Do we have any idea how the new "Low-Lag" feature announced will work?

Discussion in 'Videocards - AMD Radeon' started by BlindBison, Jul 6, 2019.

  1. BlindBison

    BlindBison Ancient Guru

    Messages:
    2,404
    Likes Received:
    1,128
    GPU:
    RTX 3070
    https://www.pcgamesn.com/nvidia/nvidia-anti-lag-sharpening-amd-radeon-alternative

    AMD recently announced that they'd have some kind of automated Sharpening solution as well as a new "Low Lag" feature.

    In the article above Nvidia seems to be saying that it's all marketing BS -- that it's likely the same as "Max prerendered frames" / "max flip queue size", but this doesn't make sense to me because AMD already has that setting.

    Could this be some kind of dynamic flip queue size or some such? I don't know how this could work, but in theory something like that could be pretty awesome.
     
  2. OnnA

    OnnA Ancient Guru

    Messages:
    17,792
    Likes Received:
    6,691
    GPU:
    TiTan RTX Ampere UV
    We need to wait for reviews :p
    HH will tell Us what's UP....
    7.7 is tomorrow.
     
  3. JonasBeckman

    JonasBeckman Ancient Guru

    Messages:
    17,564
    Likes Received:
    2,961
    GPU:
    XFX 7900XTX M'310
    I wonder if they're trying to get something like flip model presentation going, cut the queue so to speak, reduce the steps or chains a bit for how much is needed from A -> B but that's just a guess, full-screen optimization in Win10 I think attempts something similar too for when it works and for third party utilities SpecialK can override presentation from bitblit to flip model for D3D11 titles though compatibility can vary. :)

    https://software.intel.com/en-us/articles/sample-application-for-direct3d-12-flip-model-swap-chains

    Something introduced with Windows 8 and then flip discard allowed for potentially lower latency with Windwos 10 though DirectX 12 and I assume VLK is where it's mandatory with D3D support being optional and well with Windows 7 not having this and still being popular well perhaps it can begin changing now. :)

    Can make a nice improvement on erratic or high frame time values too but it needs some fine tuning and mod it all you want a problematic PC port isn't going to be perfect unless it can be patched and yeah that doesn't happen too often for these in-depth advanced issues or how to call them. (Input? Yep these exist in numerous forms, graphical sure there's a ton of problematic things here and then we have audio and numerous other flaws, OS itself and also API's and third party including Steam, lots of Steam issues actually from poor implementation and little testing so some ports well they're going to be rough and that's being mild.)


    Hmm coupled with G or Free Sync (NVIDIA or AMD.) I wonder if this could improve further and it could be something different AMD exposes or automates which I wouldn't want them to do what NVIDIA is doing with the flip model and now also the render threading option bundling together and giving a vague name though later posts and forums clarified how it works but how many are going to read that.

    Render ahead isn't bad and just putting pre-rendered down to 1 isn't a perfect solution or it'd be the default already but would something like a DXGI override for flip model work or would it be more akin to what NVIDIA already has and also exposes to users via it's settings that now AMD is also allowing at least for NAVI GPU's. (Wonder if that's going to be cited as a hardware limit too but eh I think they only pulled that on VSR to be fair.)


    EDIT: But the possible upscaling from a lower resolution and sharpening to mitigate blur ("kinda") well that's a different thing entirely, will be interesting to read more on the subject once details are available. :)

    Intel now also supporting integer scaling is a curious development as well but again a different thing though upscaling has numerous modes and drawbacks though I would imagine bilinear and sharpening effect of some kind and nothing too fancy but eh I can be wrong so we'll see although I do not expect anything special but who knows.

    Not too long now until reviews and drivers appear, part of NAVI support has landed in the Linux drivers already too and from there users can dig up what is doing what and how things really work or so I hope and then we'll get some clarification on what's going on with these features.
     
    OnnA and SpajdrEX like this.
  4. signex

    signex Ancient Guru

    Messages:
    9,045
    Likes Received:
    297
    GPU:
    RTX 2080 Ti
    I haven't noticed any difference when toggling it on/off in AC Odyssey.
     

  5. BlindBison

    BlindBison Ancient Guru

    Messages:
    2,404
    Likes Received:
    1,128
    GPU:
    RTX 3070
    Alright, now that more information on this has come out, has anyone been able to track down discussion of what the feature actually does/how it really works?

    Is it the same as setting the Flip Queue / Max Prerendered Frames to "1"? Or, is it a different feature entirely? Or, does it dynamically adjust the size of the flip queue?

    How it works and how it relates to the flip queue / max frames to render ahead is important because prerendering more frames / upping the flip queue size has many notable pros -- it can improve frame pacing (smoothness), increase overall performance (fps), and reduce stuttering/hitching (small freezes) -- which is why a lot of outlets recommend just leaving the value at "Let the 3D App Decide" / "Default".

    If AMD's new AntiLag feature is literally just setting the flip queue to 1, than that's nothing new and probably may not always be worth it in due to the tradeoffs it can result in (lower fps + more discernible hitching/stutter). If it's something else or some sort of "Dynamic" flip queue size, than that is far far more interesting and that's what I want to know about since some sort of Dynamic flip queue size might be able to decrease input lag while also avoiding the notable downsides to reducing the flip queue size.
     

Share This Page