nVidia 471.96 and DXGI_DEVICE_HUNG resolved

[Ray Proudfoot 8,942]

1 hour ago, ha5mvo said:

No Ray, no registry hacks or tweaks.

I try to avoid those as this *may* open a can of worms that's difficult to track down later. Besides, if this was really the silver bullet for the hangs, it would have probably been a common knowledge by now I reckon.

I’ve flown 14 flights now with this entry in the Registry. No DXGI errors whatsoever with the PMDG737. All it does is inhibit the message that would otherwise appear and consequently terminate your flight.

Here is the setting. Backup your Registry first...

HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\GraphicsDrivers
- On the Edit menu, click New, and select the QWORD (64-bit) value from the drop-down menu,
- Type ‘TdrLevel’ as the Name and click Enter,
- Double-click TdrLevel and set the value as 0 (it is set to that by default, but double check) and click OK,
- Close the registry editor and restart your computer for the changes to take effect!

Why not try it for a few flights? If it works then spread the word. It’s worked for FS Labs Airbus pilots.

I’ve flown enough test flights now to consider this a permanent fix / workaround until such time as whoever is responsible fixes it.

If you don’t what’s the alternative? Just accept it? Not for me I’m afraid.

[Cruachan 705]

Another interesting and well-informed article that provides further clarity to the origins of this somewhat irksome issue:

https://www.pugetsystems.com/labs/hpc/Working-around-TDR-in-Windows-for-a-better-GPU-computing-experience-777/

Note the caveat: “Just remember that these won’t only affect the CUDA code you run: they would also allow a legitimate graphics software glitch to potentially cause the system to hang.”

If I understand the content correctly, it would appear that in fact the prime responsibilities for triggering the problem lies both with the software developer, in this case LM, and our predilection for loading our simulation installations with complex Addons that individually or in combination can present the GPU with an excessive work load if or when the flight situation presents itself. Periodic brief spikes and bottoming of GPU usage manifesting as pauses in the simulation or severe stuttering without triggering a TDR/driver reset would tend to confirm this view. In such instances the GPU workload is close to but insufficient to trigger the timeout before the particular flight situation passes and everything settles down once again.

Also, and this might be a more contentious suggestion, could it be that our use of the Affinity Mask for Prepar3D is either alleviating or exacerbating the problem? Maybe Steve would be kind enough to chip in here.

Mike

Edited September 28, 2021 by Cruachan

[SteveW 1,792]

Affinity masks won't interfere with it Mike.

[SteveW 1,792]

I have a PC here with stock P3D v4 and DXGI_DEVICE_HUNG comes up when the Vehicle window is displayed in the Scenario screen, any aircraft, and the sim closes. That won't be worked around by the registry hack. Changing the Nvidia driver version cures it. There's many ways to get the problem so what might work for one might not work for another.

[ha5mvo 251]

35 minutes ago, SteveW said:

I have a PC here with stock P3D v4 and DXGI_DEVICE_HUNG comes up when the Vehicle window is displayed in the Scenario screen, any aircraft, and the sim closes. That won't be worked around by the registry hack. Changing the Nvidia driver version cures it. There's many ways to get the problem so what might work for one might not work for another.

I concur.

For one, the FSL configurator sets its own affinity as far as I know. I very rarely have DXGI issues with the FSL. The Leonardo MD80 on the other hand suffers much more from that issue. I would assume that its PBR interior flight deck might have something to do with it.

Even so, reloading the exact same scenario that had crashes results in  perfectly running sim in the second time. Those errors are difficult to preproduce.

[SteveW 1,792]

On 9/27/2021 at 10:24 PM, Ray Proudfoot said:

Steve, I’ve had the case opened once in three years to check for dust etc. There was a minimal amount. I’ve never touched the graphics card.

Your suggestion about removing and reseating to clean contacts is valid and I’ve done it with my hi-fi loudspeaker cables. But not computer components.

Oxidisation seems very low risk especially when contacts appear to have gold-plated connections. Not 24-carat I imagine. 😁

Although Gold does not tarnish, there can still be problems associated with those contacts, one I mentioned already is fingerprints which can turn brown eventually and can even be felt in relief sometimes. Other factors to consider include smoke or moisture in the environment, or even fluff or dust getting caught in between contacts during assembly. Another aspect to be aware of is with heat and cold cycling, PC cards can edge their way out of the slot or misalign slightly and might break contact after a certain warm up period, some might not have been inserted properly in the first place. I've seen them all during my time in field maintenance. The comment that Kevin was referring to was in relation to a PC that was unexpectedly shutting down and this was cured by re-seating the cards and cables and been OK since. Your problem I am sure is most likely due to software, GPU driver plus P3D. Even so it is worth remembering that over time problems can emerge with power connections and card slots which are easy to check and costs nothing, although it is something to be very careful with since the cards and components in the PC can be damaged easily so better left alone unless experiencing problems.

[Ray Proudfoot 8,942]

Steve, I used to take my PC plus TV, yoke, throttles and pedals to a monthly flight sim meeting. Fortunately it never suffered any physical damage but transporting it in a car wasn’t ideal.

I’ve stopped doing it now because of its weight and packing up and unpacking twice in 6 hours was too much like hard work. No chance of hot/cold conditions here but dust etc. will always be a pain.

I left my PC. running for over 2 hours today in slew / pause mode and no DXGI problems which was nice.

[SteveW 1,792]

12 minutes ago, Ray Proudfoot said:

I used to take my PC plus TV, yoke, throttles and pedals to a monthly flight sim meeting...

A lot of effort and I suspect you very kindly helped a lot of flight simmers there. 👍

35 minutes ago, Ray Proudfoot said:

No chance of hot/cold conditions here...

I was referring to the PC going cold overnight when turned off then heating up in use during the day. Having said that, PC cards and cables are pretty good at staying put in their slots these days though and are better designed to avoid movement so that would be rare problem cause but some parts can be prone to movement.

 

18 minutes ago, Ray Proudfoot said:

I left my PC. running for over 2 hours today in slew / pause mode and no DXGI problems which was nice.

I'm glad you found a work-around to avoid the issues for the time being. 🙂

[Ray Proudfoot 8,942]

17 minutes ago, SteveW said:

A lot of effort and I suspect you very kindly helped a lot of flight simmers there. 👍

We are primarily a bunch of old gits sharing the enjoyment of simming. Being the only Concorde pilot I always had visitors.

18 minutes ago, SteveW said:

I was referring to the PC going cold overnight when turned off then heating up in use during the day.

Well after installing AI SUITE the fans start up very quickly. Water cooled too. But it does help keep the room warm when the outside temp drops.

[fluffyflops 1,856]

18 pages and people think it's a hardware issue to do with with fans, room temperature and stuff. 

You chaps do make me chuckle. 

Yes the fans and heat are so clever they can tell when some dodgy memory leaking bit of coding causes the sim to ctd.  Aka old versions of real light and true glass. 

I hate to break it to you. It's not your gfx card or a bit of dust Ray. 

 

[Ray Proudfoot 8,942]

1 hour ago, fluffyflops said:

I hate to break it to you. It's not your gfx card or a bit of dust Ray. 

I'm well aware of that. Lose the sarcasm fluffyflops.

[SteveW 1,792]

16 hours ago, fluffyflops said:

18 pages and people think it's a hardware issue to do with with fans, room temperature and stuff. 

No one said that.

[Bigmack 29]

On 9/25/2021 at 9:13 AM, Ray Proudfoot said:

You’re the first person to point the finger at AI Suite. Any evidence to substantiate this claim?

Can you describe “hosed”?

For some reason AI suite once installed would not allow me to go back to default settings. It kept my system in overclocked state. It interfered with my video driver. I wasn't able to removed it or update it. My system kept shutting down and restarting. I believe because the AI suite overclock settings was too aggressive. I tried uninstalling it from my system altogether but it still held on and would let me do anything. I called Asus support and they were not able to figure out why AI suite went rouge but my only option was to reinstall windows and take back control. No issues since then. I can control all my overclocking from BIOS including adjusting memory clock speeds

[Ray Proudfoot 8,942]

2 hours ago, Bigmack said:

For some reason AI suite once installed would not allow me to go back to default settings. It kept my system in overclocked state. It interfered with my video driver. I wasn't able to removed it or update it. My system kept shutting down and restarting

That has not been my experience. It should be possible to uninstall any software that you have installed. My system is absolutely fine with the BIOS settings set by the program.

[Pete Dowson 558]

2 hours ago, Bigmack said:

For some reason AI suite once installed would not allow me to go back to default settings. It kept my system in overclocked state. It interfered with my video driver.

That makes no sense. It doesn't do anything when you don't run it. If you run it to set overclocking it does that via the BIOS settings, and you can undo them all either in AI Suite or in the BIOS.

And it doesn't have anything to do with video -- that's MSI Afterburner.

BTW I saved several different sets of settings in AI Suite ("profiles"), with one which was actually underclocked from the default for the mobo -- just in case I ran into an problems and needed to test things in a presumably less stressed setup.

Pete