Current setup display driver crashing every 5 minutes - Get a new GPU?

Gingervitis

Hello,

I really don't know what to do at this stage, other than get a new graphics card.

Every time I run something relatively graphically demanding, I get a black screen for 3 seconds, which crashes the game. Event Viewer classifies it as Event ID 4101: Display driver amdkmdap stopped responding and has successfully recovered.

Current System specs:

• Acer GN246HLBbid 24 inch Wide FHD LED Gaming Monitor with 144 Hz, 1 ms, 350 nits, DVI, HDMI, Acer EcoDisplay - Black
• ViewSonic 22" Monitor
• Samsung 2.5 inch 250 GB 850 EVO Solid State Drive
• Seagate 1TB HDD
• MSI AMD AM4 B350M Bazooka Ryzen 7th Gen DDR4 VR LED M.2 Micro ATX Motherboard - Black
• Corsair CMK8GX4M2B3000C15 Vengeance LPX 8GB (2x4GB) DDR4 3000Mhz CL15 XMP 2.0 High Performance Desktop Memory Kit, Black
• AMD Ryzen 5 1600 Desktop CPU - AM4/Hex Core/GHz/16MB/65W
• Seasonic G550 550W 80+ Gold Certified PSU Semi Modular Power Supply
• Sapphire 11260-07-20G AMD RX 480 Nitro+ 8 GB GDDR5 Memory PCI-E Graphics Card - Black

I have formatted my SSD, reinstalled Windows, reinstalled the motherboard chipset drivers, rolled back my AMD drivers from 17.7.2 to 16.11.5 using DDU, and the problem keeps coming back.

I have had this GPU since October 2016 and it has never acted like this until I replaced my motherboard, memory and CPU. I have not overclocked or underclocked it. Has it been damaged? Is there an issue with the PCI express mounting? Is there a power issue?

At this point, the only solution to me is to ditch the graphics card and shoot for a 1060. Is there anything else I can do?

ReZonator

Trouble shooting can be tricky, but in a situation like this you have to rule out suspected components to find the fault. The first place to start would be software, so try an older GPU driver to see if it makes a difference.

Next, if you can, borrow another graphics card to rule out a GPU fault. If another card works then you have a bad GPU. You could also try your GPU in a friends system. If the GPU works in another system then you move on the the next component.

The next most likely issue would be the PSU. If another PSU stabilises the system then get another reputable PSU, if not you move on.

If system is otherwise stable it probably isn't RAM, but you can check your RAM using free software, and I think you can do it with Windows software.

Next in the suspect list is your Motherboard. Last would be your CPU. Motherboards die, but CPU's are generally thought to be resilient, if not abused by overclocking. So you can assume the motherboard is at fault after ruling out the Drivers, GPU, PSU, and RAM.

Gingervitis

Regarding the BIOS, I downloaded MSI Live Update 6 and downloaded all the essential motherboard drivers. I had already flashed the BIOS of the B350m before I did a re-install of Windows 10. Unless you're talking about the GPU specifically, then no, I haven't, nor know how.

I rolled back the driver to 16.11, and the same issue happened, only this time, there was artifacting on the screen.
I updated the drivers to 17.8.1, and was able to get a few games of Overwatch in without crashing, but that may be a once off. I'd need to test it more.

The Motherboard, CPU and RAM were all purchased in the last month, so unless I'm very unlucky, I'll rule them out of the time being.

I was told that perhaps the new configuration is drawing more power from the PSU, and causing the GPU to not get enough power briefly, causing the unresponsiveness. Maybe if I get a dual 6pin to 8pin power connector, that may stabilise the system. Failing that, I'll have to loan the card to someone else and see if the issue can be replicated.

Thanks for the replies.

jumbobreakfast

The MSI motherboard you have comes with some MSI gaming boost application. Make sure that no overclock or gaming mode is set there. Also doublecheck the BIOS to see if there is some overclock running there and disable it.

You only have 1 PCIe slot but you could try using a different 8-pin power connector as PSUs usually share power connectors off different rails so that particular rail might have reached it's powerdraw limit.

Probably a stupid thing to ask but have you plugged in both power connectors to the motherboard? The long 24 pin and small 8-pin on the other side...It probably wouldnt work without it anyway.

CruelCoin

It's an old thread (2011) but guy there calling your issue a TDR issue, which ReZonateor above says you need to troubleshoot, but misses a few bits that the poster mentioned:

https://www.sevenforums.com/drivers/148860-error-4101-display-driver-amdkmdap-stopped-responding.html
"Poor Cooling
Problems with the power supply
Overclocking Issues
Bad System memory or incorrect memory timings
Defective PC Components"

Check that temperatures are ok, check that the card doesn't have some sort of power saving mode. If it does, turn it off.

The guy suggests you use something like Prime95 to stresstest your system, which should highlight errors for you.

Yeah, he gives a huge list of things to check...

Personally, i would be leaning towards the PSU not being beefy enough. Feed your components into this, and it will give you figures on what you need for your PSU. https://outervision.com/power-supply-calculator
I fed yours in and (without wireless/sound cards/liquid cooling) it's saying it needs 480w @ 100% CPU use.
I had read somewhere that you should get a PSU with 25% above the draw, to cater for momentary spikes in draw, meaning you'd need a 600w PSU to allow for whoopsies. The page itself is reccomending a 650w PSU for you.

Riesen_Meal

I was banging my head off a similar PC problem where mine was overheating and turning off for no good reason the last few days, turned out it was the stock fan on top of the heatsink on the CPU, I never even thought it could have been that only for I went to see was there any BIOS updates and saw huge temps on the CPU....

Replaced it with an old stock fan I had spare and it's fine since...

The fan was spinning and "looked" fine while operating also with the cover off, just wasn't cooling the way it should have...

Worth a look also to eliminate another part...

CruelCoin

Another thing to try:

A TDR error can be caused when the graphics card does not report to Windows for 2 seconds or longer.
When windows detects this, it kills the graphics driver.

Try upping this kill time from 2 to 8 seconds to give your graphics card more time to do what it needs to do before being killed off.

Increasing the default TDR setting to 8 seconds gives the graphics card longer to complete its task. How to do that:
Open registry editor..run > regedit > HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\GraphicsDrivers. In the box to the right you may or may not see TdrDelay with a data value of (2). If you see that, double click and change the value from (2) to (8). If you don't see TdrDelay...right click on a open space and create a new Dword and name it TdrDelay . Double click that and set a value of 8. I do not recommend using a 'tool' to reset this value.
Note: The computer must be rebooted after making this change.
https://community.amd.com/thread/180166

Gingervitis

Thanks for the replies everyone, I appreciate it.

Regarding the temperatures, I'm monitoring them using the MSI Command Center and Sapphire Trixx. Temperatures are about 51-55C and ~75C under load for CPU and GPU respectively.

Everything is plugged in, and running for the most part fine (130fps on Overwatch etc). I've checked the BIOS, and disabled game mode and any overclocking. It's running at base 3.2GHz.

What I'm going to do this evening is have a look at the power cables running from the PSU. It could be that the rail supplying power to the GPU is also powering something else like the SSD or HDD, which would be the culprit.

As regards the PSU not being beefy enough, I hope that's not the case. Would a Ryzen 1600 and b350M motherboard draw significantly more power than an i7 2600k and Biostar H67MH combination? I never had power issues with the previous set up (for the 6 months I had the previous setup)

I tried setting the TDR to 8 seconds, but that just increased the black screen duration to 8 seconds!

I'll try Prime95. I tried to use Furmark, but that wouldn't even start ("could not create a render window. furmark startup failed")

CruelCoin

Gingervitis wrote: »

As regards the PSU not being beefy enough, I hope that's not the case. Would a Ryzen 1600 and b350M motherboard draw significantly more power than an i7 2600k and Biostar H67MH combination? I never had power issues with the previous set up (for the 6 months I had the previous setup)

I tried setting the TDR to 8 seconds, but that just increased the black screen duration to 8 seconds!

Hopefully not, but i'm not that knowledgable.

That's really weird with the 2-8 seconds thing. No idea whats going on there.

There is an article i really like that has the author saying: "computers are like schizophrenic children with auto immune diseases.". Couldn't be more true.

srsly78

CruelCoin wrote: »

Another thing to try:

A TDR error can be caused when the graphics card does not report to Windows for 2 seconds or longer.
When windows detects this, it kills the graphics driver.

Try upping this kill time from 2 to 8 seconds to give your graphics card more time to do what it needs to do before being killed off.

Increasing the default TDR setting to 8 seconds gives the graphics card longer to complete its task. How to do that:
Open registry editor..run > regedit > HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\GraphicsDrivers. In the box to the right you may or may not see TdrDelay with a data value of (2). If you see that, double click and change the value from (2) to (8). If you don't see TdrDelay...right click on a open space and create a new Dword and name it TdrDelay . Double click that and set a value of 8. I do not recommend using a 'tool' to reset this value.
Note: The computer must be rebooted after making this change.
https://community.amd.com/thread/180166

Only people doing GPU compute need to deal with this. Should never be an issue for normal usage (graphics) - should not be taking 2+ seconds to render a frame

CruelCoin

srsly78 wrote: »

Only people doing GPU compute need to deal with this. Should never be an issue for normal usage (graphics) - should not be taking 2+ seconds to render a frame

Yeah, just throwing out simple things to try, in the hope that he doesn't need to start dissasembling his system to test components.

Throw enough ****, something will stick.

Danjalical

I had an issue like yours, I bought a GTX 1080ti and long story short, turns out i had damaged one of the pcie 8pin adapters. I could tell because a simple tap on the end of the cable pluged into the gpu would cause the power to flicker from the 'white' okay light to the 'red' bad light on the gpu. thankfully mine is modular so I just replaced it with another in the box.

Also looking at your PSU, 550w might a bit low for your setup too.

Hope this helps out

Wossack

get a system monitor and check your +12V rail to see if it dips under load, less then 11.4 and thats an issue

Gingervitis

Ok, I'll download HWinfo this evening as well. I'll also swap the cable onto a different output from the PSU, see if that makes a difference.

Gingervitis

Wossack wrote: »

get a system monitor and check your +12V rail to see if it dips under load, less then 11.4 and thats an issue

Played TitanFall 2 and it crashed within about 5 seconds of starting a match:

HWiNFO:
GPU VRM Voltage IN (VIN/+12V): Current = 12.125V | Minimum = 12.125V |Maximum = 12.126V | Average = 12.126V

PSU has 2 available slots for PCI-E power cables. I've tried both. Game crashed both times.

SickBoy

Gingervitis wrote: »

Played TitanFall 2 and it crashed within about 5 seconds of starting a match:

HWiNFO:
GPU VRM Voltage IN (VIN/+12V): Current = 12.125V | Minimum = 12.125V |Maximum = 12.126V | Average = 12.126V

PSU has 2 available slots for PCI-E power cables. I've tried both. Game crashed both times.

Connect 2 pcie cables to the PSU and join them using a 2 female 6 pin to one male 8 pin and see how that goes.

Gingervitis

SickBoy wrote: »

Connect 2 pcie cables to the PSU and join them using a 2 female 6 pin to one male 8 pin and see how that goes.

Yeah, that's my next step.
If that doesn't work I'm going to lend the card to someone else and make sure that it's the point of failure. If that works on another PC with no issue then it must be insufficient power and I'll get a 600 / 650W PSU; it would seem that the replacement mobo and CPU must have tipped the balance.

SickBoy

Gingervitis wrote: »

Yeah, that's my next step.
If that doesn't work I'm going to lend the card to someone else and make sure that it's the point of failure. If that works on another PC with no issue then it must be insufficient power and I'll get a 600 / 650W PSU; it would seem that the replacement mobo and CPU must have tipped the balance.

50% more cores and not as power efficient as Intel would do that.

Gingervitis

Update: 2 PCI-E power cables combined using a 2 female 6 pin to one male pin does not help. Tried 2 quick games, and it crashed within 5 seconds.

dreamers75

Similiar issue last year and it was the power supply.

Gingervitis

dreamers75 wrote: »

Similiar issue last year and it was the power supply.

Defective? Or just underpowered?

TerrorFirmer

Defective, if it is the PSU. 550W is loads for that setup, especially with a higher end supply.

dreamers75

Gingervitis wrote: »

Defective? Or just underpowered?

Old I reckon, the power coming from it was intermittent.

Your MSI board should have a BIOS code that may help diagnose it.

Gingervitis

TerrorFirmer wrote: »

Defective, if it is the PSU. 550W is loads for that setup, especially with a higher end supply.

Would that account for everything working fine 90% of the time, except when the GPU is under load? Or, could the problem be with the GPU? (Guess I won't know until I test the part in another system...)

dreamers75

Gingervitis wrote: »

Would that account for everything working fine 90% of the time, except when the GPU is under load? Or, could the problem be with the GPU? (Guess I won't know until I test the part in another system...)

Or test a card in your system, it would solve something anyway.

EoinHef

Gingervitis wrote: »

Would that account for everything working fine 90% of the time, except when the GPU is under load? Or, could the problem be with the GPU? (Guess I won't know until I test the part in another system...)

Yeah. CPU clock speed boosting and GPU being put under load in games will give you your highest power draw in a gaming system. Web browsing,watching videos and daily use everything is pretty much in powersaving mode. Modern hardware has something called power states. The cpu and gpu scale the power and clock speeds based on what the PC is doing so they are always as power efficent as they can be.

Man these problems can be a bummer,changed a PSU for someone recently with a very similar problem. Driver wasnt crashing but the whole PC would just turn off under load and couldnt be turned on again for about 24hrs. Presume the capacitors had to drain in the PSU to be able to work again.

Gingervitis

Well, it's not the GPU anyway!

Put the 480 into a new system:

Xeon E5-2650
64GB RAM
635 watt PSU for a Dell T3600 Precision
128GB SSD

and the result is no crashes on Overwatch. So, it looks like I'm picking up this: EVGA SuperNOVA 650 G2

Thanks to everyone for getting to the bottom of this, googling error codes on other fora is not infallible!

SickBoy

That's good news.
Nice PSU choice also.

CruelCoin

Oh for the days when error-codes are written by people who understand others may need to read them!

Post back to let us know if it worked for you!

Taylor365

Gingervitis wrote: »

Well, it's not the GPU anyway!

Put the 480 into a new system:

Xeon E5-2650
64GB RAM
635 watt PSU for a Dell T3600 Precision
128GB SSD

and the result is no crashes on Overwatch. So, it looks like I'm picking up this: EVGA SuperNOVA 650 G2

Thanks to everyone for getting to the bottom of this, googling error codes on other fora is not infallible!

How old is your setup by the way?

Gingervitis

Taylor365 wrote: »

How old is your setup by the way?

Original setup was 2011.
Replaced GPU twice (560TI for 280x in 2014, then 480 last year)
Replaced PSU once (Some Korean brand for a Seasonic 550 this year)
Replaced CPU (i7 2600k for Ryzen 5 1600 last month)
Replaced mobo (Biostar H67MH for Bazooka B350M last month)
Replaced RAM (8Gb DDR3 1300Mhz for 8GB DDR4 3000Mhz last month)
Added SSD last year.
The only things from the original build now are the case and the HDD.

Taylor365

Gingervitis wrote: »

Original setup was 2011.
Replaced GPU twice (560TI for 280x in 2014, then 480 last year)
Replaced PSU once (Some Korean brand for a Seasonic 550 this year)
Replaced CPU (i7 2600k for Ryzen 5 1600 last month)
Replaced mobo (Biostar H67MH for Bazooka B350M last month)
Replaced RAM (8Gb DDR3 1300Mhz for 8GB DDR4 3000Mhz last month)
Added SSD last year.
The only things from the original build now are the case and the HDD.

Well hope you get a refund on that PSU!

I built my rig in 2012 and have only upgraded the GPU. My Superflower 550w still going strong, even with an OC on the cpu and a 390! (Was pulling near 460w from the wall.)

Gingervitis

Ok, new PSU has come in and is installed. I'm not getting random black screens any more, and all my games are working without issue (higher temperatures not withstanding). EXCEPT Titanfall 2. Getting the exact same problem - black screen, error 4101 in the Event viewer. I have the two 6 pin to 8 pins adapter in use for the 480, and TF2 ran perfectly on the older system (albeit a little slower due to the weaker single core performance of the Xeon.
I'm at a loss.

Edit: Now Overwatch is crashing again. I'm right back where I started, only out of pocket on a new PSU.

jumbobreakfast

I'm afraid you'll need to do more work to narrow down the problem (trying someone else's card in your system, running with only 1 RAM stick at a trime, that sort of painful stuff)

Windows 7/8/10 has a built in memory tester you could run too:
http://www.thewindowsclub.com/windows-memory-diagnostics-tool-in-windows-7

Also check the Event Viewer in Windows:
https://www.howtogeek.com/school/using-windows-admin-tools-like-a-pro/lesson3/

Gingervitis

Ran the memory tester, no issues.
If it's not the GPU, which works fine in the other system, and it's not the PSU (which was bought today), then it must be the CPU-motherboard-memory combination, which was only bought last month!

Anyone in the Limerick area willing to lend me a spare GPU?

Moon54

OK. Try rolling back the AMD driver to 17.6.2.
Choose clean installation when installing.

I was having driver crashes in Overwatch but since installing the 17.6.2 one I have been rock solid.
https://support.amd.com/en-us/download/desktop/previous/detail?os=Windows%2010%20-%2064&rev=17.6.2

Gingervitis

I finally got it fixed! (I hope)

The solution was to flash the BIOS of the B350M Bazooka to version 1.6 (release date August 2017) which automatically downclocked my 3000MHz RAM to 2166MHz, XMP turned off. I didn't make any changes to the BIOS settings and now everything is running without a hitch. (Running default AMD drivers downloaded from Windows - 17.1.1)

I just hope that running the RAM at this speed doesn't cause wear and tear. Even at 2166Mhz, it's still better than the 1366MHz my old RAM was running at. At this stage, as long as everything is running fine, I'm perfectly happy leaving it in this state.