Blur Busters Forums

jorimt wrote:Sounds like something weird is going on, or the built-in meter isn't accurate, and is giving false "100 Hz" positives; I'd defer to RTSS readouts in this instance. Either way, if you're not sustaining 100 fps/Hz for long periods, and it simply spikes infrequently, there isn't going to be time for noticeable input latency to build up, theoretically.

Afterburner/RTSS only updates every 100ms or so, not every frame/refresh like the monitor's FPS counter does.
I trust the monitor's FPS counter since it will tear as soon as it hits 100 FPS on that if I have V-Sync disabled, yet RTSS doesn't catch it.

I recorded a short video of Dishonored 2 when I got the monitor showing how slow Afterburner is to update compared to the hardware display, since people were complaining about Afterburner reporting good framerates/frame times but the game appearing to stutter badly:

Dishonored 2 is an exceptional disaster when it comes to frametime delivery. If at any point it misses its internal target (which it does frequently, regardless of the system running it; the game's poor asset streaming management doesn't help either), it outright skips or repeats frames. The engine is a stutter-y mess.

It's still strange that you're seeing the built-in meter hit 100 in other games, however, and even with a hard fps limit set below it. I have a built-in meter on my 144 Hz display, and it never goes so far over the set fps limit in any game. Not certain what's going on, at least with the information you've provided thus far.

jorimt wrote:Dishonored 2 is an exceptional disaster when it comes to frametime delivery. If at any point it misses its internal target (which it does frequently, regardless of the system running it; the game's poor asset streaming management doesn't help either), it outright skips or repeats frames. The engine is a stutter-y mess.

Yes, I was just showing how much of a discrepancy there can be between what RTSS reports and what the hardware monitor shows, and how much slower RTSS is to update - so it would completely miss any time that it hits 100Hz for a single frame.

The performance in that game is so disappointing though.
id Tech 5 games run perfectly fine, and they broke that when making the "Void Engine" from it.

jorimt wrote:It's still strange that you're seeing the built-in meter hit 100 in other games, however, and even with a hard fps limit set below it. I have a built-in meter on my 144 Hz display, and it never goes so far over the set fps limit in any game. Not certain what's going on, at least with the information you've provided thus far.

The easiest way to reproduce it is to find a scene which is GPU-limited, and then quickly point the camera straight at the floor.
That will momentarily push the framerate way above RTSS' framerate limiter.
That's not the only time it happens - I originally saw it happen during gameplay when I was limiting to 95 FPS. It's just that doing this makes it easier to reproduce consistently.

Trying to avoid 100Hz is also easier to monitor than trying to avoid 144Hz, since you're looking out for a two-digit refresh rate counter turning into a three-digit counter which is very easy to spot compared to looking for a specific value.

I suppose the question is whether it matters that a game briefly hits the refresh rate cap and engages V-Sync, or if it only matters to keep the game below it 99% of the time instead of reducing framerate by 10%.

You can increase the OSD update interval in RTSS by editing the "Profiles\Global" file in the RTSS directory:
N is the interval in milliseconds. Default is 500 (half a second.) 50ms should be fast enough for you case.

Is there somewhere that documents all these hidden options for RTSS?
I tried setting it to 1ms, and it still doesn't catch these fluctuations.
Loaded up Dishonored in a spot where I can just flick the mouse back and forth and cause the monitor to report 88-94Hz, meanwhile RTSS just reports a constant 91 FPS / 10.9ms frame time. (the framerate limit set in RTSS)

1ms is 1000FPS. So unless you have a 1000FPS monitor, 90% of the updates are not visible, they just eat CPU. Why don't you try 10ms, which is 100FPS.

Also, just for testing, try the nvidia frame capper (through Inspector.) It adds more input lag, but this is just a test. Does it have the same issue? (Don't forget to set the RTSS cap to 0, which disables it.)

"Or unless you're using VSYNC OFF or triple-buffering with older games"

You knew that already, just making sure that base is covered for other readers -- because 1000fps can benefit in reducing lag for VSYNC OFF and triple-buffering(freshest-buffer technique, not queue technique) situations, for older game engines such as CS:GO capable of running at that framerate on modern GPUs.

Even though It's not applicable here for this situation.

@Chief: true, but in this case just making sure the RTSS overhead of the OSD doesn't interfere with performance.

Another thing to test is a different FPS indicator overlay. Try Fraps or Dxtory too. If all of them report the same frame rate, then we have to assume that there's something wrong with the monitor's FPS indicator.

If you really want to dig deeper into this, then you can use FRAPS and its "frame time benchmark" feature. Then, use this:

https://sourceforge.net/projects/frafsbenchview/

To get a graph out of the FRAPS frame times file. This should show you if the game truly hits 100FPS even though you capped it lower.

This is still software based of course, so it's still not 100% fail-safe. The really "true" method of testing this would be an FCat setup, but that's not something I expect people to just have sitting around in their house... This would be up to review sites to test. But lacking that, a FRAPS benchmark should be "good enough."

I'm still not sure why you think the display's monitor would be wrong, since it consistently tore when V-Sync was disabled as soon as that hit 100Hz.
All it does is report the refresh rate that the panel is currently updating at.

Anyway:
I removed the custom update rate for RTSS and set the framerate limiter to 91 FPS.
V-Sync was enabled in the NVIDIA Control Panel and disabled in the game.
Performance was measured with FRAPS and graphed with FRAFS.



I don't really like FRAPS data, as the place it measures in the rendering chain does not seem representative of what ends up on your display.
However it clearly shows the game running well above the 91 FPS limit that was set by RTSS.


I checked out both the NVIDIA Framerate Limiters, and while I did not get any FRAPS measurements, the display's refresh rate monitor was absolutely locked to 91Hz.
Nothing I did caused it go above 91Hz for even just a moment.

It also seemed to produce far more consistent performance overall.
With RTSS limiting to 91 FPS, the framerate was constantly hovering around maybe ±5 FPS of that target.
It seemed that any time it exceeded the limit, it would then drop a similar/equal amount below the 91 FPS target immediately after that.
The NVIDIA Framerate Limiters did not have that behavior. It still dropped to 87/88Hz very briefly once or twice, but it was almost a constant 91Hz for the entire test, while the refresh rate was constantly fluctuating around 91Hz when using RTSS.

Hm, my last guess is RTSS 7.0.0 isn't working as intended on your specific hardware configuration, because RTSS is usually the most stable framerate limiter out there (within <1 frame fluctuation). You could try rolling back a version (6.6.0), and see if that solves the issue.