Madvr report randomly incorrect refresh rate resulting in lots of framedrops

[VTS]

Can someone help me find out what's going on in my core i7 + geforce 660ti rig.

I have been using mpc-hc+madvr+lavfilters+CUVID now for one year without any problem.
Primary display is Asus 22 monitor. Nvidia displaydriver is set to 72Hz refresh rate.
Second display is optoma hd33 projector and displaydriver is here set to 23.976Hz refresh rate.
Audio is bitstreamed forward to home theater amp and reclock is not in use.

Few weeks ago something in my system changed. And so far I don't know what.
I notice massive framedrops and stuttering when playing 23.976 mkv files on second display, fullscreen or windowed.


Madvr OSD has always been useful and now it show very high clock deviation. OSD display line
show totally wrong framerates. Also those values are different after every win7 reboot when playing same mkv clip.
Values are like this on primary disp:
reboot1 (disp=139Hz, Clk dev=93%)
reboot2 (disp=117Hz, Clk dev=63%)
reboot3 (disp=129Hz, Clk dev=79%)

The real problem is second display: disp=36Hz, clk dev=46%.
At the same time Optoma projector on second disp show always right refresh rate 24Hz(23.976) and it would never sync to 36Hz.
Same for primary disp Asus. It would never sync to anything over 72Hz.

It seems that refresh rates nvidia is outputting to displays are ok.
But somehow madvr or evr-cp renderer, both report them wrong.

I have updated mpc-hc, madvr, LAVfilters and nvidia drivers to newest versions but the problem still remain.
Unfortunately because of SSD optimizing I do not have system restore turned on to give a way to revert back to older working state.

Last year before this occurred I had very smooth rendering with clock deviation well under 0% to second display and no framedrops when
playing 23.976 MKV files.

I would be thankful if you experts on this forum could give some hints how to fix this problem.

[nevcairiel]

Some other people have reported the same problem recently. It makes me suspect that a windows update or gpu driver update broke it.
You wouldn't happen to have "adaptive vsync" activated in the nvidia driver, would you? (globally for all applications, that is)

[Plutotype]

Hi,
Cyberbeing saved me on slysoft forum, I will try save you:

https://forum.slysoft.com/showthread.php?58710-Having-Weird-Issue-With-Wrong-quot-Display-quot-and-quot-Refresh-Rate-quot-Getting-Desperate

In short:

1. Make sure you have HPET activated in BIOS ( 32-bit for 32-bit OS, 64-bit for 64-bit )
2. With HPET enabled, boot into desktop, start cmd.exe as admin and paste following command:
bcdedit /set useplatformclock true

Click enter to execute it. This will activate HPET in Windows as well.

3. Reboot

Let us know if it worked please.
Thanks

[VTS]

Thanks Plutotype, you succeeded in saving me.

I was not aware that in modern chipsets exists this kind of event timer.
Asus P8z77-v bios had HPET enabled. I only activated HPET in windows, rebooted and that’s it! Now madvr OSD show right refresh rate to both displays and clock deviation is 0.00xx%.

I have manual control over updates. When this problem occurred gpu was using 314.22 driver. And no windows updates have been performed during that time.
But, I did a game installation at that time and directx log show some files had been installed. Tried once to update it again but
installer refused to do it due to already having newest version.

Interesting that others have recently reported about same clock problem.

It would be nice to now the root cause why this happened.
Thank you all for posting. I would never have solved this on my own.

[Plutotype]

Well, it was a nightmare for me as well. I was asking a question - what software or hardware component translates the refresh rate readings and passes them to Windows/programs incorrectly? That must be the missing screw! :-)

Actually the fear of a clean OS reinstall with 100 apps and full customisation always motivates you to find a effective solution for any weird issues like this. So the help from other users came right in time.

[cyberbeing]

Glad to hear that forcing HPET worked for both of you.

I originally suggested it, since I've seen my average refresh rate fluctuate by small amounts +/- 0.2Hz between reboots with the default Windows timer. For a couple years now I've been forcing HPET since it always gave me an identical refresh rate reading each reboot. It does make me curious as to what has been causing these recent reports of massive variations between reboots. It makes it sound like your CPU is going bad or there is some compatibility issue with your motherboard. I'd expect all modern CPUs from the past few years to support Invariant TSC, and shouldn't really be having such issues.

In general I wouldn't recommend forcing HPET, unless you are doing so to resolve a particular problem like this. Some motherboards have a poor implementation of HPET which can causes issues with DPC latency. With a proper HPET implementation, DPC latency shouldn't change at all compared to the Windows default hybrid timer or possibly be slightly lower.