Question / Help Stream Goes Down but OBS Looks Live and There's No Crash Report

[Cute_Spide]

So this has been happening for almost every stream for 2 weeks now. Everything seems fine until suddenly it isn't. I first noticed this (while playing Dark Souls) in interaction with my bot. My bot (streamlabs chatbot) would get a white boarder around it like something had went wrong. Then I was told by somebody in the room with me that either my stream froze or went down. If it freezes it's likely I have to restart that and the bot anyway because the latency after that is horrible. Because I thought maybe my bot and OBS were having strange interactions, I uninstalled both and reinstalled (restoring backups). I thought maybe it lessened because it doesn't crash when I'm playing Pokemon, but this isn't the case. For whatever reason, it's fine when playing that game.

Last night, however, it went down while trying to play Paper Mario. Specifically, there were scenes in Paper Mario where a 3D object would shift and a huge amount of lag would happen in-game. This made it go down every time and so I stopped playing.

However, the thing is that, if nobody told me I was down, I would have NO idea. OBS doesn't tell me about any frame drops and it doesn't attempt to reconnect (as far as I'm aware). I was told to check out OBS crash logs and there's none since last year.

If anybody could help me, I would be eternally grateful! This is driving me crazy because I just put so much work into bumping up stream quality and streaming is something I really love doing. I have no idea what the issue is >.< Let me know if you need more information! Attached are 3 log files from my most recent stream (last night). I can link some older ones that come from these times when it crashes as well, if need be.

What I have open during streams is usually OBS, Streamlabs Chat Bot, MS Paint (during Mario and Pokemon), sometimes Discord, and a 3rd party controller thing that lets me play with a PS3 controller (has never given me issues)

Things I did at the beginning of the 2 week time frame is kind of a lot because I recently brought the quality of my stream up a bunch and also had to download some things for my job. I had to update Skype from classic Skype to the new Skype via a forced update. I also had to download a program called ManyCam because of the way my employer wants our Skype business to be setup. I followed this guide and got a program called Python and OBS Websockets to enable OBS sources being activated using commands (I feel like this isn't the issue but I thought I would list it here regardless). I can't seem to think of anything else at the moment but I hope this is at least helpful.

 

[koala]

I see error messages like this in 2 of your logs that make your stream stop:
18:47:06.670: error: Failed locking bitstream buffer: invalid param (8) 18:47:06.706: [NVENC encoder: 'simple_h264_stream'] nvenc_encode: Error encoding: Invalid argument 18:47:06.706: Video stopped, number of skipped frames due to encoding lag: 5/88156 (0.0%) 18:47:07.134: Error encoding with encoder 'simple_h264_stream'
It seems some kind of internal error within the nvenc encoder. I've seen this in other logs before, with the same issue (stream stops without noticing), but I never read anything about a fix. The error is very rare.
I guess (but do not know) this might be some kind of hardware error within the nvenc circuit, resulted perhaps by overheating, low power voltage or by overloading the GPU in a way not foreseen by the chip manufacturer.
Check your hardware, revert any overclocking, check temperatures, check power supply, verify correct seating of GPU card in the mainboard slot. If this doesn't help, I don't know more.

You might try a newer or older GPU driver, but I doubt that will change anything.

The GPU is a bit older and you are using a huge amount of sources - may be it's something totally different, may be the GPU or nvenc is running out of GPU memory and there isn't a proper error message about this. In this case, reducing the number of sources and/or filters might help.

 

[Cute_Spide]

I see error messages like this in 2 of your logs that make your stream stop:
18:47:06.670: error: Failed locking bitstream buffer: invalid param (8) 18:47:06.706: [NVENC encoder: 'simple_h264_stream'] nvenc_encode: Error encoding: Invalid argument 18:47:06.706: Video stopped, number of skipped frames due to encoding lag: 5/88156 (0.0%) 18:47:07.134: Error encoding with encoder 'simple_h264_stream'
It seems some kind of internal error within the nvenc encoder. I've seen this in other logs before, with the same issue (stream stops without noticing), but I never read anything about a fix. The error is very rare.
I guess (but do not know) this might be some kind of hardware error within the nvenc circuit, resulted perhaps by overheating, low power voltage or by overloading the GPU in a way not foreseen by the chip manufacturer.
Check your hardware, revert any overclocking, check temperatures, check power supply, verify correct seating of GPU card in the mainboard slot. If this doesn't help, I don't know more.

You might try a newer or older GPU driver, but I doubt that will change anything.

The GPU is a bit older and you are using a huge amount of sources - may be it's something totally different, may be the GPU or nvenc is running out of GPU memory and there isn't a proper error message about this. In this case, reducing the number of sources and/or filters might help.

So Based on what you said, I tried a few things. I did a test stream and used Paper Mario as the game to coax out the issue. Removing scenes and whatnot didn't seem to do much and temperatures on all hardware seem fine. The stream went down each time I reached the animation I mentioned before, so it's a good thing to reliably test this issue on.

However, I then switched my encoding to software and while the cutscene was still broken (as expected, it's an emulation issue), the stream didn't crash. In fact, watching a live rundown of my stream's performance showed that under the GPU's encoder we were only at around half the bit rate I am set to stream at (it was running at around 2K and I had it set at 4K). On the software encoding, it was a pretty stable 4K bit rate.

So temps are fine, I don't overclock, and I am willing to bet my voltage is fine. How does one go about updating or reverting GPU drivers? I have an issue sometimes while trying to update Nvidia driver. 2 or 3 error messages pop up and I'm wondering if it just never installed correctly. I also know that I attempted to fix this not long ago and I'm wondering if this coincides with the time my issues started happening. That is, I've had it give me error messages for like a year but I did a fix recently and everything seemed to work when updating.

Does this point to GPU issues then? Am I going to have to replace it? Also does closing scene sources out when they aren't being used/visible help? Also, weird question, but is it possible in any way that this is a RAM issue?

Attached are my test stream logs, the last one not resulting in a crash.

(thank you so much for your help thus far!)

 

[Cute_Spide]

As an update to this situation, I thought I might list some of the things I've done since to try and fix this problem, but with no such luck so far.

-I did a clean install of new and much older drivers for my GPU
-I did the DDU method of scrubbing everything from my GPU to REALLY install clean drivers
-I changed a setting somebody thought my have been causing it with my webcam
-I tried different encoding settings in the advanced tab
-I tried my GPU in another PC and the way in which I get a consistent crash on my rig didn't happen (that is, Paper Mario's emulation was smooth and didn't have the issue in the emulation that I expected to happen) and so, I'm unsure if this is conclusive of anything
-I reseated my GPU
-I cleaned my PC
-I changed my scene collection to that of a clean one with just cam and window capture to make sure my many sources weren't the cause

I'm unsure what to do next at this point. Any helps would be greatly appreciated. Attached are some more recent logs that are less cluttered since I was in the empty scene collection. Again, there's no crash log as OBS doesn't create one or even indicate that it has had an issue at all.

EDIT (adding to the list) 1/28/19
-I "upgraded" to WIndows 10
-I tried OBS Classic ****NO CRASH**** (idk what this means tho?)
-I uninstalled ManyCam, a program I got for work around the time the issues started
-I uninstalled new Skype that I also got for work around the time the issues started
-I did a clean install of OBS and didn't import anything
-I uninstalled Python
-I uninstalled OBS' websocket plugin
-I cried really hard and it still didn't fix itself

 

[Cute_Spide]

-I tried a Shadowplay stream and no crash
-I tried this beta NVENC encoder thing and still a crash

This log should be the one while using this beta encoder.

 

[Narcogen]

This pops up as being more particular than the earlier NVENC related error:

12:55:24.736: [jim-nvenc] init_encoder: nvEncGetEncodePresetConfig failed: 15 (NV_ENC_ERR_INVALID_VERSION)

This might help Jim out.

 

[Narcogen]

How does one go about updating or reverting GPU drivers? I have an issue sometimes while trying to update Nvidia driver. 2 or 3 error messages pop up and I'm wondering if it just never installed correctly. I also know that I attempted to fix this not long ago and I'm wondering if this coincides with the time my issues started happening. That is, I've had it give me error messages for like a year but I did a fix recently and everything seemed to work when updating.

This says to me: clean driver install.

https://obsproject.com/forum/resources/performing-a-clean-gpu-driver-installation.65/

 

[Cute_Spide]

This says to me: clean driver install.

https://obsproject.com/forum/resources/performing-a-clean-gpu-driver-installation.65/

I've since talked in the OBS Discord and did the whole DDU thing to no avail. Successful changed drivers up and down in number and nothing worked, so I'm willing to rule out a simple driver issue.

 

[Narcogen]

I'm afraid that's all I've got. :(

 

[Cute_Spide]

I'm afraid that's all I've got. :(

That's fine. After over 2 weeks of trying to figure this out, I've basically done it all. Thank you for your help, regardless <3

 

[Cute_Spide]

This pops up as being more particular than the earlier NVENC related error:

12:55:24.736: [jim-nvenc] init_encoder: nvEncGetEncodePresetConfig failed: 15 (NV_ENC_ERR_INVALID_VERSION)

This might help Jim out.

What does this mean, exactly? I'm unsure what I'm looking at with most of this. I've been told that the "crash" is very vague and, therefore, very difficult to track down.

 

[Narcogen]

I don't know, I'm not Jim.