GPU keeps getting crashed

Technical Support
1

Hi there!
For 2 days i have a problem with HotS. During the match the screen goes black, everything else is working (music keeps playing, i still can communicate with friends via discord, but HotS itself is silent during this). The only solution is to reboot my PC.

(it seems i cannot attach link to my DxDiag file, so here’s some specs:

  • Processor: Intel(R) Core™ i7-7700 CPU @ 3.60GHz (8 CPUs), ~3.6GHz
  • Memory: 16384MB RAM
  • Operating System: Windows 10 Pro 64-bit (10.0, Build 17763) (17763.rs5_release.180914-1434)
  • Card name: NVIDIA GeForce GTX 970
  • Driver Version: 26.21.14.3086
  • Current Mode: 2560 x 1080 (32 bit) (75Hz)

So i’ve been doing some digging, this is what i’ve found:

  • This is from the log file from HotS folder (2019-06-15 13.20.21 Graphics.txt):

Blockquote
GFX 13:20:21.070 ====================================================================================================
GFX 13:20:21.070 Heroes of the Storm (B74238)
GFX 13:20:21.070
GFX 13:20:21.070 Executable C:\Games\Heroes of the Storm\Versions\Base74238\HeroesOfTheStorm_x64.exe
GFX 13:20:21.070 -sso=1 -launch -uid heroes
GFX 13:20:21.070 Parent Executable \Device\HarddiskVolume13\Games\Heroes of the Storm\Support64\HeroesSwitcher_x64.exe
GFX 13:20:21.070 Grandparent Executable \Device\HarddiskVolume5\Games\Battle. net\Battle.net.exe
GFX 13:20:21.070 LocalTime 2019-06-15 13:20:21.070
GFX 13:20:21.070 noobe
GFX 13:20:21.070 DESKTOP-3MLESGT
GFX 13:20:21.070 <Exe.Architecture> x86_64
GFX 13:20:21.070 2.45.1.74238
GFX 13:20:21.070 B74238
GFX 13:20:21.070 branches/Heroes.45
GFX 13:20:21.070 589768
GFX 13:20:21.070 <Locale.Assets> enUS
GFX 13:20:21.070 <Locale.Data> enUS
GFX 13:20:21.070 <Locale.Install> enUS
GFX 13:20:21.070 RUS
GFX 13:20:21.070 2.17.2.6700
GFX 13:20:21.070 TelemetrySessionId 6A85F79C-4BA7-4BF4-B8F0-B000E178E857
GFX 13:20:21.070 ====================================================================================================
GFX 13:20:21.070 DXGI: Available
GFX 13:20:21.079 VRAM Override Variables: Dedicated [0MB] Shared [0MB]
GFX 13:20:21.198 Adapter “NVIDIA GeForce GTX 970” Vendor 0x4318 Device 0x4318
GFX 13:20:21.198 Adapter “Microsoft Basic Render Driver” Vendor 0x5140 Device 0x5140
GFX 13:20:21.198 Skipped adapter “Microsoft Basic Render Driver” because no ouputs
GFX 13:20:21.198 Attempting D3D11 device creation on “NVIDIA GeForce GTX 970”
GFX 13:20:21.258 Device creation success.
GFX 13:20:21.259 VRAM Override Variables: Dedicated [0MB] Shared [0MB]
GFX 13:20:21.262 VRAM Override Variables: Dedicated [0MB] Shared [0MB]
GFX 13:20:21.262 Initialized D3D11 GraphicsDevice: NVIDIA GeForce GTX 970 (Driver Version: 0.43086.0.0, VRAM: 4043MB, SVMEM: 8162MB).
GFX 13:20:21.262 Setting up swap chain.
GFX 13:20:21.263 Found matching display mode width: 2560 height: 1080 RRN: 74991 RRD: 1000
GFX 13:20:21.263 Resizing D3D11 swap chain (2560x1080)
GFX 13:20:21.265 Hardware Depth Sampling: Available
GFX 13:20:21.284 Resource Cache Override Variable: 0MB
GFX 13:20:21.284 Resource Cache: 2048MB Win64 (16324MB Physical, 7096MB Process Virtual)
GFX 13:20:21.284 Resource Cache Sound: 407MB
GFX 13:20:21.284 Resource Cache Common: 450MB
GFX 13:20:21.284 Resource Cache Texture Max: 1500MB
GFX 13:20:23.998 PerfNotifiers initialized
GFX 13:20:24.032 Focus Event. Changing to fullscreen.
GFX 13:20:24.273 Setting up swap chain.
GFX 13:20:24.273 Trying fullscreen
GFX 13:20:24.276 Found matching display mode width: 2560 height: 1080 RRN: 74991 RRD: 1000
GFX 13:20:24.277 Resizing D3D11 swap chain (2560x1080)
GFX 13:20:24.316 Resource Cache Texture: 1350MB
GFX 13:20:24.316 Resource Cache Texture Used: 0MB
GFX 13:22:42.253 Focus Event. Changing to windowed.
GFX 13:23:35.581 Focus Event. Changing to fullscreen.
GFX 13:23:36.327 Focus Event. Changing to windowed.
GFX 13:23:54.884 Focus Event. Changing to fullscreen.
GFX 13:24:00.241 Focus Event. Changing to windowed.
GFX 13:24:33.596 Focus Event. Changing to fullscreen.
GFX 13:24:34.884 Focus Event. Changing to windowed.
GFX 13:24:49.672 Focus Event. Changing to fullscreen.
GFX 13:27:04.795 DirectX call failed with error code -2005270523
GFX 13:27:04.796 PerfNotifier: [Render] [APIErrors] Occurrences [1] Peak [1.000000]
GFX 13:27:04.823 Attempting device reboot
GFX 13:27:05.742 Windows Timeout Detection and Recovery (TDR) detected!
GFX 13:27:05.746 Windows Timeout Detection and Recovery (TDR) detected!
GFX 13:27:06.117 Final device refCount is 15
GFX 13:27:06.117 Adapter “NVIDIA GeForce GTX 970” Vendor 0x4318 Device 0x4318
GFX 13:27:06.117 Adapter “Microsoft Basic Render Driver” Vendor 0x5140 Device 0x5140
GFX 13:27:06.117 Skipped adapter “Microsoft Basic Render Driver” because no ouputs
GFX 13:27:06.117 Attempting D3D11 device creation on “NVIDIA GeForce GTX 970”
GFX 13:27:06.118 Device creation failed.
GFX 13:27:06.118 Attempting D3D11 device creation on “NVIDIA GeForce GTX 970”
GFX 13:27:06.121 Device creation failed.
GFX 13:27:06.121 Device reboot failure, error code 537329682

  • This is the log collected by GPU-Z:

(and i cannot include the log file or screenshot of it either, so in short at 6/15/2019 13:27:01 PM everything was normal, GPU Clock was at 1278.3 MHz, Memory at 1752.8 MHz, GPU Temperature at 73 °C, GPU Load at 68%. Next entry is at 6/15/2019 1:27:04 PM, and everything is empty. CPU temperature was both at 68 °C)

I’ve tried reverting GPU driver to a previous versions, update them, redownloading both Battle net client and the game itself, disabling all overlays (discord, Windows game bar, Nvidia Experience), nothing helps.

Tried to play other games ofc, the result is perfectly normal (tried playing Witcher 3, GTA5, Divinity: Original Sin 2).

As can be seen, this is not happening due to power or temperature throttling, something else keeps crashing GPU driver.
Any help, guys?

ps. while trying different “solutions”, i’ve also managed to be thrown into a leavers-pool, ofc. Awesome. :+1:

2

even though you say you have done this already, just to be sure things are all cleaned up and stuff can you do it again anyway? :slight_smile:

download DDU and the driver.
uninstall the current driver.
restart in safemode and run DDU.
use DDU to clean up the driver remnants.
install the new driver.

Display Driver Uninstaller (DDU):

https://www.guru3d.com/files-details/display-driver-uninstaller-download.html

Display Drivers:

419.67 - http://us.download.nvidia.com/Windows/419.67/419.67-desktop-win10-64bit-international-crd-dch-whql.exe
3

I agree, a DDU is the best way to cleanly re-install a potential corrupted or troublesome driver and is strongly advised when TDR errors are occurring. Other than the suggestion made by Morogoth, the one crash we were able to locate on our end called out the DiscordHook, from the 11th to the 15th. Would you happen to have tried uninstalling Discord instead of disabling just the overlay to test?

This may not be the cause since TDR (Timeout Detection and Recovery) crashes can be difficult to pinpoint at times since the causes can be anywhere from other applications and overlays to hardware concerns. Could you provide DxDiag files? This should help with checking for clues in the error reporting, place the DxDiag between two ~~~~ like so:

~~~~
DxDiag here
~~~~
4

Rule out that it is a hardware problem by running furmark stress test. If that stress test causes a driver crash, artefacting or other kinds of errors then it is a hardware problem.

StarCraft II and Heroes of the Storm can generate work loads similar to those produced by furmark. Due to how efficient the shaders are at using parts of the GPU in parallel the power usage can reach near theoretical maximum. Even if the GPU die remains cool, the VRM might not and start to overheat. Few games create such workloads. For example an 8800GT (an old card now, poorly engineered by modern standards) could run Elder Scrolls Oblivion maxed out perfectly and remain cool but StarCraft II would cause it to quickly become unstable and even permanently damage the card’s VRM.

5

Here you go:

[DxDiag](https://drive.google.com/open?id=1s2tbv7mtYEfacsgAYu_6iUWi78kD5hgf)

I’ve tried DDU, used it in safe mode, reinstalled drivers, nothing.

Also i’ve tried another GPU from my friend, and the problem goes away. But still i find it very weird that HotS is the only game that produces this problem (tried FurMark on a maximum settings, temperature went for 77 degrees, but GPU is stable).

Uninstalling Discord, and also setting the game to Windowed mode and limiting FPS didn’t help.

Also now i need to play 4 games with another leavers. :confused:

6

If we’re running into issues like this we do recommend testing in a different mode such as AI to avoid leaver penalties. Unfortunately we’re not able to assist with those.

If the game is working fine on a different GPU then that does narrow down the issue quite a bit. is your friend by chance able to test with your GPU to see if they run into the issue on their end?

7

haven’t thought about it, we will test my GPU on his rig tomorrow, thank you

Edit:
Well, we switched GPUs with each other, and here’s the thing, now my GPU is working flawlessly in his rig, and his GPU in mine. No crashes, nothing. But switching it back and i have crashes again.

8

my immediate thoughts on something like that happening is power supply issues and/or windows OS.

can you tell us what exactly the GPUs are that are being used here and the powersupply sizes?

[Brand Model Memory/Power]

==> ex. Powercolor Radeon 390 8GB 
==> ex. Silverstone SX550 550W
9

sure, here it is:

My rig:
GIGABYTE GeForce GTX 970 4GB Mini ITX
Corsair RM670x 650 W

My friends rig:
GIGABYTE GeForce GTX 1060 WINDFORCE OC 6GB
don’t know the exact model of the PSU, it’s Corsair with 600 W

10

i found some things that you may want to try.

this will start up without all the normal background applications starting. if there is an application that is normally starting when windows starts up that is interfering with HotS this should help find that.

another thing to check is cabling. though that may sound obvious, make sure everything is completely plugged in firmly and screwed down. loose connections cause problems. – additionally if you are using HDMI, try a DVI connection if you are able to.