Jump to content
AIDA64 Discussion Forum
Cat_in_window

Troubles with GPU stability test

Recommended Posts

Theres something wrong about my system (whih is Windows 7 SP1) i guess.

When I just start the stability test of any GPUs, no matter discrete or built-in, i got my appropriate videodriver crash. I'm assuming that the problem may be caused by the conflict of the drivers, but still theres no same problem if i use any other GPUs testing app such as FurMark or Video Memory stress Test.

I want to know, what errors of my system can cause this crashes?

Share this post


Link to post
Share on other sites

It means the OpenCL layer of your video card driver crashes when AIDA64 tries to create an OpenCL compute context.  Try to update the video drivers, it should help to fix this up.

Regards,
Fiery

Share this post


Link to post
Share on other sites
29 minutes ago, Fiery said:

It means the OpenCL layer of your video card driver crashes when AIDA64 tries to create an OpenCL compute context.  Try to update the video drivers, it should help to fix this up.

Regards,
Fiery

I did update my GTX 660 driver to last version after total wipe (deleted all components including .dll) of previos version but this not solved my problem.

Maybe theres might be something else with corrupted .dll files? 

Share this post


Link to post
Share on other sites

It certainly can not be a videodriver due to the fact that an attempt to test Intel HD built-in videocard couses the same instant crash :\

I was tried to check my OC memory moduls on the errors, and also reduced their frequency to default values, but this also had no effect.

Share this post


Link to post
Share on other sites

If it crashes with both Intel and nVIDIA GPU, then it can be the OpenCL platform that's got corrupted somehow.  Maybe by uninstalling both video drivers and reinstalling them would help.  I'm not sure if there's an OpenCL driver scrapping/repair tool -- that certainly would help in this particular case...

Share this post


Link to post
Share on other sites
10 hours ago, Fiery said:

If it crashes with both Intel and nVIDIA GPU, then it can be the OpenCL platform that's got corrupted somehow.  Maybe by uninstalling both video drivers and reinstalling them would help.  I'm not sure if there's an OpenCL driver scrapping/repair tool -- that certainly would help in this particular case...

Iam just delete the OpenCL.dll and reinstall it via nVIDIA driver update, but Problem stll here :C

Also sfc.exe never find any errors after 2 launchs.

Share this post


Link to post
Share on other sites

Just by removing OpenCL.dll you cannot reset the whole OpenCL stack. The actual OpenCL driver is not in the file OpenCL.dll.  That file is just a platform passthrough module that calls the other OpenCL platform drivers made by AMD, Intel or nVIDIA.

Share this post


Link to post
Share on other sites
1 hour ago, Fiery said:

Just by removing OpenCL.dll you cannot reset the whole OpenCL stack. The actual OpenCL driver is not in the file OpenCL.dll.  That file is just a platform passthrough module that calls the other OpenCL platform drivers made by AMD, Intel or nVIDIA.

Then it means that i can't update or find the troubles by scanning this thing and the only way to refresh it - reinstall the OS system, right?

I have no idea which file is call for other files and by that i can't even replace them manually..

Share this post


Link to post
Share on other sites
2 hours ago, Cat_in_window said:

Then it means that i can't update or find the troubles by scanning this thing and the only way to refresh it - reinstall the OS system, right?

I have no idea which file is call for other files and by that i can't even replace them manually..

I'm not sure if you need to be as drastic as to reinstall the whole OS.  But, it should help if you could completely remove both video drivers, find the latest version of both, and reinstall them.

As for OpenCL, after removing both video drivers and if there's a remaining OpenCL.dll file, that file too... It may help to remove the traces of OpenCL from the Registry, by removing the following Registry entries and its sub-entries:

HKEY_LOCAL_MACHINE\SOFTWARE\Khronos\OpenCL

HKEY_LOCAL_MACHINE\SOFTWARE\WOW6432Node\Khronos\OpenCL

 

  • Like 1

Share this post


Link to post
Share on other sites
13 hours ago, Fiery said:

I'm not sure if you need to be as drastic as to reinstall the whole OS.  But, it should help if you could completely remove both video drivers, find the latest version of both, and reinstall them.

As for OpenCL, after removing both video drivers and if there's a remaining OpenCL.dll file, that file too... It may help to remove the traces of OpenCL from the Registry, by removing the following Registry entries and its sub-entries:

HKEY_LOCAL_MACHINE\SOFTWARE\Khronos\OpenCL

HKEY_LOCAL_MACHINE\SOFTWARE\WOW6432Node\Khronos\OpenCL

 

I was tired of all this and just reinstall the system one day. The problem solved now, but my videodriver still crashes in the games, so, i guess, it's just videocard death has came :(

Since it's not my DROM, which i was replaced by other module, it must by only videocard. Now i want to find same model of VC and test it.

Thanks for all of your help then. I wish all TechSupps work as good as you did here.

Also, i found that AIDA is an excellent choice for comlex tests like this. So from now on, I'am gonna advise it to more peoples for similar purposes.

Share this post


Link to post
Share on other sites

Similar situation here. Immediately after I start a GPU stability test, my video driver crashes with Event ID 4101 "Display driver nvlddmkm stopped responding and has successfully recovered". My video card is a GeForce GTX 1060 6GB (Pascal) on Windows 7 64 Bit, and I get the same result with driver 378.78 and 372.90 (did a complete uninstall with DDU prior to installing the older driver). Running a GPGPU benchmark works fine, I can loop 3DMark and play games like Rise of the Tomb Raider all day long, my rig is ancient (Core 2 Quad 3.0 GHz on Asus P5E with 8GB DDR2-1066) but rock solid, not overclocked, low CPU (24 degC idle) and GPU temps (34 degC @ 0 RPM idle) and this is the first time I experienced a video driver crash since I installed the GTX 1060 approximately two months ago.

I looked through the "Manage 3D Settings" section in the NVIDIA Control Panel to see if there are any OpenCL-related options. There is a setting "CUDA - GPUs" that I tried both with value "All" and "None" (by unticking the GPU) but doesn't make any difference. Please let me know if you can think of anything else I could try and in case you guys happen to have a GeForce GTX 1060 6GB laying around, please confirm that the GPU stability test definitely works for you under Windows 7 64 Bit (and which driver you are using).

Thank you.

 

Edit: For good measure, I downloaded the LuxMark 3.1 OpenCL benchmark/stability test (http://www.luxmark.info) and all 3 models render without error. At this point I am convinced the AIDA64 GPU stability test needs a patch to run on GTX 1060 under Windows 7 64 Bit, at least for recent drivers including 378.78 and 372.90. Fiery, would be good if you could look into this and please let me know if I can provide any other information.

Edited by Buhntz
Added additional information

Share this post


Link to post
Share on other sites
4 hours ago, Buhntz said:

Similar situation here. Immediately after I start a GPU stability test, my video driver crashes with Event ID 4101 "Display driver nvlddmkm stopped responding and has successfully recovered". My video card is a GeForce GTX 1060 6GB (Pascal) on Windows 7 64 Bit, and I get the same result with driver 378.78 and 372.90 (did a complete uninstall with DDU prior to installing the older driver). Running a GPGPU benchmark works fine, I can loop 3DMark and play games like Rise of the Tomb Raider all day long, my rig is ancient (Core 2 Quad 3.0 GHz on Asus P5E with 8GB DDR2-1066) but rock solid, not overclocked, low CPU (24 degC idle) and GPU temps (34 degC @ 0 RPM idle) and this is the first time I experienced a video driver crash since I installed the GTX 1060 approximately two months ago.

I looked through the "Manage 3D Settings" section in the NVIDIA Control Panel to see if there are any OpenCL-related options. There is a setting "CUDA - GPUs" that I tried both with value "All" and "None" (by unticking the GPU) but doesn't make any difference. Please let me know if you can think of anything else I could try and in case you guys happen to have a GeForce GTX 1060 6GB laying around, please confirm that the GPU stability test definitely works for you under Windows 7 64 Bit (and which driver you are using).

Thank you.

 

I got GTX 660 with last driver version (378.92), but i had this problem in AIDA until i reinstall whole system to the Windows 10. This was the only way i could solve this issue, so it might be just AIDAs incompatibility.

As i see now, crashes in games isn't connected to crashes during AIDAs GPU tests.

Share this post


Link to post
Share on other sites

I am still experiencing the same issue in the new AIDA64 version 5.90.

Fiery, I'd appreciate if you can run a quick GPU stability test with a ***GTX 1060 6GB*** (no other internal or second display adapter) on ***Windows 7 64-bit*** and report your findings.

Thank you.

 

Share this post


Link to post
Share on other sites
On ‎2017‎. ‎03‎. ‎28‎. at 4:23 PM, Buhntz said:

I am still experiencing the same issue in the new AIDA64 version 5.90.

Fiery, I'd appreciate if you can run a quick GPU stability test with a ***GTX 1060 6GB*** (no other internal or second display adapter) on ***Windows 7 64-bit*** and report your findings.

Thank you.

We've tested AIDA64 v5.90 on two Windows 7 64-bit SP1 based systems.  One of them had a Gigabyte GV-N1060WF2OC-6GD video adapter, while the other had a Asus GT610-SL-1GD3 video adapter.  They both used ForceWare 378.92 WHQL drivers.  Both system ran the GPU subtest of the AIDA64 System Stability Test properly, and they also ran the GPGPU Benchmark suite properly.  We've seen no crashes or other issues.

Share this post


Link to post
Share on other sites

Fiery,

Thank you very much for looking into this. The specs of your Gigabyte GV-N1060WF2OC-6GD are very close to my Asus DUAL-GTX1060-O6G card, so it is quite bizarre that the GPU stability test always causes the display driver to crash with Event ID 4101 (screen goes blank for a few seconds, then comes back with message that the display driver successfully recovered, but AIDA64 thinks the test is still running and will not let me close the stress test window), yet the GPGPU benchmark within the same application, the LuxMark 3.1 OpenCL benchmark/stress test, 3DMark. GPU-Z, graphic-intensive games like RotTR, etc. run without problem and I never experience driver crashes elsewhere. Other things I tried since my last post:

  1. Uninstalled ForceWare driver through Programs and Features in Control Panel, reboot, 2nd cleanup pass with DDU in safe mode, reboot, manually removed one leftover Khronos key incl. subkeys from registry per your guidance for Cat_in_window above, installed ForceWare 378.92 => no change
  2. Uninstalled AIDA64, reboot, manually removed all remaining folders/files under C:\Program Files (x86)\FinalWire\, fresh reinstall of latest version, re-entered product key but did not touch any AIDA64 settings (so everything at factory default) => no change
  3. Flashed GTX 1060 video adapter with backup of original BIOS (I previously flashed the card with a BIOS update from Asus that enables 0 RPM feature during low GPU utilization/temperature) => no change

At this point I pretty much reached the end of my wisdom. If you have any other ideas/suggestions/debug aids or know of a code difference between the GPU stability test and the GPGPU benchmark that could explain why on the same hardware/environment one always works and the other one never, please let me know. Again, this is an ancient Asus P5E motherboard with Intel X38 chipset, PCIe 2.0 x16 interface (the video  card is PCIe 3.0), 3GHz Core 2 Quad and 8 GB DDR2 1066, but since the rig runs everything except for the AIDA64 GPU stability test I don't think age of the MB is the problem.

As it turns out, my issues were rooted in the timeout detection and recovery (TDR) feature of WIndows 7 and the AIDA64 GPU stability test happens to be the only application on my machine that exceeds the 2 second response limit which causes Windows to reset the display driver. I added a new DWORD "TdrDelay" under HKLM\SYSTEM\CurrentControlSet\Control\GraphicsDrivers with an arbitrary value of 8 (seconds), rebooted and problem solved! GPU stability test runs just fine now and heats the card from 36 deg C to a maximum of around 70 deg C after a few minutes. I'm all set.

Thank you again.

 

Share this post


Link to post
Share on other sites
10 hours ago, Buhntz said:

Fiery,

Thank you very much for looking into this. As it turns out, my issues were rooted in the timeout detection and recovery (TDR) feature of WIndows 7 and the AIDA64 GPU stability test happens to be the only application on my machine that exceeds the 2 second response limit which causes Windows to reset the display driver. I added a new DWORD "TdrDelay" under HKLM\SYSTEM\CurrentControlSet\Control\GraphicsDrivers with an arbitrary value of 8 (seconds), rebooted and problem solved! GPU stability test runs just fine now and heats the card from 36 deg C to a maximum of around 70 deg C after a few minutes. I'm all set.

Thank you again.

Thank you for letting us know about the resolution. I've checked, and both our Win7 64-bit test systems had the TdrDelay value set to 30 seconds. So that explains everything :) We'll display a warning message when the value is set below 8 seconds in the Registry, and offer a quick fix.

Share this post


Link to post
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

Loading...

×
×
  • Create New...