Lowest input lag on normal 60Hz monitor with fps limiter

[hejonar]

Greetings, I'd like to know which settings should I use to achieve lowest input lag possible. I am using old 1080p 60Hz HDMI monitor with Pascal GPU and setting this just for single application(game) that experiences huge variables in frame rate(it's most likely programmable issue with application because people with best RTX cards report same fps drops as people with low end GPU's while average and max fps are proper for given hardware). Because of those frame rate drops I am forced to lock game to certain frame rate limit to minimize input stutter. So, sometimes game drops to around 30fps but that's brief and rare, more likely it will very often drop to around 60fps even when using GPU's that normally run it at 300fps+. In order to prevent this ugly variable I locked the game at 60fps, but I also want to reduce input lag to lowest possible. I don't care about graphical artifacts like tearing, and I prefer low input lag to stable frametimes. Hence my priority is low input lag>stable frametimes>artifacts. And I have no idea how to measure input lag.

So after checking some guides I went with most basic options and have forced v-sync off and used in-game fps limiter. But is it better than forcing v-sync off and using RTSS for fps limiting, where frametimes look definitely smoother? What about RTSS scanline sync, I've read somewhere that it's as smooth as RTSS frame limiting while having less input lag and in situations where GPU is overloaded it's just turning itself off resulting in normal v-sync off scenario?

Also, which NV Control Panel/Inspector/Windows 10/Game Mode/borderless etc. settings would give me lowest input lag? Is it true that RTSS scanline sync only works when "disable fullscreen optimizations" is selected and when you alt-tab it disables itself until application restart?

[RealNC]

In-game limiters usually reduce input lag more than RTSS does. But not all games are equal. Some games have better in-game limiters than others. The only way to test this without special equipment is to cap your FPS to 20 with the in-game limiter and disable the FPS cap in RTSS (set it to 0 to disable it) and play a bit. Then alt-tab to RTSS and set it to 19FPS. Alt+tab back to the game and see if you get more or less input lag. That way you can tell if RTSS is better than the in-game limiter.

RTSS s-sync is only useful if you can maintain 60FPS at all times, and on top of that your GPU must not be loaded too much. If you can't maintain that, then it's going to result in heavy stutter. So s-sync is not for this game, it seems. You can try it, but it's most probably going to give way worse results than normal FPS capping.

As for anything else, to get the lowest input lag possible, play in exclusive fullscreen mode and make sure you set "Maximum pre-rendered frames" to "1" in the nvidia panel's 3D settings section.

[hejonar]

In-game limiters usually reduce input lag more than RTSS does. But not all games are equal. Some games have better in-game limiters than others. The only way to test this without special equipment is to cap your FPS to 20 with the in-game limiter and disable the FPS cap in RTSS (set it to 0 to disable it) and play a bit. Then alt-tab to RTSS and set it to 19FPS. Alt+tab back to the game and see if you get more or less input lag. That way you can tell if RTSS is better than the in-game limiter.

I've heard about it before but I have hard time testing it. Those input delay differences aren't that big and hence hard to catch. Is there any way of checking it? I know that Overwatch game had some option that allowed to display "SIM", perhaps there is third party tool that would be able to check this for other applictions?

RTSS s-sync is only useful if you can maintain 60FPS at all times, and on top of that your GPU must not be loaded too much. If you can't maintain that, then it's going to result in heavy stutter. So s-sync is not for this game, it seems. You can try it, but it's most probably going to give way worse results than normal FPS capping.

Actually to test it I have used "x/2" option for 30fps. Outside those fps drop spikes my GPU usage was at around 20-30% with 30fps lock. I was more interested what technically happens when GPU is unable to keep up with s-sync. Does s-sync turn itself off, does it affect input lag, does it affect frametimes(I mean s-sync effect itself not GPU slowdowns). Additionally, assuming that my GPU would be able to keep GPU usage at stable 50% in 60fps scenario, would s-sync offer better input times than just plain RTSS framelimiter?

[RealNC]

I've heard about it before but I have hard time testing it. Those input delay differences aren't that big and hence hard to catch. Is there any way of checking it? I know that Overwatch game had some option that allowed to display "SIM", perhaps there is third party tool that would be able to check this for other applictions?

You can't measure input lag through any kind of software. You need to be measuring the time between input and when the pixels change on the screen. Without a setup to do that, like a 1000FPS camera to record the screen, all you can do is make frame times as large as possible by applying a very low cap. Like 20FPS, or even lower. At 20FPS, a difference of one frame of latency equals 50ms. That is high enough to be felt without the need to measure it. If you can't feel a difference between an in-game 20FPS cap and an RTSS 20FPS cap, then I'd say there's no big difference between RTSS and the in-game limiter.

Unfortunately, many in-game limiters can't be configured to such a low FPS cap. So this method isn't usable with all games. Sometimes you can tweak an ini file to set a 20FPS cap though. My method is to set the in-game cap to 20FPS, get a feel for the input lag, then quickly alt+tab to RTSS, change from "0" to "19" (so that the RTSS cap overrides the in-game cap) and quickly alt+tab back to the game and see if I feel a difference. If not, then it's a safe bet to say that RTSS is just as good as the in-game limiter in that game.

Actually to test it I have used "x/2" option for 30fps. Outside those fps drop spikes my GPU usage was at around 20-30% with 30fps lock. I was more interested what technically happens when GPU is unable to keep up with s-sync. Does s-sync turn itself off, does it affect input lag, does it affect frametimes(I mean s-sync effect itself not GPU slowdowns). Additionally, assuming that my GPU would be able to keep GPU usage at stable 50% in 60fps scenario, would s-sync offer better input times than just plain RTSS framelimiter?

S-sync will not turn itself off. It will still sync to the monitor's vsync, but if the frame rate target is not reached (30FPS in x/2 mode, 60FPS in normal mode) then you get heavy stutter because the sync will now happen on the next vsync. It looks horrible. Input lag doesn't even matter at that point, because the stutter makes the game borderline unplayable. You can test that by playing a more demanding game, or by bumping the graphics settings of the game to ultra so that 60FPS can't be maintained and/or GPU load gets high. S-sync will just result in unplayable stutter in that case.

And yes, if your GPU can keep up, then s-sync will result in the lowest input lag possible. Input lag will be as low as vsync OFF + normal 60FPS RTSS cap (which also happens to be the same as g-sync or freesync; basically RTSS s-sync kind of does what g-sync/freesync do when it comes to syncing, except it does it through software and outside the driver. But contrary to gsync/freesync, s-sync falls apart when the framerate drops or when the GPU can't flush the frame buffers fast enough.)

[hejonar]

Thanks, this quite explains a bit. After some tests I guess I will stick to normal in-game limiter, even while using it I can feel some delay. I guess next step would be running game at higher frame rate, but it's almost impossible because of frame rate drops. I have also noticed small thing when using ssync or rtss when compared to ingame limiter, that is CPU usage. My results might be somehow skewed as I didn't test exactly same scenes but it still seems that ingame option has lower CPU usage, in all cases there's 30fps lock(ssync is x/2 on 60Hz monitor). Also in all cases I have selected parts where fps drops *almost* didn't occur.

https://imgur.com/a/0hp1j7X

SSYNC:


INGAME:


RTSS:

[RealNC]

Higher CPU usage with RTSS is normal, and it doesn't have ill effects. The CPU usage is due to RTSS doing an idle loop for better frame timing accuracy. This generates a kind of "fake" CPU load. It doesn't result in high temperatures or anything. It just loops the CPU doing nothing, but this still registers as CPU load. Basically, RTSS loads the CPU with nothing. It just kills time. The game is blocked by the RTSS frame limiter anyway, so this doesn't take away CPU resources, as the game is prevented from doing anything anyway by the frame limiter.

It's normal.

[hejonar]

What about setting my monitor refresh rate higher or lower, would that affect anything? For example I can lock fps at 30 and set monitor to refresh at anything from 30 to 75Hz, would using 30fps lock and 30Hz be better than 30fps lock and 60Hz or 75Hz? Same question for 60fps lock. Obviously I don't use any form of vsync.

[RealNC]

If FPS does not match Hz, you will get some stutter from time to time, unless you use 1/2 vsync in nvidia inspector to sync to half the refresh rate. So 30FPS @ 60Hz 1/2 vsync is as good as 30FPS @ 30Hz.

You can also use RTSS scanline sync x/2, which is even better (if it works.)

[hejonar]

I am not using any form of sync, I was more interested in knowing would it affect input lag. Also, I've heard about frame skipping issue, so I guess that running 30 or 60fps at 75Hz would be worse than running 30/60fps at 60Hz?

[RealNC]

Without sync, you can pretty much cap to whatever you want, and the higher your Hz, the better.