F@H 7.6.13 + RTX 2080 Ti 456.55 Windows Crash

[bruce]

It comes to mind that there have been weird behaviors observed on Windows with FAH and GOG, something linked to GOG using .dll similar to what FAH is using, or in the same directories, or something.

viewtopic.php?f=108&t=34665

Yes, GOG is replacing certain .dlls that FAH needs. I suggest you uninstall both sets of Apps then install FAH without installing GOG. My guess is that FAH will run. After verifying that, reinstall GOG and check and report what .dll have been added to the FAH loads.

I'm certainly not knowledgable in GOG code paths or why they find it necessary to disrupt FAH's code, I just know we were able to fix a similar problem some time back by preventing GOG from messing with FAH

[VxJasonxV]

The 2080 Ti has 3 DP ports (2 useable with my case), 1 HDMI, and a USB-C port. https://www.nvidia.com/content/geforce- ... .pdf#page7 The third offset DP port is completely obscured by my case.

I have my monitor in the port closest to the USB-C port, and my Valve Index in the opposite end. But the Index remains unpowered when not in use.

Nothing is connected via HDMI nor USB-C. (I didn’t even know I had that USB-C port until just tonight researching all this…)

I don’t do a lot of tweaking, the most I have is GeForce Experience, and letting it “automatically optimize games”, but note that this is supposed to be limited to editing games’ configuration files so they output in optimal quality.

I had used Hue Sync for the two Hue Play lights on the back of the monitor, but games started disliking that, Doom Eternal in particular would go down to sub-10 FPS with it running, so Hue Sync has been installed but not run in quite some time now.

I deleted Razer Synapse a while ago, not that I used their whatever boost software (Cortex, I think?) they bundle.

I listed everything that shows up in the taskbar when I lock my computer. The only things to add would be things running only when I’m actively using them, various other game launchers, Discord, etc., but no tweaking software aside from Hue Sync, which I’m reasonably sure is not permanently resident a process.

I haven’t changed anything with the nVIDIA Control Panel besides enabling G-Sync only in full screen mode. I don’t know if I’ve ever looked at the other panes of that app.

I don’t know if I have another DP cable, I could flip the ports I guess, but I have to state that I can’t fathom how this is an issue when the system is locked and the display is off. And doubly, since F@H isn’t rendering anything for display at the time. FahControl isn’t open, no browser is open, I don’t leave the protein viewer up or anything like that. I lock the computer, shut off the display, and when I come back to my computer, I’ve since been logged out unexpectedly, with those video output crashes as listed earlier in this thread.

This has happened mid-use if I turn F@H on while barely using my computer, it happens 100% of the time overnight since F@H tends to stay on “on idle”, and kicks in to high gear when my computer is locked.

I mentioned Star Wars: Squadrons earlier in the thread, those were Star Wars: Squadrons crash bugs and have not recurred since patches.
It has never happened during any other VR game.
It has never happened during Rocket League.
It happens 100% of the time otherwise idle overnight which is the primary time F@H runs. Or any other time I let F@H run. I even turned it down to “Low” Folding Speed and it still happens.

Typing this up, it just occurred to me that I should try turning off the GPU workers and seeing if it happens, and if it doesn’t, see if I can find a reduced number of GPU workers where it doesn’t happen.

[ajm]

I would try to disconnect the Valve Index. Maybe it can be activated in some conditions, or maybe it is somehow faulty. I have seen crashes in similar situations, when a device is considered as a "display" by Windows and is activated or disactivated in special circumstances.

[VxJasonxV]

On the one hand, it’s unpowered at the mains. On the other hand, it has a USB plug too. ┐('～`；)┌

It won’t be detrimental to unplug it, so I have. Unplugged, turned computer on, logged in, locked computer. We’ll see what happens in the morning.

[VxJasonxV]

No changes in behavior. Crashes still coming fast after I close up apps and lock my computer. I left my computer around 18:10, came back just 5 minutes ago, and it was logged out though not slept yet. My computer is set to sleep after 1 hour, and it would only do that if F@H isn't running, which the crash, logging me out of my session, causes. The computer will not sleep as long as F@H is running, which stops and logs out my entire session when whatever crash happens, happens.

I've deleted my GPU slot and am going to see if it survives over night. I expect it will.

I need to re-iterate the peculiarity of the crash:

Windows is NOT Blue Screening and restarting. The computer remains on, Windows remains working, but I'm logged out of my session, all user processes are shut down abnormally.

[ajm]

Maybe you'll find a solution there:
https://answers.microsoft.com/en-us/win ... 13bfb23f88
https://answers.microsoft.com/en-us/win ... 52d?auth=1
https://answers.microsoft.com/en-us/win ... 89bcc57ff4

This kind of "log out after a while when the PC is locked" can also be a feature, ie enterprise admins would want that for some of their users. Maybe you or someone else installed that once in the past?

[PantherX]

Here's another test case:

Leave your system logged-in, exit all applications including the ones running in taskbar with the exception of F@H. Turn off your monitor(s) and ensure that you don't have auto-lock on your system. See what happens over lunch time and if it works, repeat the same at night-time.

[VxJasonxV]

It isn’t an auto-logout policy because it is my personal PC, I’ve never set that setting, I live alone, no one else has ever used this PC, it’s not joined to a domain/directory, etc. Again, this crash behavior happens no matter whether I let F@H run in the foreground, or run when the session is locked. I keep bringing up the background crash because I keep F@H set to “On idle”, but the same thing happens if I turn F@H on while I’m using the computer. At some random time, under no specific load (besides F@H) the screen just goes black. When it comes back, I’m at the login screen.

It is not a blue screen because 1. I have automatically restart off so I would see the blue screen, and 2. I never see the SBIOS/UEFI/whatever. The fans also never spin down, Windows doesn’t stop running, this is a type of crash I’ve never seen before in Windows. I’ve had X server crashes in Linux and thats the closest analogy I can draw. That said, the video card doesn’t stop outputting because my monitor never goes to sleep or indicates any sort of a change in state. Of course, the black screen is only there for a couple of seconds, but there are no tell tail signs of video output changes.

Windows just closes my session out suddenly. No warning message, no indicator, no nothing. Just •

I’ve never set any sort of usage limits on my account, and if I had surely some amount of gaming would have tripped it as well. Lots of gaming and many other random whatevers should have caused it. But no, nothing expect for F@H, and from what I can tell, within an hour of being allowed to run, it crashes.

7 hours after disabling the GPU slot my computer is still on as I expect. I’ll turn them back on tomorrow.

[PantherX]

Correct me if I am wrong but it seems that this is the current status of your system:
System not folding -> Functional
CPU folding only -> Functional
CPU & GPU folding -> Problematic
GPU only folding -> Problematic

If the above is true, then it would mean we need to investigate the issue on your GPU only when folding. I am aware that you mentioned gaming without issues (once the patches were applied) but gaming is less intensive than folding. On Nvidia GPUs, it uses CUDA which pushes the GPU even harder. Thus, try:
1) Running your GPU on Nvidia stock frequencies
2) Check power delivery to your GPU; are there any fluctuations? anything out of normal?

[VxJasonxV]

I have never done GPU only folding, but the other 3 are true.

I have never attempted to overclock anything on my system. It should be as-configured from the BTO vendor I bought it from, which is not known to me to be a “performance pusher”. That said, I’ve also never reviewed the settings. GPU Z was recommended earlier in the thread, which I’ve opened a few times but have never really known what to look for. I also haven’t looked at it mid-fold.

Those are on the list for me to do after I re-add the GPU slot tomorrow. I’ll open GPU Z, record lows and highs, and re-add the GPU slot and keep taking notes.

Is there a reference point for what sort of frequencies and voltages I should be seeing? I have no other frame of reference besides without Folding and with Folding. But naturally with Folding is going to be higher due to power demands.

[ajm]

The 2080ti is always more or less "overclocked" out of the box. If you want to look into this, I'd recommend using MSI Afterburner, rather: https://www.msi.com/Landing/afterburner
It is also free, it shows better graphs (detachable section underneath) and it allows to easily correct an excessive overclock.

[PantherX]

For the reference speeds, look at this page which has them listed: https://www.techpowerup.com/gpu-specs/g ... 0-ti.c3305

In GPU-Z, you will see a section on the first tab which shows current frequencies (GPU Clock, etc) and stock frequencies (Default clock, etc). MSI Afterburner does allow modification of those clock speeds.

[ajm]

Afterburner also shows the base and boost speed of the card. It is in the blue zone of the left circle: white triangle = base speed, blue triangle = boost speed.
Here, on the image above, mine has a boost speed of 1650 MHz and an effective frequency at 100% Power Limit (= out of the box) of 1950 MHz. If I reduce the Power Limit to about 65% I can "respect" that boost speed.

[VxJasonxV]

The inevitable update: The same behavior occurred last night after re-enabling the GPU slot. Came back during the day and my computer was asleep.

In the process of collecting data, testing something with GPU-Z's log-to-file feature I stumbled on something strange. I locked my screen and walked away for maybe 15 minutes, came back and my session had already been logged out… ?!

If I lock my screen for less than a minute it doesn't happen. Tomorrow I might sit down, lock my screen, and just wait and see if I can see the same screen blanking thing that I talked about happening mid-use. The problem might be if there is something other than the screensaver (which I have disabled) to do with a normal screen blanking that occurs on a locked screen after some amount of time.

I do have my device set to screen blank after 15 minutes, sleep after 1 hour, but I doubt that's related considering that this behavior can also happen mid-use, logged-in. Again, the sudden, unexpected sign-out of my session from the lock screen or logged-in is the strange behavior I'm attempting to diagnose.

And, and, again-again, the computer never slept with CPU-only folding!

I never ever did wind up posting F@H, sorry. See https://gist.github.com/VxJasonxV/53808 ... 2a6ba07585 for Oct 27 @ 00:00Z forward to literally as I write this, including the sudden logout that happened to me just now too.

[ajm]

It may comes from an issue with your 2080ti. I've read elsewhere about such a behavior (random log-out) caused by display crashes (MacOS).
As said above, all 2080ti are overclocked, that is, they will go beyond their standard boost clock (to a point, determined in their specific bios) if they have enough power and cooling. And it can work very well with games, but not at all with FAH, where the slightest error can be fatal. Then you have the silicon lottery: some cards of the same model will be able to handle higher (boost) frequencies than others.
I would try to lower the Power Limit using Afterburner. First a lot, so that your card clocks at her standard boost speed. If the problem disappears, you then can test which frequency you can safely use, or just leave it at standard boost speed for FAH, and get back to 100% for gaming.
Another test would be to disable cuda, which is more demanding.