Mirror refresh rate and frametime uneven graph

I did some experiments with the new mirror refresh rate limit option. I also have 140fps in game limit set.
I can say that the new option has good impact on the gpu load (the lower mirror refresh rate is the lower the gpu usage is). But it also have some impact on the frametimes "distribution", average frame time value probably is the same, but the frames are more uneven. See the first screenshot (mirror refresh limit off) - the frametime graph is almost flat, and see the others - clearly there is a difference.

I dont feel stuttering or similar in game. I post here just as an interesting observation. I kind a like more the flat line than the shaggy line with mirror refresh limit


off (same as before acc 1.6):


auto:


20:


30:


60:


90:

 

Stands to reason that the frames when the mirrors are rendered will take longer.

 

I've always found RivaTuner the only way to get a smooth frametime graph.

I dont have a video in ACC using it but i do have this, as you can see without it cpped its rough but capped its a straight line Better than in game caps also.

 

Using RT cause up to one frametime delay. I always prefer to use the ingame fps limiter.

 

Really, I never noticed, lol

 

There are many users of RTSS which can confirm you are getting less input lag by capping with RTSS than many ingame limiters. There might by one game or two which does it even better, but general speaking you can't go wrong using RTSS to limit FPS.
The latest RTSS beta has massive improvements for fps limiting:

you may even head over to blurbuster.com to read RTSS Scanline Sync HOWTO - Blur Busters Forums

 

I set it to 60 FPS. Should be good to know how much frames are being rendered in "off" mode. I can´t notice uneven frametimes with 60 though.

 

There are no frames rendered in "off mode" as you say. When using RTSS to limit fps it holds the render queue until the time has passed to render a new frame. That's why the gpu utilization goes down. It has less to do.
Ingame limiter can do the same, but they also could use that "waiting" time to do other stuff, like AI, physics, etc... better.

The thing is,ingame limiter mostly working on a timer event message basis. The render thread gets a notification "He, time is up!" and starts to present the frame then goes to the next frame. This method is a tiny bit imprecise why you get some little jitter in the frametime graph.
RTSS also can do this method, which unloads the cpu (wait state). But it can also do high precision frame limit, then using a busy loop to wait for the time to pass. This keeps the cpu in a somewhat loaded state and the cpu utilization is displayed high up to 99% but RTSS doesn't use the cpu exclusively, it's not hogging it.

 

I had previously used nVidia Inspector to cap my FPS in some games that RTSS didnt work (eg.Wolfenstein II) nd i could really feel input delay while using that compared to the in game vsync.

 

But you know that RTSS is capable of displaying OSD in Wolfenstein II?
You only have to enable it via profile config file and set PresentFromCompute=1 in the section [RendererVulkan]

It was disabled by default, because to display the OSD means to flush the compute queue and this results in a little performance hit, which many "normal" users then would complain about, because they don't know why.

 

I didn't know that but I've played through that game now anyway. What I do remember is I had to use inspector because nothing else removed tearing. That was my main goal at the time.