Is it better to cap frame rate with RTSS or NVCP v3 Limiter?

[DemaRCO98]

This question popped into my head after watching Battle(non)sense’s video about this here:

It’s been 3 years since he posted that video, and now I’m wondering if anything has changed since then.

With all the updates the RTSS received, and NVIDIA drivers released, have these frame rate limiters improved in a way so that the input lag is now closer to that of the in-game’s FPS limiters?

Also, what about frame times? Which of these two provides more consistent and stable frame times?

I don’t have proper tools to monitor input delay, and as frame times are concerned, I can only use MSI Afterburner’s OSD, so I was hoping you can help me with this, as it seems you’ve done plenty of testing with these tools,

So, out of these two, which one provides the least amount of input lag, and consistent, stable frame times?

 

[natr0n]

I do it in the driver and rivatuner for redundancy.

There been a few times where driver setting didnt take and rivatuner always works for limits.

Use the same limits of course.

frametime is easily understood like this
60fps=16ms/ 120fps=8ms/ 240fps=4ms

It will dip during loading and such, overall if its consistant nothing to worry over.

Testing with online games is not a good idea too many variables test with a offline/ racing game for example

 

[Klemc]

Special-K can have handy OSD ?.. for frametimes...

 

[droopyRO]

If you hit Alt+R with a Nvidia GPU, you get a nice graph that shows you your latency, no need to install additional software.

 

[Chomiq]

If you hit Alt+R with a Nvidia GPU, you get a nice graph that shows you your latency, no need to install additional software.

That's IF you have geforce experience installed.

I stick to ingame frame limits. Ready or Not has a handy latency stats and it clearly shows that hard locking to 141 fps on a 144 Hz display with RTSS gives extra 2-3ms of lag in comparison to ingame fps to limit set to 141.

 

[droopyRO]

Yes, but i guess that over 99% percent of Nvidia users have it installed by default.

 

[Klemc]

Yes, but i guess that over 99% percent of Nvidia users have it installed by default.

What can we do with GFE ?.. i don't want it but i don't know all about it.

 

[DemaRCO98]

Testing with online games is not a good idea too many variables test with a offline/ racing game for example

Hmm, that's interesting.

Part of the reason why I asked which of these two is better is because I wanted to see which one would provide better game responsiveness (that is, less input latency), especially in competitive games.

This idea came to me from one of Battle(non)sense's videos which he made back in 2019 here:

TL;DW: In that video, he concluded that Ultra Low Latency Mode reduces input latency only while you're playing games with high GPU usage (97-99%), whereas if you drop below that value, the ULLM does nothing, and can sometimes have the opposite effect, where it introduces more input lag.

However, he also found out that if you limit your game to a certain frame rate that your PC can maintain consistently, you will get an immensely lower input lag as compared to running the game without a limiter (even with ULLM), and not only that, but at a lower GPU load as well.

Now for me for example, some of the games that I play are: CS GO, Starcraft 2, fighting games, DotA 2, etc.

Most of them aren’t GPU bound, in fact, only a few of them are.

For example: Starcraft 2, and many fighting games are locked to 60 FPS by default, and my GPU usage in those games hovers around 22-50% (depending on the game), while CPU usage percent varies across different games and different cores, as some of these games utilize either only one core, two cores, or all 4 of them.

One thing that I noticed on these 60 FPS-locked games while testing both RTSS and NVCP MFR was that:

• Without limiters, the frametimes fluctuated around the 16.6ms range by 0.1-0.9ms (meaning every second, or split second it went from, 16.6 to 16.5, 16.7, 16.2, 16.9, 17.2…)
• When either of the two limiters were enabled, and set to 60 FPS (even tho the games are already capped at that value) the frametimes immediately stabilized at 16.6ms and didn’t move at all from there. Like, the OSD frametime graph was a beautiful flat line, and never had any spikes in it

Again, most of the tests were monitored via MSI Afterburner’s OSD as I, unfortunately, lack the dedicated tools to test input latency to see if frame-limiting these games does indeed have an effect on reducing input lag, as I would really like to know.

If you hit Alt+R with a Nvidia GPU, you get a nice graph that shows you your latency, no need to install additional software.

Do you mean the Render Latency option?

If so, I don't think that's the sort of thing I'm looking for because as far as I've read, the Render Latency is just an option that displays how long it takes for a frame to go from sitting in the frame buffer to it being rendered. It doesn't say anything about input latency, and how many milliseconds it takes from a button click to register on the screen.

So, with all of this in mind does anyone know if frame-limiting improves latency in FPS-locked games that have low GPU load, and does the Low Latency Mode feature of NVCP also have any effect on input latency (whether set to On, or Ultra)?