All my GPU WU finishes "computation error" after a few seconds !!!

\n studio-striking\n

Message boards : Number crunching : All my GPU WU finishes "computation error" after a few seconds !!!
Message board moderation

To post messages, you must log in.

AuthorMessage
Profile marsinph

Send message
Joined: 25 Feb 18
Posts: 6
Credit: 1,717,092,450
RAC: 0
Message 7992 - Posted: 10 May 2018, 14:53:33 UTC

Look host 992563.
Till 19 april all perfect.
No any changes since then. C++ again reinstalled (x64 and x86)
Now all crashes after 11-12 seconds ! So I needed to cancel the others.
Who have idea ?
ID: 7992 · Rating: 0 · rate: Rate + / Rate - Report as offensive
Profile bcavnaugh
Avatar

Send message
Joined: 25 Dec 15
Posts: 7
Credit: 556,523,753
RAC: 0
Message 7993 - Posted: 11 May 2018, 1:01:07 UTC

Seeing this on some Hosts as well ATM

Crunching@EVGA The Number One Team in the BOINC Community. Folding@EVGA The Number One Team in the Folding@Home Community.
ID: 7993 · Rating: 0 · rate: Rate + / Rate - Report as offensive
Profile bill1024

Send message
Joined: 24 Apr 17
Posts: 1
Credit: 167,780,016
RAC: 1
Message 7994 - Posted: 11 May 2018, 1:07:39 UTC

Seeing the same thing, computational error after 12 seconds
I have C++ 8, 10, 13 and 15 both 86 and 64 installed.

I did DDU clean driver and reinstall driver and physx only. 390.65
Not sure what to do next.
ID: 7994 · Rating: 0 · rate: Rate + / Rate - Report as offensive
No_Name

Send message
Joined: 14 Nov 17
Posts: 11
Credit: 1,959,292,242
RAC: 1,233,956
Message 7998 - Posted: 23 May 2018, 8:14:30 UTC

The same problem here https://moowrap.net/forum_thread.php?id=527, but only with opencl.

With a new driver version, the problem ist still there.
ID: 7998 · Rating: 0 · rate: Rate + / Rate - Report as offensive
QuintLeo

Send message
Joined: 26 Jun 16
Posts: 53
Credit: 1,866,706,325
RAC: 0
Message 8002 - Posted: 1 Jun 2018, 17:56:22 UTC - in response to Message 7998.  

The core issue is that Moo Wrapper continues to use the old, outdated, and flat out buggy in certain ways 520 Dnet client instead of the current 521 client.
ID: 8002 · Rating: 0 · rate: Rate + / Rate - Report as offensive
No_Name

Send message
Joined: 14 Nov 17
Posts: 11
Credit: 1,959,292,242
RAC: 1,233,956
Message 8005 - Posted: 20 Jun 2018, 6:44:18 UTC

For my 1050 TI, the solution looked like this:

- if present, the Intel video card driver must be uninstalled
- under
HKLM\SOFTWARE\Wow6432Node\Khronos\OpenCL\Vendors
you have to delete all entries except the one from the manufacturer of your videocard

I hope i this will help you.
ID: 8005 · Rating: 0 · rate: Rate + / Rate - Report as offensive
Profile marsinph

Send message
Joined: 25 Feb 18
Posts: 6
Credit: 1,717,092,450
RAC: 0
Message 8007 - Posted: 4 Jul 2018, 14:48:23 UTC - in response to Message 8005.  

For my 1050 TI, the solution looked like this:

- if present, the Intel video card driver must be uninstalled
- under
HKLM\SOFTWARE\Wow6432Node\Khronos\OpenCL\Vendors
you have to delete all entries except the one from the manufacturer of your videocard

I hope i this will help you.



Hello,
I tryed this solution.
It seem to works !
Look on my hosts 992563. I renamed the line "Intel..." (in case of master crah)
On host 1012044 i also have the same line : "intel..."
Not on my two other hosts.

[HKEY_LOCAL_MACHINE\SOFTWARE\Wow6432Node\Khronos\OpenCL\Vendors]
"IntelOpenCL32.dll"=dword:00000000
"C:\\Windows\\SysWow64\\nvopencl.dll"=dword:00000000

It is to hope nothing else is affected In Win7 x64, also not to other projects !?
Wait and see....
ID: 8007 · Rating: 0 · rate: Rate + / Rate - Report as offensive
bluestang

Send message
Joined: 31 Mar 18
Posts: 7
Credit: 1,788,023,946
RAC: 7,628,428
Message 8011 - Posted: 13 Jul 2018, 2:28:01 UTC

Same thing happened to me after not running for a while. Error at 11 seconds on tasks. AMD GPU and the tasks from the application "Distributed.net Client v1.04 (ati14) windows_intelx86" is the culprit. Resetting project didn't work. Didn't have the Intel entry in my registry so it wasn't that either.

Only thing that worked was removing project, shutting down and restarting BOINC and adding project again. Then only "Distributed.net Client v1.04 (opencl_ati_101) windows_intelx86" tasks download and run properly. The "ati14" tasks are the bad ones.
ID: 8011 · Rating: 0 · rate: Rate + / Rate - Report as offensive
dduggan47

Send message
Joined: 17 May 11
Posts: 1
Credit: 25,912,412
RAC: 0
Message 8026 - Posted: 15 Oct 2018, 15:58:23 UTC - in response to Message 8011.  
Last modified: 15 Oct 2018, 16:00:37 UTC

Others who have posted in this thread are far more technical than I am but here's my experience.

These errors started happening 2 or 3 days ago on several projects, Moo, Collatz, SETI, Primegrid, ... anything that uses the NVIDIA GPU. I didn't get around to dealing with it until today.

Before looking at the message boards (many, probably all, of those other projects have similar conversations going on) I decided to try the easiest potential fix (i.e. the only one I knew how to do). I'm running Windows 10 Pro and I updated the NVIDIA driver.

It appears to have worked. Moo is the only project that's completed a GPU task since I updated but it was pretty clear right away that it would work once it had run for a full minute. Nothing else has completed but another Moo task has a couple of minutes under its belt and ditto for an Asteroids and a Collatz. Looks like all is well.

My GPU is GTX 745 and the driver is 24.21.13.9836.

- Dick

Edited twice because I've never learned to proof read before posting.
ID: 8026 · Rating: 0 · rate: Rate + / Rate - Report as offensive
Profile [AF>EDLS]GuL

Send message
Joined: 11 May 17
Posts: 2
Credit: 45,042,502
RAC: 221
Message 8031 - Posted: 2 Nov 2018, 1:01:25 UTC - in response to Message 8005.  

Hello,
This seems to be related to https://moowrap.net/forum_thread.php?id=510&postid=7846
I have a very similar problem : with Dnetc 1.04 and AMD application, it is impossible to compute on AMD GPU. The application is rather using the integrated GPU (Intel HD Graphics 4600 on a Core i7 4790K). I have tested separately with two GPU : a HD 7970 and a R9 Fury.

I am currently testing with a Nvidia GPU, and results are similar: only the Intel GPU is computing. In fact, when we start the client in command line, we obtain no presence about the nvidia gpu, at least in OpenCL mode. However, it is correctly detected in boinc, both as Cuda and Opencl device.
dnetc520-win32-x86-opencl.exe -cpuinfo --> Intel GPU only
dnetc520-win64-amd64.exe -cpuinfo --> Intel GPU only
dnetc518-win32-x86-cuda31.exe -cpuinfo --> Nvidia GPU only

The problem is the used distributed.net client itself that seems to be from 2012 and not even know the -gpuinfo option. Thus, it is detecting the first opencl platform (intel) and not the following ones (nvidia or amd). The afore mentioned solution is not suitable as it will perturb opencl detection on other projects. The best solution would be to update and recompile the dnetc clients. In the mean time, on Nvidia one can use an app_info.xml file forcing the use of the cuda client. On AMD, one need to remove intel opencl drivers, as proposed above.
For my 1050 TI, the solution looked like this:
- if present, the Intel video card driver must be uninstalled
- under
HKLM\SOFTWARE\Wow6432Node\Khronos\OpenCL\Vendors
you have to delete all entries except the one from the manufacturer of your videocard
I hope i this will help you.

ID: 8031 · Rating: 0 · rate: Rate + / Rate - Report as offensive

Message boards : Number crunching : All my GPU WU finishes "computation error" after a few seconds !!!


 
Copyright © 2011-2024 Moo! Wrapper Project