Nvidia Inspector Frame Rate Limiter different versions explanation?

Discussion in 'Videocards - NVIDIA GeForce Drivers Section' started by 321Boom, Jun 25, 2017.

  1. 321Boom

    321Boom Member Guru

    Messages:
    122
    Likes Received:
    12
    GPU:
    GTX980 Ti
    Hello,
    I just updated to the latest version of Nvidia Inspector (2.1.3.7), and under Frame Rate Limiter, there is an option called Frame Rate Limiter mode.

    Can anyone explain the difference between these 3 options:
    Default (does this mean V1?)
    Limiter V2 - Allow All
    Limiter V2 - Force Off [Enables Limiter V1 (Less Input Lag)]

    I was using Frame Rate Limiter set to 60 fps (v2) on the older version I had (2.1.3.0) which is quite straightforward, but I'm finding these new options in this new version a bit unclear.

    Thank you for any help and clarification :)
     
  2. RealNC

    RealNC Ancient Guru

    Messages:
    4,945
    Likes Received:
    3,223
    GPU:
    4070 Ti Super
    This was tested here:

    http://forums.blurbusters.com/viewtopic.php?f=5&t=3441#p26810

    The new setting is the frame limiter that gets automatically used with fast sync. For it to work, you need to also force fast sync.

    Note that forcing fast sync is just a "secret" switch to make the "v1" limiter actually work. If your FPS cap is about 10FPS lower than your refresh rate, fast sync isn't used at all, but the v1 limiter still is.

    However, it still has more lag than the RTSS frame limiter though, as the tests show. In general:

    In-game limiter (if the game has one): Best option (reduces input lag).
    RTSS: Good option (does not reduce but also doesn't add input lag.)
    NVidia v1 limiter + fast sync: Not so good option (increases input lag by at least 1 frame.)
    NVidia limiter (v1 or v2) without fast sync: Worst option (increases input lag by at least 2 frames.)

    In practice though, RTSS still reduces input lag by a bit. For example, if your game runs at 140FPS uncapped, you get somewhat lower latency if you cap to 139FPS with RTSS. The NVidia limiter (as configured in Profile Inspector) is not as good with this, even the v1 + fast sync option.

    However, the v1 + fast sync option can still be "good enough" if you don't want to use RTSS and prefer the nvidia limiter. You need to cap well below refresh rate though, otherwise fast sync kicks in and you'll get stutter. RTSS does NOT require fast sync.


    Edit:

    These are g-sync tests, but I think the results are in all likelihood also applicable even without g-sync. G-Sync should in theory not have any impact on the frame limiters.
     
    Last edited: Jun 25, 2017
    RodroG and mirh like this.
  3. 321Boom

    321Boom Member Guru

    Messages:
    122
    Likes Received:
    12
    GPU:
    GTX980 Ti
    Hello,
    thank you for your detailed reply, while very informative, it still doesn't answer my main question though.

    Which is the option that comes equivalent to the V2 I was using in the older version (2.1.3.0)?

    By new setting I take it you mean Force Off?
    Is Default V1 or V2?
    What is V2 - Allow All?
    I am currently using V2 - Allow All. It seems to be giving the results closest to before updating.

    I cannot use Fast Sync, the games I play are STGs (shmups) which require a steady 60 fps.

    I cannot use RTSS on certain games, I still get screen tearing with it, Nvidia Inspector eliminates most screen tearing so for some games I'm forced to use that as it seems like the best option. I know about the input lag it adds but it varies by game. For some it's bearable, for others it's not even an option to use it.
     
  4. RealNC

    RealNC Ancient Guru

    Messages:
    4,945
    Likes Received:
    3,223
    GPU:
    4070 Ti Super
    V2 and V1 should give the same result when not using fast sync. There does not seem to be a difference. You can just pick v1 in your case and the results should be as before.

    If you want low-latency vsync (no tearing), you can follow my guide here:

    http://forums.guru3d.com/showthread.php?p=5380262#post5380262

    If you get it right, you will get smooth 60FPS vsync'ed motion, with reduced input lag and no stutter.
     
    Last edited: Jun 25, 2017

  5. 321Boom

    321Boom Member Guru

    Messages:
    122
    Likes Received:
    12
    GPU:
    GTX980 Ti
    I had read on these forums that V2 should lock the frames better: showthread.php?t=409652 (post 7)

    By V1 you mean Default? I'm still curious as to what the V2 - Allow All does though, especially since on the previous version I was using V2.

    Wow that DEFINITELY is an interesting read, I'll be sure to try that out next time I'm on a Windows game! (currently I'm playing through an emulator and RTSS does not recognize it for some reason, no change in frame rate, even if I do a stupid number like 27 or 38 fps :/). Regarding your suggestion, I'd prefer using step 2, having the game capped to 60 fps, and increasing the monitor's refresh rate instead of using a fractional cap. So instead of using CRU, would a custom refresh rate (with an increase of 0.007 Hz) set in a custom resolution in Nvidia Control Panel do the same? I used the vsynctester site you mentioned, my refresh rate to 3 decimal points is coming at 60.000 Hz, so I could just set mine to 60.007 Hz? (Note I'm on a tv not a monitor in case it makes a difference).

    Thanks for your suggestions, greatly appreciated :D
     
  6. RealNC

    RealNC Ancient Guru

    Messages:
    4,945
    Likes Received:
    3,223
    GPU:
    4070 Ti Super
    v2 is the default. v1 seems to behave like v2, unless you select fast sync. In other words: v1 without fast sync is the same as v2. v1 with fast sync has lower input lag.

    The "lock the frame rate" better comment refers to nvidia's limiter ignoring your frame cap and instead capping to the same amount as your refresh. If you select 59FPS, it ignores it and picks 60FPS instead. There seemed to be a difference in that behavior with v1 and v2 according to some people, but I can't see it.

    (Yes, it's confusing and weird... NVidia's limiter does what it wants and seems to have its own mind :p)

    Not sure if TVs can accept non-standard signals. You'd have to try and see. In any event, you'd need to modify the existing mode and raise it to 60.008 or so. I don't think it's possible to have two modes that use almost the same refresh rate.
     
    Last edited: Jun 26, 2017
  7. MrBonk

    MrBonk Guest

    Messages:
    3,385
    Likes Received:
    283
    GPU:
    Gigabyte 3080 Ti
    I have never heard or seen evidence of the driver FPS limiter adding a frame of latency.

    If you would point to some actual tests done. I'd love to see them.
     
  8. 321Boom

    321Boom Member Guru

    Messages:
    122
    Likes Received:
    12
    GPU:
    GTX980 Ti
    @RealNC: As you said, for some people they reported a difference between V1 and V2. I notice a difference too. Currently using V2 - Allow All is producing the same/closest results to what I was getting before I updated. I would still like to know what this 'Allow All' is doing if you can clarify it though please. Using Default felt different for me :/

    Yes my TV accepts non-standard refresh rates, and yes I can have two resolutions that use almost the same refresh rate. I have a custom resolution set to 60.001 Hz alongside a standard 60 Hz resolution with no problem. I do notice that screen tearing is minimized when I'm on the 60.001 Hz resolution for some reason. Why did you suggest setting it to 60.008 HZ instead of 60.007 Hz as stated in your guide? Isn't the difference meant to be 0.007 Hz?

    Also, does maximum pre-rendered frames need to be set to 1? I'd like to avoid this if possible to have less chances of stutter and dropped frames. Will your suggested method still work effectively if max pre-rendered frames are not at 1, or will they throw off the timing or something? (I usually leave them at Default/Use 3D Application Setting). The results at the end of the game show slowdown/dropped frames as a %, so the lower I keep this % the better.

    @MrBonk: I'm sorry but I've never seen tests, I'm just speaking from my own personal experiences, but just from playing a few different games you can easily feel the difference with and without it. In some games the difference is almost unnoticeable (which is nice when that's the case), in some games it makes the game unbearable to play it with Nvidia Inspector cause you feel the lag so much. It really is a case that differs by game.
     
  9. Agent-A01

    Agent-A01 Ancient Guru

    Messages:
    11,628
    Likes Received:
    1,119
    GPU:
    4090 FE H20
    It's all here.

    http://www.blurbusters.com/gsync/gsync101-input-lag/

    If you have to use out of game FPS limit use RTSS
     
  10. RealNC

    RealNC Ancient Guru

    Messages:
    4,945
    Likes Received:
    3,223
    GPU:
    4070 Ti Super
    See the link by Agent.

    However, here's a test though that you can do yourself. Configure a 23FPS limit in profile inspector. Have RTSS running, but without frame capping (set it to 0). Start a game. Check out the lag.

    Now, alt+tab to RTSS, set a frame cap of 22FPS. Back to the game. Now that RTSS takes over frame limiting (because of the 1FPS lower cap), you will see hugely reduced latency. At these low frame rates, you can observe the input lag very easily; nvidia's limiter will give you about 90ms worse lag than RTSS (at least 2 frames, and each frame is about 45ms.)

    Very easy to test and verify this yourself.
     

  11. RealNC

    RealNC Ancient Guru

    Messages:
    4,945
    Likes Received:
    3,223
    GPU:
    4070 Ti Super
    I'm not aware of a difference. I'm not sure what "feels different" means :p

    How do you select between then?

    Doesn't matter. 0.007, 0.008, 0.009, all these work fine. The point is that it should be somewhere in that range. The exact value doesn't matter. It shouldn't be too low or too high. Anywhere between 0.007 and 0.012 should be fine. If you go higher, it will result in microstutter.

    If you don't mind input lag, sure. The higher the number, the more input lag there is once the game cannot reach the FPS cap. If the FPS is actually reached, MPRF doesn't matter. It only comes into play with uncapped frame rates (which happens if the game falls below 60FPS if you use a 60FPS cap.) This includes uncapped vsync on, btw, not just vsync off.
     
    Last edited: Jun 26, 2017
  12. 321Boom

    321Boom Member Guru

    Messages:
    122
    Likes Received:
    12
    GPU:
    GTX980 Ti
    Haha you're right sorry, I should explain feels different better :p By feels different I mean it's like the game is skipping more frames on Default than on V2 Allow All. You'd notice it cuz it feels like everything (the whole screen and what's going on in the game) 'jumped' or hiccuped for a very brief moment (around 1/4 of a second) but it's very tedious when that was happening. V2 Allow All seems to be working much smoother and better.

    I select between the two resolutions in Nvidia Control Panel -> Change Resolution, same as you'd change from a standard resolution to another. Once I create a custom resolution it just ends up at the top of the list of resolutions with the standard resolutions below it. It's very easy to identify which is the custom resolution with the custom refresh rate I set since it's always at the top of the resolutions list.

    Thanks for this tip, I will DEFINITELY give this a try soon! If it could get RTSS to work better cause of it, it would be so worth it! Thanks again! :D

    Oh ok, then it shouldn't really matter lol, a GTX980 Ti on a shmup should easily keep up with the 60 fps the games require :)
     
  13. Mda400

    Mda400 Maha Guru

    Messages:
    1,089
    Likes Received:
    200
    GPU:
    4070Ti 3GHz/24GHz
    v1 in profile inspector 2.1.3.6+ behaves the same as with profile inspector 1.9.7.6 and pre-368 drivers. Moreover that since Pascal GPU's were released, The frame limiter was updated with a v2 version that caps the frame rate using the third buffer introduced with fast-sync (which is why it adds more delay). Not sure if profile inspector or the Nvidia driver between that time frame is the cause for only applying v2, but its now working as intended.

    v1 does what RTSS does. It slows the render rate at the CPU level instead of the GPU level like with v2. It is more inconsistent in frame-time, but less of a delay.

    v1 is not triggered by fast-sync. You can force vsync off in the NVCP and compare v1 and v2 tearing behavior (v1 should variate where as v2 holds the tearing line).
     
    Last edited: Jun 28, 2017
  14. Martigen

    Martigen Master Guru

    Messages:
    534
    Likes Received:
    254
    GPU:
    GTX 1080Ti SLI
    Interesting, thankyou.

    If you don't use Fastsync (I have SLI, and can't) does V2 serve no purpose then? Or does it work without Fastsync?
     
  15. GroinShooter

    GroinShooter Master Guru

    Messages:
    440
    Likes Received:
    98
    GPU:
    GTX 1080 SC
    How does borderless fit in with all this?

    Meaning is there a difference in input lag etc. running a 80Hz monitor as follows (vsynctester reports Hz at 79.999):

    Fullscreen + RTSS 79.983 fps limit + NVCP Vsync
    VS
    Borderless + RTSS 79.983 fps limit
     

  16. RealNC

    RealNC Ancient Guru

    Messages:
    4,945
    Likes Received:
    3,223
    GPU:
    4070 Ti Super
    Borderless is like fast sync, except it has more input lag. Unless the game in question is allowed to use W10 CU optimizations, in which case borderless is the same as exclusive fullscreen without fast sync.
     
  17. Mda400

    Mda400 Maha Guru

    Messages:
    1,089
    Likes Received:
    200
    GPU:
    4070Ti 3GHz/24GHz
    I personally would not use v2, but its meant for a more consistent fps limit since the GPU controls the limit (something that might help with SLI microstuttering). It works without fast sync. But they probably meant for it to be used with fast sync or even G-Sync so that fluctuations didn't throw off the the timing of each technology.
     
    Last edited: Jun 30, 2017
  18. Mda400

    Mda400 Maha Guru

    Messages:
    1,089
    Likes Received:
    200
    GPU:
    4070Ti 3GHz/24GHz
    The v2 limiter does as it holds the tearing line if you are limiting at/near your refresh rate (for example on a 60hz display, 59 to 62fps limit with vsync off). This wouldn't be possible without adding another frame buffer.
     
    Last edited: Jun 30, 2017
  19. RealNC

    RealNC Ancient Guru

    Messages:
    4,945
    Likes Received:
    3,223
    GPU:
    4070 Ti Super
    It adds two frames of latency, not just one.
     
  20. Mda400

    Mda400 Maha Guru

    Messages:
    1,089
    Likes Received:
    200
    GPU:
    4070Ti 3GHz/24GHz
    Let's agree to disagree.

    Also:
    What was the frame rate limiter option in the profile inspector doing before they added fast-sync to the drivers?

    There has always been a "v1", why would they make it depend on fast-sync?
    It did not work as intended unlike pre-fastsync drivers until either Nvidia fixed it in a recent driver or Orbmu2k modified profile inspector in version 2.1.3.6+ to better distinguish which one for the driver to use.
     
    Last edited: Jun 30, 2017

Share This Page