NVIDIA’s New Control Panel FPS limiter VS Rivatuner VS In-Engine

Discussion in 'Videocards - NVIDIA GeForce Drivers Section' started by BlindBison, Jan 7, 2020.

  1. BlindBison

    BlindBison Ancient Guru

    Messages:
    2,404
    Likes Received:
    1,128
    GPU:
    RTX 3070
    Hi there guys,

    I read here: https://www.engadget.com/amp/2020/01/06/nvidia-geforce-drivers-framerate-cap-ces-2020/

    That Nvidia has added FPS limiting to their control panel in their latest driver. Is this one of the limiter options from Nvidia Inspector or something new?

    Something I wonder about is how it stacks up against Rivatuner and In-Engine limiters.

    My understanding is things used to breakdown as:

    In-Engine > RTSS > Nvidia Inspector for input lag, where RTSS was known for the best framepacing/frametimes.

    Have any new tests yet been done with the new limiter from Nvidia’s control panel? I haven’t been able to find any as of yet.

    Thanks for your time,
     
    Robbo9999 likes this.
  2. Mda400

    Mda400 Maha Guru

    Messages:
    1,087
    Likes Received:
    199
    GPU:
    4070Ti 3GHz/24GHz
    As far as the new control panel option goes, It does not seem to be the same.

    Inspector seems to have a different Frame Rate Limiter option exposed.

    Tested with Gears 4, it seems to be limiting at the CPU level, but i'd also wait to see what those with high-speed camera will come up with when testing.

    RTSS still has the scanline sync option, which is more accurate than time-based frame limiting.

    [​IMG]
    [​IMG]
    [​IMG]
    [​IMG]
     
    Last edited: Jan 7, 2020
    BlindBison likes this.
  3. janos666

    janos666 Ancient Guru

    Messages:
    1,645
    Likes Received:
    405
    GPU:
    MSI RTX3080 10Gb
    Some preliminary test results suggest the V1 driver limiter has the same 0-1 frame latency as RTSS since 441.08 (when LLM Ultra started applying an auto fps cap for G-Sync+V-Sync): https://forums.blurbusters.com/viewtopic.php?p=45250#p45250
    So, in-game limiter should be superior (when available and works as expected). And I think this driver limiter doesn't work with DX12/Vulkan either.

    [​IMG]
     
    Robbo9999 and BlindBison like this.
  4. BlindBison

    BlindBison Ancient Guru

    Messages:
    2,404
    Likes Received:
    1,128
    GPU:
    RTX 3070
    @janos666 @Mda400 Thank you very much! All of that info/the screenshots are very helful.

    One thing I have been wondering about for sometime now is why the Ultra Low Lag mode limits you to 138 rather than 141 as suggested by BlurBusters. Is it just to be on the very very safe side or what?

    Also, strange that the new Nvidia Driver Limiter doesn't work with Vulkan/DX12, huh -- I mean, RTSS does so surely it should be possible.

    Thanks again, all that is very helpful to know, much appreciated.
     

  5. Mda400

    Mda400 Maha Guru

    Messages:
    1,087
    Likes Received:
    199
    GPU:
    4070Ti 3GHz/24GHz
    My speculation on the ULLM frame cap is that Nvidia themselves may have took notice of what BlurBusters discovered?

    DX12/Vulkan are low-level API's and disregard high-level options such as those set in the driver software.
    The application's developer has more control over what happens with various hardware to make it perform similar between different manufacturer's hardware.
    But they have to be more careful/have more knowledge about what they are coding for each manufacturer's hardware.

    RTSS is likely hooking/injecting itself to the application at a low-level, to allow control of this feature. Some programs may treat this as a sign of a hack.
     
    Last edited: Jan 7, 2020
    BlindBison likes this.
  6. janos666

    janos666 Ancient Guru

    Messages:
    1,645
    Likes Received:
    405
    GPU:
    MSI RTX3080 10Gb
    Those numbers are not some de facto standards but mere suggestions, so yes, the very first implementation by nVidia is on the very safe side. But a few drivers later they gave us manual control over the limiter (although I didn't check if that new parameter overrides the G+V-Sync auto-cap or Ultra overrides the manual setting in that case with the old auto default because I am testing the old 445.23 right now).
     
    BlindBison likes this.
  7. -Tj-

    -Tj- Ancient Guru

    Messages:
    18,095
    Likes Received:
    2,601
    GPU:
    3080TI iChill Black

    Hi,

    it does indeed have 1fps fluctuations, I tested it vs internal fps limiter by Destiny2. In one moment I saw it shoot up to 169fps, but only in one loading section, otherwise it was 144-145 by those scenarios, while gaming it never went over fps cap, I used 143fps driver cap.


    Just a quick question about RTSS scanline sync, does it still work if overlay doesn't work, e.g. by Destiny2? I really want to give that a try too, so for 143-144fps cap, I need it to set to? Will it still sync bellow that, idk if i have 70-80fps drops?
     
  8. Cave Waverider

    Cave Waverider Ancient Guru

    Messages:
    1,876
    Likes Received:
    662
    GPU:
    ASUS RTX 4090 TUF
    The new Nvidia FPS limiter introduced with the 441.87 drivers seems to work fine for me in Red Dead Redemption 2 under Vulkan and DX12, so it doesn't seem like using a low-level API itself is the issue.
     
    Last edited: Jan 8, 2020
    Smough, Mda400 and BlindBison like this.
  9. Ital

    Ital Master Guru

    Messages:
    202
    Likes Received:
    86
    GPU:
    ZOTAC GTX 1080 AMP
    The newly updated NVIDIA Profile Inspector 2.3.0.3 added a new setting "Frame Rate Limiter V3" which in all likelihood refers to the new FPS limiter in 441.87.

    The previous "Limiter V2" option is still there in the "Frame Rate Limiter Mode" setting, so that could mean that the FPS limiter in 441.87 is a new (or updated) implementation compared to the previous V2 one.
     
    Mda400 and BlindBison like this.
  10. Smough

    Smough Master Guru

    Messages:
    983
    Likes Received:
    302
    GPU:
    GTX 1660
    That's really interesting. I will install this driver when I have the time and I will report back how it stacks against RTSS and nVidia Inspector frame limiter V2, which is the next best thing after RTSS since it also gives perfect frametimes, but at some games it does strange things.

    Maybe nVidia finally stepped their game up and their new limiter it's actually efficient.
     
    BlindBison likes this.

  11. Mda400

    Mda400 Maha Guru

    Messages:
    1,087
    Likes Received:
    199
    GPU:
    4070Ti 3GHz/24GHz
    I feel silly, I don't know why i said that after i tested the frame limiter in a low-level API!... i had forgotten i even did that.
     
  12. janos666

    janos666 Ancient Guru

    Messages:
    1,645
    Likes Received:
    405
    GPU:
    MSI RTX3080 10Gb
    Keep in mind that RTSS can't self-validate it's own performance. That is, the frame time graph will represent the "what ought to be" when the RTSS limiter is in use. It will draw the graph it targets with it's limiter. That's why the RTSS graph might looks better with the RTSS limiter compared to the in-game limiter even if the in-game limiter is factually superior (in case of the in-game limiter, the RTSS graps still track the real world rather than the expectation).
     
  13. jorimt

    jorimt Active Member

    Messages:
    73
    Likes Received:
    69
    GPU:
    GIGABYTE RTX 4090
    ^ this.

    I try to remind others about this fact when they're saying how much better the frametime graph looks when using RTSS to limit FPS, when in fact, while yes, RTSS does have extremely stable frametime performance, in actuality, even it can't provide frametimes stable enough for a perfectly flat graph; it can't read fluctuations happening above its own limiter.

    We also have to keep in mind that using the system to monitor itself subjects the results to further possible inaccuracies as well; Afterburner is far better than nothing where monitoring is concerned, but it can only get you so far. You're eyes are actually ultimately a better meter for framerate/frametime performance where what actually ends up showing on the display is concerned.
     
    BlindBison likes this.
  14. Unwinder

    Unwinder Ancient Guru Staff Member

    Messages:
    17,083
    Likes Received:
    6,567
    Nonsense, sorry. There is no any "real word" vs "expectation".
     
  15. Smough

    Smough Master Guru

    Messages:
    983
    Likes Received:
    302
    GPU:
    GTX 1660
    That absolutely makes zero sense. Where do you get this "theory" from? I can tell with my eyes alone. Any in.-game limiter is trash, for not using other word, both RTSS frame rate limiter and nVidia inspector V2 frame rate limiter give a flat 16.6 ms frame time line at any game, in-game fps limiter or just v-sync gives a horrible frame time or frame pacing with micro-stuttering. Windows Full Screen Optimizations also manage to do this role quite well, makes sense since it literally mixes FullScreen with Borderless. But they can introduce strange CPU usage bugs in certain games. What I am saying is that MSI Afterburnber frame time graph IT'S accurate. When I feel a game stuttering for any reason, it's very well reflected on the graph. When it's smooth, the graph is as flat as a pancake.

    You are saying isn't accurate without any valid proof. No in-game limiter it's factually superior, ever. Not any chance. They are literally a mess.
     
    RodroG likes this.

  16. jorimt

    jorimt Active Member

    Messages:
    73
    Likes Received:
    69
    GPU:
    GIGABYTE RTX 4090
    Ouch, guys; did I severely misunderstand @janos666 comment (if so, my bad), or are you beating him up on possibly poorly worded semantics?

    Because I took this:
    As him inferring that the RTSS frametime graph can't reflect any fluctuations occurring above its own FPS limit (which, as far as I'm aware, is true; feel free to correct me if I'm wrong @Unwinder).

    And this:
    As him inferring that if the in-game limiter did have superior frametime performance to the RTSS limiter (not that they ever do), that the frametime graph wouldn't necessarily accurately reflect this, as it would show any fluctuations occurring above the in-game limit on the graph, but not when using RTSS to limit FPS in the same instances, making the RTSS limiter's graph, in effect, look flatter than it actually is (so long as the framerate is sustained at the set limit in both scenarios).

    That said, I can see where this would be inaccurate phrasing:
     
    BlindBison likes this.
  17. Robbo9999

    Robbo9999 Ancient Guru

    Messages:
    1,855
    Likes Received:
    442
    GPU:
    RTX 3080
    This is a really interesting thread, because the past 3 days I've been curious about RTSS vs this new NVidia driver fps limiting in reference to input lag and frametime stability vs the in-engine BF1 limiter. I'm testing in BF1, and RTSS frame rate limiting provides ultra stable frametime (completely flat graph). The in-game engine limiter is completely unstable and sees fluctuations of +- 20fps around the target framerate (fps monitoring tools will show stable fps but the RTSS graph will show these wild fluctuations). Mouse movement felt a little off with RTSS framerate limiting though, I don't think I'm playing as well with it activated. I'm gonna try NVidia Control panel framerate limiting and I'll update this post.

    EDIT/UPDATE: Sticking NVidia Control Panel on Ultra Low Latency and not using any other framerate limiters seems to be working best for me. On a 180Hz monitor Ultra Low Latency results in it capping the fps at 171fps automatically and the frametimes in RTSS graph are perfectly flat...mouse movement also feels better in game. This is good because it means the preferred option for me is just sticking it on Ultra Low Latency and the NVidia driver is choosing the lowest possible latency settings overall then it seems, as well as providing a really stable and consistent frametime. I'm still looking forward to the "official testing" on renowned websites re input lag differences & frametime stability differences....I want to make sure I'm using the best option for my G-sync screen, but I have a feeling that this new NVidia Driver is a bit of a game changer when it comes to limiting framerate for low latency...in as much as we probably won't need external or in-game limiters. As I said this limiter is way better than the in-game limiter for BF1 which shows large frametime variances.

    EDIT #2 (a couple of days later): tried the BF1 in-game fps limiter again and frametimes almost as stable as limiting through NVidia driver or through RTSS. Not sure why the in-game fps limiter was producing such unstable results a couple of days ago. I've decreased fps to 171fps down from 178fps, 171fps was the "auto-cap" that NVidia 'recomends' if you use Ultra Low Latency. Oh yeah, the other difference was that I'm now not selecting "Low Latency On" in the NVidia driver too. It's one or a combination of those things that is letting the in-game fps limiter perform very stable for frametimes....this theoretically should be the lowest input lag option too, ie limiting with in-engine limiter. That's how I'm gonna run it from now on then, and in games that don't have a good in-game limiter then I'll use the NVidia Control Panel frame limiter.
     
    Last edited: Jan 11, 2020
  18. Robbo9999

    Robbo9999 Ancient Guru

    Messages:
    1,855
    Likes Received:
    442
    GPU:
    RTX 3080
    Auto fps capping is what I observed and talked about in my previous post. Thanks for this post of yours because I didn't realise until I read it! So I tried it...and it works well (described in my previous post). It works better than the in-engine BF1 framerate capping, more consistent frametimes, and out of all the different frame capping option LLM Ultra in the driver with no other caps feels best with the mouse, it auto caps at 171fps on my 180Hz monitor = job done!
     
  19. janos666

    janos666 Ancient Guru

    Messages:
    1,645
    Likes Received:
    405
    GPU:
    MSI RTX3080 10Gb
    I am not sure how well Ultra works with Frostbite, even in DX11. I don't have any recent BattleField games but used to play a lot with Bioware games and was always unsatisfied with Dragon Age Inquisition's smoothness (lots of areas with reproducible stutter), so I keep that game around for testing. Even though Ultra caps my framerate to 117 in DAI, changing the RenderDevice.RenderAheadLimit from default -1 (assumed to result in 2) to 1 while the game is running (started with Ultra) has a clear immediate effect on the GPU utilization (drops from ~99% to ~86% or so, assuming a taxing scene with <117 fps) and measured fps (the decrease is comparable to the drop in the GPU utilization) which makes it seem like Ultra can't override the queue length (well, may be unless Ultra did change the queuing behavior during the program's start but this parameter now restored the regular behavior but I don't think that's the probable answer). This was the same back in the days when we had the old prerender limit setting in the driver. These Frostbite games (DAI, MEA) seemed to ignore it.
     
  20. Robbo9999

    Robbo9999 Ancient Guru

    Messages:
    1,855
    Likes Received:
    442
    GPU:
    RTX 3080
    In BF1 the Ultra Low Latency settings doesn't affect the fps, I always have GPUz Sensors open during a gaming session, and average GPU utilisation is always the same at about 86-89% for a whole gaming session, I also have a framerate counter on screen whilst playing, and I'm not seeing lower fps. Ultra Low Latency doesn't seem to have any negative affects on my rig in BF1.
     

Share This Page