Question / Help Same old settings, new "encoder overloaded" problem

[Rorax]

Computer Specs:
OBS version: 23.1.0
OS: Windows 10 Pro, 64bit
CPU: Intel Core i5-6600K 3.5GHZ 4 cores
GPU: GeForce GTX 980 Ti
Driver Version: 430.86
Memory: 16gig
Audio Card: ASUS Xonar Essence Stx
Driver Version: 8.1.8.1823

As of a couple of months ago my OBS setup has completely bombed. When the problem first started I was running the same settings I always ran that used to work fine. I have no idea where this issue came from whether it was an OBS update, a GPU driver update or a Windows 10 update. I wish I had payed more attention.

Anyway, now when ever I record I will intermittently have massive "encoder overloaded" spikes which last anywhere from 1-10 seconds, I have looked at task manager and OBS stats pages while these occur and my GPUs encoding graph never peaks over 13% my CPU barely over 20% and its writing to a pretty dang fast SSD which reports no faults when tested. And these are the stats when the error "encoder overloaded" appears in OBS. As far as I as a lay person is concerned it has more than enough resources that its just not using? And just to paint the picture, all this can occur recording nothing but an unmoving desktop.

I have tried both the old and new NVENC Encoders, I have tried testing with psycho tuning on and off on the new encoder, I have tried using my original CBR rate control set to 2500. I have tried using CQP with the CQ setting set to 15. preset, Quality, profile, High.
Because of the latest windows 10 updates futzing with sound a lot I made sure bit rates all matched across the board incase it was a sound issue, all have been set to 44.1
I have tried changing back to simple and using the wizard assigned settings of high quality medium file size.
I have tried disabling audio devices, I have tried stripping back to a bare bone scene.
I even tried restricting OBSs core affinity.

Nothing so far I have tried has produced satisfactory results. Changing to CQP seemed to briefly work in one test and then every other test bombed all the same.

I'm at my wits end. Here is a log of a test record its one of the most well behaved logs I have but it still has the problem in it.

 

[Narcogen]

14:52:59.195: Windows Version: 10.0 Build 17134 (revision: 829; 64-bit)

Windows is out of date. But this would most likely cause rendering lag, not encoding lag. You still might want to update as this version is over a year old. If you're saying Windows just updated, it may be that it tried and failed, which could be part of the problem.

AFAIK, GPU utilization graphs don't include encoder activity since that's a separate chip. You can't just say, "GPU utilization is lower than X so it shouldn't be overloaded, it must be wrong." (Might include I/O, not sure).

But generally speaking, for NVENC encoder overload means you can't do that frame size, at that frame rate, with those settings, and you need to change something.