NV_low_latency2 which is then handled by the nVidia driver. For this to work on AMD the AMD drivers have to expose similar functionality since this is all handled by the driver and not by VKD3D (it just handles the translation between DX12 and Vulkan).
          Yup, this was covered by the article.

          The interesting part is, what type of modifications are done that allow this massive latency reduction even in Linux. Reducing amount/sizes of buffers in internal drivers? Tuning power management? Why can't these tweaks be toggled by the user without Reflex library?

          I suppose answers to this lies in how the FOSS version works, LatencyFlex

          Comment


          • #15
            Does this open the door for Nvidia frame generation? My understanding was that depends on reflex being available

            Comment


            • #16
              Originally posted by Kjell View Post
              The interesting part is, what type of modifications are done that allow this massive latency reduction even in Linux.
              Killing CPU prerender without killing performance via kind of a dynamic in-game fps limiter (probably). You can already have the same dramatic drop in latency when using e.g. Unreal Engine's own fps limiter, but of course that's missing the dynamic part.

              Always been wondering why AMD & Intel can't get their crap together in mimicking this feature. It's just disastrous marketing by leaving all of the "e-sports" stuff to Nvidia. Just utterly stupid.

              And now Nvidia also uses it with VRR + vsync to avoid vsync backbuffer queue latency when vsync would cap fps, and also with DLSS 3 FG. How "unexpected"...
              Last edited by aufkrawall; 16 March 2024, 10:43 AM.

              Comment


              • #17
                Originally posted by aufkrawall View Post
                Killing CPU prerender without killing performance via kind of a dynamic in-game fps limiter (probably). You can already have the same dramatic drop in latency when using e.g. Unreal Engine own's fps limiter, but of course that's missing the dynamic part.

                Always been wondering why AMD & Intel can't get their crap together in mimicking this feature. It's just disastrous marketing by leaving all of the "e-sports" stuff to Nvidia. Just utterly stupid.

                And now Nvidia also uses it with VRR + vsync to avoid vsync backbuffer queue latency when vsync would cap fps, and also with DLSS 3 FG. How "unexpected"...
                isn't AMD antilag+ similiar ?

                Comment


                • #18
                  fsr3 fg is supported under linux, but dlss fg isnt. fyi

                  Comment


                  • #19
                    Originally posted by luno View Post

                    isn't AMD antilag+ similiar ?
                    Game bans were issued over it since it wasn't natively supported by games

                    This is why there needs to be a industry standard like Vulkan protocols or LatencyFlex.. The end result is the same, difference is wasted developer time from having to maintain multiple implamentations
                    Last edited by Kjell; 16 March 2024, 09:29 AM.

                    Comment


                    • #20
                      Originally posted by theriddick View Post
                      fsr3 fg is supported under linux, but dlss fg isnt. fyi
                      Yes, but it currently works only with very poor frame pacing in conjunction with vsync, which also kills experience on Wayland and on Xorg with VRR (as that forces TearFree).
                      Maybe some "HAGS" equivalent is missing, I'm also getting worse results without it on Windows. When FSR 3's FG part works correctly on Windows with HAGS (with driver-forced vsync), it's mighty impressive though.

                      Comment

                      Working...
                      X