nVidia 471.96 and DXGI_DEVICE_HUNG resolved

[Luke 782]

3 hours ago, fluffyflops said:

my point is ray if they wont fix bgl data, why do you think they will fix dxgi errors.

You keep saying that you're certain that it's LM's fault. Unlike the data in their own app, they don't write the drivers, they don't manufacture the GPU hardware and they certainly don't create the power supplies to power them. If you're certain it's LM's fault, then you must know what the specific issue is. What is it?

There's a basic rule in software - userspace cannot break the kernel or the drivers. A misbehaving application cannot bring down the driver or the OS, and if it does that's a driver bug. Given how intermittent the situation is, I suspect it's some combination of GPU hardware, driver and power supply under instantaneous load. Triggered by P3D? Sure. But not caused.

We tend to focus on the last thing that happened as the cause, when it is merely a trigger. It's akin to having a bomb placed in a car, triggered on turning the ignition and then blaming the unlucky driver (or the manufacturer or the ignition lock) for the resulting explosion. To say that 50% of the blame should be leveled at Umberto for an obvious P3D bug is unfair and merely encourages people to not work around other people's issues (since that implies that their previous code was at fault) or not write client software in the first place.

Cheers!

[Cruachan 705]

On 10/31/2021 at 4:22 PM, Luke said:

No one knows the trigger or where the problem resides, but you're certain why it's not getting fixed?

Hi Luke,

You sure about that? I think it’s far more likely that the cause IS known, whereas the fix is either beyond their current expertise or would involve a major and costly rewrite. Either way such an admission would undermine faith in (most probably) DirectX 12.

The level and sophistication of software debugging tools these days is such that failure to identify the cause/causes of these errors is, quite frankly, unbelievable. ‘They’ know alright, they are just not saying as it does not fit their current agenda, whatever that may be.

Regards,

Mike

[Luke 782]

1 hour ago, Cruachan said:

You sure about that? I think it’s far more likely that the cause IS known, whereas the fix is either beyond their current expertise or would involve a major and costly rewrite. Either way such an admission would undermine faith in (most probably) DirectX 12. The level and sophistication of software debugging tools these days is such that failure to identify the cause/causes of these errors is, quite frankly, unbelievable. ‘They’ know alright, they are just not saying as it does not fit their current agenda, whatever that may be.

"I don't know what the cause is, but I am certain they know it and there's a conspiracy behind why it's not fixed!"

I'm aware of the sophistication of software debugging tools. I've also seen first hand how hard it is to debug issues caused by an interaction of software from three different vendors coupled with hardware from another three, and that's just if you've spec'd it all yourself and it's the same. Heaven help you when it's all different like here.

I challenge everyone to put up or shut up - whose code is faulty, and what is the trigger? If they can't answer that basic question, then there's no point assigning blame or motive.

Edited November 2, 2021 by Luke

[Ray Proudfoot 8,903]

Things may become clearer with the next release of MSFS when there will be support for DX12. If lots of people start screaming about DXGI issues with that sim then it will be interesting how Microsoft / Asobo respond.

[Cruachan 705]

2 hours ago, Ray Proudfoot said:

Things may become clearer with the next release of MSFS when there will be support for DX12. If lots of people start screaming about DXGI issues with that sim then it will be interesting how Microsoft / Asobo respond.

What’s the betting that there are no DXGI related issues. Microsoft / Asobo are aware (if not then what planet have they been on?) and at release will have engineered the heck out of this issue.

Mike

[Ray Proudfoot 8,903]

6 minutes ago, Cruachan said:

What’s the betting that there are no DXGI related issues. Microsoft / Asobo are aware (if not then what planet have they been on?) and at release will have engineered the heck out of this issue.

Mike

If that is indeed the situation then it puts the ball firmly back in LM’s court.

[Cruachan 705]

6 minutes ago, Ray Proudfoot said:

If that is indeed the situation then it puts the ball firmly back in LM’s court.

Did I hear anyone say Prepar3D v5.3 or maybe even v6.1? Lol! Why do we do this to ourselves?!

I consider myself as being fortunate as I’ve been taking a break from flight simulation recently. Instead, my time has been spent more profitably by brushing up my guitar licks and all those neglected rusty tunes. No sleepless nights, no anxiety, just oceans of calm and oodles of pleasure and satisfaction 🤓

Mike

[planechaser 46]

I do not suffer from DXGI errors ( so far ) knock on wood.

I think this boils down to:

1.Faulty drivers ( remember newer is not always better ) . I use 471.91 ?  seems to be fine.
2.DX12 - too much processing being done by the GPU and combined with an old 2005ish engine causing issues.
3. Incompatible add-ons.( less likely)

5.2 most likely offloaded quite a bit of processing back too the CPU from the GPU, this greatly reducing these DXGI hung errors. Unfortunately , this has resulted in sometimes a significant FPS issue -especially with some add-on sceneries. in 5.1 , FPS were amazing and you could definitely see the difference. With 5.2 ( CPU hit harder than before0 , while still better than 4.5 , the gap between the 2 is narrowed. When you are running the latest and greatest CPU at 5.1 GHZ all cores plus a still-good 1080TI and barely hitting 30 FPS , that is a problem. And that is not with maxxed out AG settings.

with MSFS maturing and a lot of the add-on devs ( Orbx, REX ) moving on to bigger and better things. Combine that with more regular sim updates and an up to date SDK, P3d's days are numbered. This is 2006 FSX ( yes 2006 ) dressed up. Take a look at weather in P3d with white mountain syndrome and visibility issues - then go to FSS2020, you see the 15 year difference in tech. When video card stock comes back plus study-level aircraft come available , MSFS will pull away further from the pack. 

Scenery in P3d - 2009
Scenery in MSFS - 2020

You smartphone in 2021 is more powerful than a fully loaded PC in 2005. You can probably play FSX on your Iphone with good performance.

Edited November 2, 2021 by planechaser
spelling

[Ray Proudfoot 8,903]

17 minutes ago, planechaser said:

with MSFS maturing and a lot of the add-on devs ( Orbx, REX ) moving on to bigger and better things. Combine that with more regular sim updates and an up to date SDK, P3d's days are numbered. This is 2006 FSX ( yes 2006 ) dressed up.

Can I remind you this topic is about DXGI problems in P3Dv5 and nothing else. Switching the discussion to P3D versus MSFS is unwelcome. Start another topic if you wish to discuss that.

[Les Parson 99]

Thank you gentlemen for this fantastic thread. I have the 3090 card w/ 24Ghz and experienced my first DXGI failure just the other day. I want to perform more flights (A2A V35 only) in P3Dv5.2 before posting all the details. 

I just wanted to post somewhat of a "pre-post" and urge the continuance of this thread as I would be in a bad position without all the excellent advice. 

Thanks again / Les Parson

 

[Bigmack 22]

On 11/2/2021 at 7:48 AM, Ray Proudfoot said:

Things may become clearer with the next release of MSFS when there will be support for DX12. If lots of people start screaming about DXGI issues with that sim then it will be interesting how Microsoft / Asobo respond.

Where are the shader folders located that get deleted when installing a newer Nvidia driver?

[Ray Proudfoot 8,903]

4 minutes ago, Bigmack said:

Where are the shader folders located that get deleted when installing a newer Nvidia driver?

They don’t get deleted automatically, you have to do it yourself. My PC is off but from memory it’s...

Local Disk (C:) ► Users ► You ► AppData ► Local ► Lockheed Martin ► Prepar3D ►  Shaders

[Bigmack 22]

5 hours ago, Ray Proudfoot said:

They don’t get deleted automatically, you have to do it yourself. My PC is off but from memory it’s...

Local Disk (C:) ► Users ► You ► AppData ► Local ► Lockheed Martin ► Prepar3D ►  Shaders

Thanks you very much.

[pgde 571]

You may also want to delete your DX12 shader cache. See 

for a tutorial. 

[Ray Proudfoot 8,903]

I've hardly bothered with the DX shader cache since adding those entries in the Registry. Not one DXGI problem.