GSYNC Breaks In-Game VSYNC

Discussion in 'Videocards - NVIDIA GeForce Drivers Section' started by PhantomGamers, May 22, 2017.

  1. Memorian

    Memorian Ancient Guru

    Messages:
    3,027
    Likes Received:
    209
    GPU:
    Strix 2080 Ti OC
    RTSS is the best frame limiter(Less input lag).
     
  2. archie123

    archie123 Ancient Guru

    Messages:
    2,476
    Likes Received:
    17
    GPU:
    Aorus 1080ti Extreme
  3. RealNC

    RealNC Ancient Guru

    Messages:
    3,150
    Likes Received:
    1,375
    GPU:
    EVGA GTX 980 Ti FTW
    At least two independent tests came to the same conclusion.

    I provided a method that everyone can use. It's easy to do and allows you to see for yourself which method has the least input lag for your case :) You can do this with any game that has a software rendered mouse cursor (examples are Fallout 4 and also Witcher 3 with "Hardware cursor" set to OFF in the game's video settings.)

    Just out of curiosity, if you do the test yourself, what results do you get? Did the previous profile inspector version give you input lag as low as RTSS?
     
  4. Mda400

    Mda400 Master Guru

    Messages:
    848
    Likes Received:
    89
    GPU:
    1660 - 2130/10ghz
    You should know I've already done both from past arguments. I said there is not a difference to me in delay between both methods (using inspector 2.1.3.4 + 59 fps v1 limit
    [~59.7 fps] + use 3d application vsync | RTSS 7.0.0 with limit=5988 and denominator=100) and again, we have different setups as you use G-Sync and I don't.

    But i'd rather have something I set and works in the driver itself than a program that runs in the background using resources.


    Also the question about the software mouse cursor test is how do you know it doesn't only affect software mouse cursor games? Directly related to how RTSS limits frames versus how Inspector and the driver do, there can be bias. One goes "im going to chop off any excess frames from this point on" and the other goes "i'm running at this many frames, better only run at this speed"
     
    Last edited: May 23, 2017

  5. RealNC

    RealNC Ancient Guru

    Messages:
    3,150
    Likes Received:
    1,375
    GPU:
    EVGA GTX 980 Ti FTW
    Well, you didn't do the test then.

    The test is to set a 23FPS cap with Inspector and 0FPS cap with RTSS, test mouse cursor movement with a games that has a software rendered cursor (the test will fail if it's not a software cursor; a game that's guaranteed to give you such cursor is Fallout 4 in its startup main manu) then alt+tab to RTSS, set a 22FPS cap, alt-tab back to the game, and see if that reduces your input lag.

    That's the actual test. If you cap to 59FPS, you're not going to actually be able to tell how many frames of extra lag there are without using a high-speed camera.

    At 23FPS, each frame adds an additional input lag of 45ms, which is why this test works without needing to use special equipment.
     
    Last edited: May 23, 2017
  6. Mda400

    Mda400 Master Guru

    Messages:
    848
    Likes Received:
    89
    GPU:
    1660 - 2130/10ghz
    Read the last thing I said above. If you are still not swayed, let's agree to disagree.

    I'm trying to say there's more to look at before you declare RTSS the better decision. But you vehemently insist it, so there is no reason to waste anymore time. Hopefully others will take that into account.
     
    Last edited: May 23, 2017
  7. Agent-A01

    Agent-A01 Ancient Guru

    Messages:
    11,363
    Likes Received:
    904
    GPU:
    1080Ti H20
    We've already had two 'professional' tests done that prove you are wrong.

    Just because you don't notice it doesn't disprove the facts.

    As always. Ingame > RTSS > NV inspector.

    Weird how you say RTSS using a few mb is stealing resources... lol
     
  8. RealNC

    RealNC Ancient Guru

    Messages:
    3,150
    Likes Received:
    1,375
    GPU:
    EVGA GTX 980 Ti FTW
    It's not about the cursor. It's about the added input lag. The software cursor just allows for observing the effect easier rather than have to actually start the main game and play it. Makes testing various settings quicker since you get to the main menu of the games faster compared to loading up a save or starting a new game or loading a map.

    You can just as well start the game fully and play it. Same result.

    And it indeed only affects software mouse cursors. That's the whole point. Hardware mouse cursors are not affected by frame limiters, so it's not possible to test frame rate limiter induced latency with a hardware mouse cursor.
     
    Last edited: May 23, 2017
  9. sniperbr0

    sniperbr0 Maha Guru

    Messages:
    1,108
    Likes Received:
    1
    GPU:
    EVGA SC BLACK 1080TI
    Pardon my ignorance, but why exactly is Gsync + Vsync ON better than Gsync + Vsync OFF if we are using a frame limiter below the refresh rate? What is the difference? I am very interested in this.
     
  10. RealNC

    RealNC Ancient Guru

    Messages:
    3,150
    Likes Received:
    1,375
    GPU:
    EVGA GTX 980 Ti FTW
    G-Sync seems to have a frame time variance compensation function which is controlled by the v-sync setting. Without it, you can still get tearing with g-sync if frame times are unstable (they vary too much) or too high (meaning very low FPS.) You can get a tearing line at the bottom or top of the screen because the frame buffers are flipped just a bit too soon or too late.

    The frame time variance protection was always enabled when g-sync was first released. It was how nvidia had originally introduced g-sync. The option to turn vsync off was added later, but turning it off means you can get tearing even with g-sync.

    This is not an issue with stable frame pacing though. The only game I know of where this happens right now is CS:GO.

    The tests performed so far have shown that having vsync enabled in the control panel (and thus having the protection mechanism enabled) does not increase input lag if you are within the g-sync refresh rate range. So if your goal is to stay within g-sync range at all times, capping the frame rate and enabling v-sync makes sure you never get tearing, even when FPS fluctuates too much. No input lag increase, so it's a win-win situation.
     
    Last edited: May 24, 2017

  11. Memorian

    Memorian Ancient Guru

    Messages:
    3,027
    Likes Received:
    209
    GPU:
    Strix 2080 Ti OC
    So the best situation is:

    G-Sync + Fast/V-Sync On + RTSS frame limiter.
     
  12. RealNC

    RealNC Ancient Guru

    Messages:
    3,150
    Likes Received:
    1,375
    GPU:
    EVGA GTX 980 Ti FTW
    Well, that's what we preached until now. But it looks like the new driver + new Profile Inspector settings might at last mean nvidia's limiter is catching up. Waiting for tests.
     
  13. Mda400

    Mda400 Master Guru

    Messages:
    848
    Likes Received:
    89
    GPU:
    1660 - 2130/10ghz
    I fail to see where this is fact... but ill leave you to it.

    And you'd be surprised on a Q9550 @ 4ghz it utilized a whole core to limit with RTSS
     
    Last edited: May 24, 2017
  14. khanmein

    khanmein Ancient Guru

    Messages:
    1,648
    Likes Received:
    71
    GPU:
    EVGA GTX 1070 SC
    @RealNC If I play MOBA games, enabled Fast Sync (NvCpl) + RTSS frame limit to 60 + V-Sync OFF (under the game setting) is better than V-sync turned it ON?

    For triple A & sports games, I set Adaptive Sync (NvCpl) + RTSS capped 60 + V-Sync OFF (in the game)
     
    Last edited: May 24, 2017
  15. Agent-A01

    Agent-A01 Ancient Guru

    Messages:
    11,363
    Likes Received:
    904
    GPU:
    1080Ti H20
    What do you mean?

    Uh facts are the professionally done tests to measure latency?

    and RTSS doesn't use any CPU for me.
    You probably have some isolated issue.
     

  16. RealNC

    RealNC Ancient Guru

    Messages:
    3,150
    Likes Received:
    1,375
    GPU:
    EVGA GTX 980 Ti FTW
    To me, it looks like busy waiting (that's a term; look it up.) This is blocking the game from rendering more frames with extremely precise timing. (RTSS's accuracy is in the microseconds level.)

    If the cap is not reached, there's no waiting, and thus no CPU use. If the cap is reached, the busy wait loop runs, which blocks the game. The faster the game would run, the more waiting is done by RTSS, the more CPU this appears to use (a busy wait loop does nothing, but it still shows up as CPU use.)

    To give an example, for a 100FPS cap, the target frame time is 10ms. If the game renders frame in 1ms (1000FPS), you get 9ms of waiting to get to the 10ms target. That appears to eat CPU. The faster the game is, the more CPU appears to be used by RTSS.

    Since I can't see the RTSS source code, I can't guarantee that this is indeed what happens, but it sure looks like it. In which case, all is fine and how it should be.


    It's on the previous page. The Profile Inspector author uploaded a new version to:

    https://ci.appveyor.com/project/Orbmu2k/nvidiaprofileinspector/build/artifacts

    Which allows you to actually force the v1 limiter in profile inspector with the newest nvidia driver. If you also set Fast Sync, it *seems* you get input lag that's as good as RTSS. Note: *seems*. My "dumb" 23FPS testing method is indication that it now works well, but not proof. Hopefully there will be some actual tests for this in the coming week.

    (Fast Sync never kicks in since I limit below g-sync max refresh, but the limiter has low input lag when selecting fast sync. It's weird, but that's how it appears to be.)
     
    Last edited: May 24, 2017
  17. Agent-A01

    Agent-A01 Ancient Guru

    Messages:
    11,363
    Likes Received:
    904
    GPU:
    1080Ti H20
    Did you post this information at blur busters so they can test it?
     
  18. RealNC

    RealNC Ancient Guru

    Messages:
    3,150
    Likes Received:
    1,375
    GPU:
    EVGA GTX 980 Ti FTW
    On 60Hz, a 60FPS cap is not going to do anything useful. If you have g-sync, you need to cap 58FPS. If you don't have g-sync, it gets complicated. You need to find out your exact refresh rate:

    https://www.vsynctester.com

    And then cap your FPS to about 0.007FPS below the refresh rate. For example, if your refresh rate is 59.869Hz, you need to cap to 59.862FPS.

    Or you can use CRU to raise your refresh rate to 60.007Hz and then cap to 60FPS.

    I posted about this here:

    http://forum.guru3d.com/showpost.php?p=5380262&postcount=232


    Yep. It's in the "G-Sync 101" thread. Chris (Battle(non)sense) also knows about it and he hopes to do a quick preliminary test too.
     
  19. cryohellinc

    cryohellinc Ancient Guru

    Messages:
    3,333
    Likes Received:
    2,655
    GPU:
    RX 5700 XT/GTX 1060
    Lots of great information here, thank you guys.

    Currently using 98fps new v1 limiter cap + Gsync + fastsync, and it does feel better. :)

    Great stuff!
     
  20. Agent-A01

    Agent-A01 Ancient Guru

    Messages:
    11,363
    Likes Received:
    904
    GPU:
    1080Ti H20
    Interesting.

    So what about Fast Sync ON + ingame FPS limiter (a couple below refresh) cap.

    Or still better to stick with Vsync ON.
    I say it doesn't matter as neither should activate but maybe there's something that's better than the other.

    Also what about frame rate stability?

    RTSS always was dead accurate while NV inspector had fluctuations and setting 142fps would result in limit to 133~ or something like that(aka not accurate)
     

Share This Page