Aweless

@glider1 Have you tried locking your gpu clocks via CMD? here a few commands that might be useful: 

Run cmd administrator navigate via CMD to cd \Program Files\NVIDIA Corporation\NVSMI\ 

nvidia-smi -lgc 4104,1620         This command will set gpu clocks 4104 Vram 1620 core clock. 

You can also check the gpu power limits: nvidia-smi -i 0 --format=csv --query-gpu=power.limit

To set power limit: nvidia-smi -i 0 -pl 160  You can define any number instead 160.

Here you can log power draw: nvidia-smi -i 0 --loop-ms=1000 --format=csv,noheader --query-gpu=power.draw

For more detailed information you can also give a look nvidia-smi pdf file that placed C:\Program Files\NVIDIA Corporation\NVSMI folder.

Regards

1 hour ago, Cruachan said:

I hope I didn’t imply that that was likely to be a sensible or, indeed, rational option. Nor do I believe that the subject of this thread was influenced by the choice of operating system. From where I stand I am content with what I have while, at the same time, having the luxury of exploring what Windows 10 may have to offer me in, say, a year or so. By that time, hopefully these issues will mostly be sorted out and I can think about transfering my allegiance. Meantime I’m in no hurry as W10 offers me nothing I don’t have already in abundance.

Mike

Nope, And i really didn't say so neither to convince you nor anybody else to switch win10 my intention wasn't that and never will be and why should it be?  We are performing brain storm here and that's just amazing.  the subject of this thread wasn't influenced by the choice of operating system i just shared what i experienced by the intention of maybe that would be a beacon for someone else who suffering from what i suffer.

Regards.

31 minutes ago, pgde said:

Just following up on my previous posts about my RMA to Gigabyte for this problem. They just notified me that the card is being replaced under warranty. And, just for the record, I run Win7 so this problem is not solely related to the OS version you use.

Happy Holidays!

P.

In new motherboards it might be related to the OS with modern hardware gpu cpu etc. And it is another story being able to install win 7 in new motherboards whatever you do despite new mobo allows win 7 installation you are facing some troubles during install and after install. 

38 minutes ago, Rob Ainscough said:

It's actually the 1809 Nov Update, Microsoft pushed this update out in Oct, it failed for many, withdrew it, and then pushed in out again in Dec.  1809 is NOT without problems still ... for example I operate several networked PC's and so far I've had to do a "Network Reset" in order for Explorer to be able to see my networked PC ... I have to do that after ANY Windows Updates related to 1809.  It's apparently a known issue that Microsoft "might" address in the future.

Why this is a problem for me is that I run three computers, one main FS computer, one network computer for weather, and one for flight planning/tracking ... When installing HiFi AS4 is wants a path to my main FS computer and "mapped" drives do NOT show in the dialog, only Network paths will work and my main FS computer wasn't showing up (yes network discover was on and running).

Soooo ... after many hours of diagnostics, checking services, etc. etc. the ONLY solution for me was to run "Network Reset" on all my PCs running 1809 after ANY Windows Update that is related to any core elements of 1809.

Cheers, Rob.

Yes i know the story Rob. it's been reported by some users that the October update had deleted user files at first release and then Microsoft cancelled the update.

It looks like one of my nightmare bug has gone after updating to 1809. I was getting some stuttering on the desktop on idle at times before the update and this is what made me think my gpu could be faulty.

i Assume for your issue you may have to wait network driver updates for the new 1809 build for all your pc's that running on the network.

I would suggest you to look for new driver updates via an app as driver easy for your network adapters if you haven't done something similar to this so. If your secondary network pc's are older then your issue may have something to do with incompatible network hardware too.

Have you ever tried running only two computers on the network and run your both weather app and flight planning app in one computer?

There might be good news in windows10 1809 October update for those who facing the dxgi error . i am up to date today and there are noticeable improvements. You can also watch the the power usage per apps in task manager. I haven't performed a flight yet but i will do by rolling back any changes i made previously.

3 hours ago, joepoway said:

Aweless

In an attempt to summarize your path to a solution tell me if this is correct, please:

Limited ( not locked to one frequency ) your GPU clock frequency using the method described above. This method helped but did not fix the problem? I saw it was not the solution for CJ

Then you went into the bios and enabled PCIE power saving features coupled with Windows 10 balanced power features turned on

You also set all NVI settings to default which would clearly not be running Preferred Power Maximum that most all of us use with NVI or NCP

So it appears ironically you are holding the frequency variations of you GPU  at higher settings (more power consumed but fewer spikes). However through enablement of the power options in the bios and Windows you are allowing the CPU to throttle up and down In a power saving mode. This is something I have always disabled in my overclocks and just run at the continuous overclock speed on all cores all the time to limit flucuations. I’m a bit surprised the PCIE bios change was what allowed your CPU to downclock usually there are other bios features to accomplish this. My Gigabyte board is 6 years old and I don’t have the PCIE feature you changed but have always disabled the CPU throttling features.

So to summarize you totally enabled CPU and PCIE Windows/Intel throttling features, tried your best to lock GPU frequencies (ironically the opposite approach to CPU), and disabled all NVI max performance settings. Is this correct?

Finally I assume the GPU frequency “locking” and no max performance NVI setting didn’t solve your issue it took the combination of limiting GPU frequency spikes with allowing Windows to use the Intel CPU power management functions, correct?

Thanks in advance

Joe

Joe there is a misunderstanding to be cleared.

To summarize my solution

1: As long as i kept my gpu clocks locked i had no dxgi error at all.

2: All i wanted to find what sort of solution i could figure rather than locking the gpu clocks not to get the DXGI so i managed this by leaving power management up to windows by doing some extra changings in bios so this is why i dig in here.

3: I should mention this additionally: in your terms before i enable the PCIE power saving feature i seen that the gpu clocks were going down to 139 MHZ on default clocks while on idle even though nvidia profile was set to prefer high performance. After enabling the PCI Express Native power management in combination with Native Active power management which lets the windows interfere to regulate the power drawing in GPU and then i seen that the gpu clocks kept remaining around 900 MHZ on default clocks while on idle even though nvidia profile set to default in global. And yes this sound way more ironic and yes i think there are some conflicts in between bios windows ncp etc who know what else trying to take the control… I guess being able to lock Gpu clocks that easy is something special to pascal cards.

4: To clear one thing that misunderstood: Enabling PCIE ASPM has nothing to do with CPU interaction. Thats something i discovered later on after i figured something must be done to reduce power drawing from the psu. To let the windows manage cpu clocks there are a few modes in my bios that i had to dig in these modes: EPU, PERFORMANCE, Max Power saving and AUTO Cpu C states and intel SpeedStep i don't exactly remember what i changed in my bios i may also have changed c states to auto or SpeedStep to Auto to let the windows take the control of cpu speeds. Here are a few screen shots to my current bios configuration to make the windows to control the cpu speeds it must look like this: 

 But i think it won't make much difference no matter i keep the cpu working @4ghz. Balanced power plan has just became my preference to avoid in case if the psu feeds insufficient power to the system or a faulty gpu wants to draw much more power from the psu at times. these are my assumptions so i go balanced that the system could stabilize it self by reducing the frequencies as needed, just a precaution. I had to perform at least one action of two options not to get dxgi, 1: Lock Gpu clocks. 2: Enable Pcie power management feature in bios and leave the gpu clocks alone. But now i use two options together both locked gpu and enabled pcie aspm also balanced power plan enabled in windows and works good and i haven't seen any fps drop in simulator comparing the previous setups. 

I hope i could clear the questions in your head.

All the best.

I got your point, to me The relationship between gpu clock speed is to the idea that i got from Joe's earlier post is when the gpu clocks going up and down it accumulates some stress on the gpu by requesting a different power level each time. There are two things to eliminate in this case, one a faulty psu that can't feed the requested power frequencies to the gpu properly, resulting in gpu to reset and disconnect from the simulator  i wish the p3d could by pass this kind of resetting and keep running as well just like in Linux systems.Well maybe in future. second a faulty gpu that requests excessively unachievable frequencies at its peak so this also leads the gpu to reset and disconnects from the sim. In my case maybe i got a faulty gpu but somehow locking the gpu clocks fixes the unachievable frequency requests so it keeps working more stable. I will know this when i add a secondary psu to my system. But let's say the psu is ok, in this case considering the gpu is faulty in terms of p3d when everything else working great its completely up to one's own wish. Maybe we shouldn't compare p3d with other games neither any other benchmark tools. If it is up to me i can live without p3d too quit the sim and consider my gpu isn't faulty. But i already consider my self won this battle against this error besides with a faulty hardware long time ago by doing some cheap tricks. I think i m done with this thread and i hope anyone who deal with this error can come up with positive results in short time. Thanks to everyone who tried to help and shared their experiences their ideas you are such a great community.

Good luck! 

I have completed one more flight today without any error the flight took 4 hours around and the sim run near 5 hours with many apps running in background. switched between full screen and windowed mode from time to time but mostly in windowed mode. If you remember the catch which i discovered the buffer too small warning found by the process monitor gave me the idea that there must something be done with power settings to reduce the power consumption because the buffer too small warning was associated with current control set registery path. And this absolutely confirms those who was keep telling this issue has something to do with power options. And then I went right in to bios looked for settings that won't force the power consumption for the better performance. Switched balanced power plan in Windows and watched cpu speeds on idle cpu kept working between 1 to 2 ghz on idle as i launch the p3d cpu speed increased to 4ghz and never dropped during under load. Also locked the gpu clocks just in case and the gpu core clocks never fall below 1900 mhz. As i stated previously before i go in bios and set some settings i don't remember exactly what i changed but no matter i switch balanced plan in Windows I watched the cpu kept working at maximum speeds. There was something not right It seems in modern motherboards things more complicated than we can imagine. There are some conflicts between modern bioses and os especially with Windows10. So just dont be offended by switching balanced power thinking that u can't take advantage of your super fast hardware and i think balanced power won't force your system beyond what your hardware can't really handle but always will run at maximum speeds as your hardware allows.

37 minutes ago, joepoway said:

I could be wrong but you are actually enabling the power “ saving” feature and this seems similar to what Chris did as the only thing that helped him by reducing the power limit to his GPU to 70% if memory serves me.

Its interesting how a few of the positive results come from either trying to limit the frequency cycling of the GPU because of GPU Boost or reducing the power to it, perhaps this feature that Windows can take advantage of that you are doing by editing your BIOS is having a similar effect?

Joe

You are right Joe. And this is what actually i try do. i should warn that if things set to performance mode in bios no matter you switch to power saver mode within windows power saver plan won't take effect at least on my bios this is what i experienced. i guess switching to auto lets windows to handle the power plan and i plan doubling my power supplies in future so one for gpu one for motherboard and see what happens no need to have more than 300 watt for a 150 watt gpu it's barely peaks at 200 watt and a used one psu would do it. 

6 hours ago, glider1 said:

Don't have Native ASPM option in my bios unfortunately.

Have you ever tried doing a test flight by switching the balanced in power option in windows? Also try enabling the power saver options in your bios settings if any available. try everything that will reduce the load of psu. if you are not sure what to do in bios don't even touch it. unplug all unnecessary devices from the usb ports. BTW i made my second flight and this time i pushed a  bit more harder than the previous flight and set my resolution to DSR X4  also live streamed on the same hardware and i got no error. still not sure but usually i was getting this error within an hour.

I completed the flight without any error this time all the gpu clocks were set to default nvidia control panel settings inspector and everything was default. I am not sure yet this is a permanent solution time will show.

here is what i did in bios. Enabled PCI express naive power management and native aspm.

Ok now i do another test again i have changed a setting in bios related to pcie power management it was set to disabled by default means that aspm will be handled by the bios enabling it allows to os to handle the power management i m flying now on windowed mode switching between other apps obs working and keeps sending data to youtube with CBR set to 6500 process monitor also working and keeps logging as well so far good if everything goes well i will share a screen shot of bios about what i changed in bios. 

Here is stream link if u are interested 

12 minutes ago, glider1 said:

Do you know of a way to do a reliability test on just the GPU's VRAM nothing else? What card do you have a brand new 1080ti?

The only reliable way to test is p3d, from now on i consider all the benchmark tools are trash. My card is Gtx 1070 ichill inno3D x4.

Ok now i can confirm the exact time when the dxgi came. i figured by watching my streamed video. when the time second was stepping to 31 from 30 i started count till dxgi appear and it came after 45 seconds and there is only one unique event logged at that time is ''BUFFER TOO SMALL'' associated with this path: HKLM\System\CurrentControlSet\Control\Notifications\41C61629A3BC60B5

here is a SS: https://prnt.sc/lvr5k8

8 minutes ago, cj-ibbotson said:

Its a Inno3D card and will have to be returned to Hong Kong

Hi cj just letting you know this might be a sign for us and my card also brand inno3D x4 and i am facing this issue too in p3d strangely. i hope i can find a way to eliminate this by the support of process monitor. have you ever run the process monitor while p3d get this error and i think we can compare our monitoring results and see what differences are there in the logs.

1 hour ago, Jim Young said:

Just make sure you immediately write down the computer time when it crashes.  You'll only have to look at that one minute.

Ok it's crashed while the monitor was capturing the system time was 8:31 by the time crash. but there are too many entries logged between 31 to 32 i am still looking into it and the only entry unusual and not similar to the others is ''BUFFER OVERFLOW'' and this result not associated only with one path. 

See the video i was streaming at the time crash happened watch starting from 4:29

in 17 minutes 1.5 gb of data logged that's crazy to trace the seed of the issue so lets hope dxgi comes less than in an hour. happy ghost bustings 🙂

27 minutes ago, Jim Young said:

Maybe this program will also tell you what was being loaded when you get the DXGI Error.  It loads thousands of objects a second but you can probably get a good idea as it is usually loading AI or an aircraft for several seconds.

Ok. Now i am into it. But one question i have read in CTD guide that running process monitor more than 60 mins can cause another crash because file size getting larger as the time pass. Can i save the log to another drive and will this prevent any crash caused by process monitor? 

12 hours ago, Jim Young said:

I understand what you are saying and that's great you are trying to help others that might be having this issue.  FYI, P3D does not load up the entire world when you start it up.  I think it is something like a 300 mile or less radius that is loading depending on what you have installed.  Photoscenery will load no matter where you are located in the world so you do not want to have very much photoscenery loaded that you will not be using in your flight plan.  The default scenery does not load worldwide.  Orbx/FTX stuff does not load unless you are near it.  Anything in the Scenery/World/Scenery folder will load.  Anything in the Addon Scenery folder will load unless disabled.  If you want to see what exactly is loading, set up Process Monitor as suggested on page 18 of the AVSIM CTD Guide.  This will only load P3D and P3D addons.  You can then see exactly what is loading during the flight.  Of course, weather and AI is loaded as flight plans have to be enabled worldwide but this is nothing.  There really is not much placed on your CPU/GPU if you configure your sim properly for your flight plan.

I once did a flight from Chicago to St Louis with only scenery covering the flight plan enabled.  I was shocked to see Orbx/FTX scenery objects loading in California and the Pacific Northwest.  Upon further research, I found these .bgl's in the scenery/world/scenery folder that Orbx installers placed in these folders.  Still, it was minuscule in comparison to photoscenery I might have accidentally kept enabled in Europe. 

Are you saying that disable any scenery remains out of our planned route? is there any tutorial about this what should be disabled and what shouldn't so the simulator won't load any unused scenery in cache? And what the simulator and pc would have profit of doing so? As far as i know the simulator loads up everything but unused objects remains compressed in memory till we reach a certain point or mile by mile the simulator decompresses the compressed textures as they step in range. 

5 hours ago, glider1 said:

I got the DEVICE HUNG again! This time no overclock on a brand new 2080. The plane was parked after completing the flight. I was in the cockpit in VR and had alt-tabbed out to Chrome browser on the virtual desktop to type my logbook up in Chrome sheets app. P3D went DEVICE HUNG but the video driver was unaffected (I know because I was interacting with Chrome in the virtual desktop with the headset on and the video driver functioned perfectly no interruption to my VR experience while P3D went DEVICE_HUNG). With headset still on (which is still using the GPU) I started MSI afterburner on the virtual desktop and noticed that when P3D reported hung, the core clock on the 2080 had throttled back to 1515MHz when I wanted it boosted to 1830MHz as it was during the whole two hour flight. The GPU temps, CPU i7700k temps all systems running fine when the DEVICE_HUNG was reported. GPU usage was probably 50% and CPU core 0 wasn't even pegged at 100% the other cores were basicallly idle.

There could be something in the theory that we have to stop the clock throttling down into idle with P3D if we are going to interact with desktop apps while the sim is running.

One thing for sure, it is possible for P3D to report DEVICE_HUNG when the device has not hung.

EDIT: from reading other peoples reports, it can be related to when there is more than one monitor. VR uses three monitors. It uses flatscreen monitor and one display for each eye.

Have you noted the time of error when occurred? i bet there is a log that indicates to: ''display driver nvlddmkm stopped responding and has successfully recovered'' go to event viewer and look at windows logs/system there must be a warning related to nvlddmkm when the sim got device hung error.  After the creator update no more messages appear related to a faulty driver in older days it would go blue screen now windows by passes many errors we don't recognize.

2 minutes ago, Jim Young said:

Unfortunately, that has been suggested before to this member by many and not a solution.  If you look in our CTD Forums (pinned topics), you will see many posts by Cj and he wants to overclock his GPU because he paid for it and it is high end and refuses to allow this error to stop him for using his "powerful" GPU with higher settings and it is only the fault of LM and they need to fix their program immediately.  There is no solution for him and there never will be one.  I have worked with him for a long time and I am finished wasting my time.  Just let him continue to blame LM for his woes.  I get more and more upset reading his issues over and over again to different people who all tell him the same thing and his only response is mostly, "sadly, this didnt work for me".

My suggestion isn't aimed on a specific person so think wide there are many people reading this forum and find their answers for their own questions by reading these threads. I know one thing very clearly when it comes to p3d. it doesn't matter how fast you run your gpu as long as you can't keep your system balanced you will get errors one way or another. So bottleneck must be minimized if you want your sim work more stable, or minimize the resource amount to pass it easier thru bottleneck. Think about it once you launch the simulator it loads up the entire world in to your system and process it. 

4 hours ago, cj-ibbotson said:

Sadly this didnt work for me 😞

Ok. i am feeling really bad anything related to this error. but if your card is new and still under warranty and if you are sure that this isn't psu issue i would suggest you to test your gpu on another pc depending on result RMA it or request new one. this may even be related with the motherboard who knows? God knows whats going on inside those chips and circuits. but the interesting thing in  my case this problem comes up with p3d only. my gpu is already out of warranty so i don't mind locking gpu clocks. actually before this solution i wanted to enable igpu feature in bios then with two hdmi plugged to the same monitor and enabling extend display feature so the p3d can recognize both integrated and discrete card and just test it this way. i will do this with clock speeds set to default this time and report back here.

1 minute ago, cj-ibbotson said:

Thanks for this, i googled it earlier and seen a video. I did notice it resumes normal behavarior if I close or restart Afterburner.  Ill look at your video.  Mine still fluctuates a little.  When I pick the dot furtherest to the right on the curve it says 2025 mhz but in Afterburner it jumps about between 1980-2025 and has dropped down to 1770 min according to GPU-Z so its not locking completely.  The only 'fix' that worked for me was reducing the gpu Power Limit to 70 or 75%, 100% guarantees a crash so testing now at 100%

To understand if MSI AB lock your clocks, apply your changes then exit any other apps such as steam and gfe and check your gpu p states via nvidia inspector and actually compare this to default settings my point is here keeping the gpu work above idle clock speeds. in my situation idle gpu clock speeds were which is P8 that pulling clocks down to 139 MHZ when i move the mouse cursor or  run any small 3d process it switches to 1620 MHZ which is defined as base speeds under heavy load it boosts to 1974 MHZ on default profile. So you should compare and make sure gpu clocks never goes down to idle state to eliminate the issue you are having with sim.

1 hour ago, cj-ibbotson said:

Im curious to know how you lock the clock speeds? I understand you can do it for the processor so that it always runs at its max but I wasn't aware its possible for the gpu.  I would be keen to test it.

Chris

i locked via msi afterburner.  run it and on msi screen press CTRL+F  and then a new screen will pop up called voltage frequency curve editor on the editor screen pick one of the range which you wish one of between first square and the last one (don't change the curve) and click on the square which you picked then press L on the keyboard then click apply on MSI AB you can also save it to a profile. here is i made a short video: (I don't know whether this method work on your gpu model. in my case once i locked the clocks no matter i go idle clocks won't change) If you can't lock just because of your gpu model try nvidiapowermizermanager if you can find it.

13 hours ago, glider1 said:

Is there some way to log the idle state that you know of? It would be interesting to see what the clock does when you switch out of the sim onto other desktop apps.

You can check it via nvidia inspector or msi AB if you expolre any value below 400 Mhz with msi AB that means your gpu works at idle speeds. Also make sure that nothing is working on the background after you switch out of the sim for example a steam app or gfe or any web browser may trigger the gpu to work at higher clock speeds. Values may vary depending on your gpu.