Windows 10 Frequent, Semi-random BSODs (Thread stuck in device driver)

SpaceKitkat

Member
Joined
Sep 17, 2016
Messages
42
As the title says, almost* every time I launch a video game (specifically, a minecraft mod), I get a BSOD with the error "THREAD_STUCK_IN_DEVICE_DRIVER".
I'm also getting them more and more frequently just using google chrome, even just now while trying to get minidump log.
Checking the log, there was something about atikmdag.sys, and my problem seems to be very similar to this thread:
However, that thread's main solution was when someone checked the logs, they listed drivers that needed to be removed, and I don't think I have any of the same drivers.

After a bit of googling, a couple "easy" solutions were to update my BIOS and reinstall GPU drivers.
So I uninstalled all of my display drivers, deleted everything AMD, used DDU (Display Driver Uninstaller), restarted into my BIOS, updated my BIOS successfully, installed the new GPU drivers (Radeon Crimson edition software, for an R9 290), and restart my computer.

I've also performed several Windows diagnostics, including a RAM check, disk checks on all of my hard drives, and a "sfc /scannow" via CMD. All have come up with no errors.
I tried to add the minidump W7F log, but it said the file size was too large.. lol (RIP).
I've uploaded it to
but if it needs to be uploaded elsewhere, just let me know.
If any other logs are needed, I'll supply them ASAP. Any help is very much appreciated, as I work from home and this is currently my only available computer!
 


Solution
I had a look through your thread and signed up to chime in.

I have an R9 290 and the exact same problem, it has plagued me for the majority of the year. You may stop butchering your PC now.

It is categorically a result of AMD's drivers.

I am not exactly sure at what point the issue arose, but the same very frequent yet erratic BSOD/crashes on wake up have been present for the last dozen or more driver versions.

A poster on another forum found a BIOS altering solution that I didn't want to replicate, but this is what he discovered. It seems related to a new driver based power saving system that lowers the clock speeds of your card during idle situations. This was "improved" at some point over the 16.xxxx series of...
Well, I bit the bullet and downgraded to 15.11.1
I'll update if I keep getting BSODs.
(I got 5+ in the last 2 days, incredibly frustrating lol)
 


Hello You two,

Basically Circinae is correct. I am not 100% sure that the option AMD suggested is exactly the registry entry changed by the tuning tool.

AMD support is very frustrating.
They have a ticket system and if I reply there are different persons in charge almost every time. Sometimes the person does not even read the ticket history and comes out with pretty standard answers.
Also very high effort recommendations are there even if they are not logical, like "Please reinstall windows".
I said, this does not sound logical as the error is gone without the gfx card and also gone for older driver. How should windows system files be the root cause then.

They also said, XFX suggested the card is broken, so bring it to your next qualified local tech support and have it checked out.

I said the argumentation of XFX that the card is broken was:
It does not work with newest drivers -> It must be broken.
Right.
Then my warranty is over. Where should I bring it now.


Overall support from AMD tech support is not very strong because not the same person is discussing the ticket and almost no continuity is in the discussion.

One time they suggested me to contact the driver team via a support form and report it to them.
They told me to ad a special line at the end of the description to bumb up the priority. Unfortunately i did not fully understand that statement until i already send the report to driver team and I got no response.

I am frustrated about all of this. For me it would be more cost efficient to by a new Nvidia card than to continue to fix this.
 


As I posted earlier a new driver came out only last week.
 


I am frustrated about all of this. For me it would be more cost efficient to by a new Nvidia card than to continue to fix this.

Yep. I tried to downgrade to 15.11 and it doesn't seem to have worked, I'm still on 16.~
I uninstalled the drivers, then used DDU in safemode, then installed the 15.11 drivers. Got a BSOD earlier, checked my version, and it's still 16.~
Weird.


Also, yeah, I'm inclined to agree with you, Sven. I really didn't want to upgrade to a new GPU so soon, but I'm about to just say ..... it and get a GTX 1080.
 


Last edited by a moderator:
Hello,
I just reply to this old thread. I rebuild my system and took away the air cooler from my R9 290X and retrofitted my EK waterblock.
Basically i had it in my cupboard for long time now but i was too lazy to rebuild my system.
With the old air cooler there was not enough space for the pump and the pump was giving a bit of physical strain onto the card.
I suspected this could be also part of my problem.
With the rebuild system i now gave it a new try with the new drivers. I just installed it (17.3.1) and I discovered the amazing Wattman!
AMD gives you complete control about the dynamic clock switching.
You can customize the frequencies of every step and the voltage of every step. This means if the frequency switching is responsible for the BSOD problems, wattman in the driver 17.3.1 should help to fix it.
I will now first try it with my new hardware setup with the unmodified driver.

BR

Sven
 


Thanks for the update Sven and best of luck with the testing. :)
 


Hey kemical,

you are still on the job ;)

All my crashes are still there .
Now i discovered that FurMark burn in Test also produced the black screen issue under full load, which I before never observed.
Before it was just happening at light load or in idle case.
I experimented with down clocking the core and the memory and it seemed to be memory related, which means down clocking the memory already made it stable, while the core was at stock speed ( I down clocked to minimum mem clock which was 150 MHz).
Increasing the memory clock (to 1400MHz) let it crash very quick .
Now I upped the voltage of the memory via WattMan for a first try by +70 mV and FurMark ran stable for around 25 minutes (until i stopped the test).
Then I went back to + 40mV and it ran stable for around 35 minutes (until I stopped the test).
 


update: All of that didnt't work.
I have a new strategy now.
I modified the VGA bios with a tool and did increase the frequency from the lowest power state from original 300 MHz to 420 Mhz.
This means in idle case core clock will never go below 420 MHz
And I upped the core voltage by +44 mV. I am not sure if that is necessary. No crashes so far yet.
It has been half a day now.
BR
Sven
 


Last edited:
Hi
update: All of that didnt't
I have a new strategy now.
I modified the VGA bios with a tool and did increase the frequency from the lowest power state from original 300 MHz to 420 Mhz.
This means in idle case core clock will never go below 420 MHz
And I upped the core voltage by +44 mV. I am not sure if that is necessary. No crashes so far yet.
It has been half a day now.
BR
Sven
Thanks for the update Sven.

Looks like you have had some success and hope it continues.. I'll keep my fingers crossed!
 


I got the idea from this youtube video:

Unfortunately it did not work with that ASUS tweak tool. It did not recognize my card.
So i hacked the bios to up the clock to 420 MHz.

BR
Sven
 


Hello kemical,
i finally gave up. I even tried it with +95mV vcore but the card is still crashing.
I ordered a new card from Nvidia and will replace the AMD card. My wife said I have already invested too much time.
The new card has 3 year warranty with watercooler already fitted, when it shows the slightest weakness i will RMA it!
BR
Sven
 


Good luck with new card Sven, you will have to post what card it is and how your getting on.. :)
 


Back
Top