GPU Folding Error

[DefiDJ]

Hi,

when trying to fold with my GPU (automatically chosen and executed by the client) an Error occurs:

UNKNOWN_ENUM (-1073740940 = 0xc0000374)

while the CPU-folding thread works just fine.

My specs are:

Computer Specs:
CPU: Intel(R) Core(TM) i5-3330 CPU @ 3.00GHz
CPU ID: GenuineIntel Family 6 Model 58 Stepping 9
CPUs: 4
Memory: 7.90GiB
GPU 0: Bus:2 Slot:0 Func:0 NVIDIA:3 GK104 [GeForce GTX 670]

Network Connection:
Cable
Firewalls: Avast and Commodo (exceptions set)

Operating System:
Windows 7 64bit

Overclocked?:
GPU is factory overclocked but i doubt it's much

Software:
Version of the client: 7.5.1
Version of Core: 21

Is my GPU unfit for Folding or is it something with the doubious overclocking?
Thanks

[Joe_H]

Welcome to the folding support forum.

You give too little of the error to place it in context, but that error could be a Windows error. If not, the error could be related to the video drivers you have installed. Folding requires a recent version of the nVidia drivers with the OpenCL support included. The drivers supplied through Windows Update by MS usually do not include the OpenCL components.

[JimboPalmer]

Both Kepler GPUs (the K in GK104 is for Kepler) and Win 7 64 bit are supported well. https://en.wikipedia.org/wiki/Kepler_(m ... hitecture)

I would encourage you to get a 391 version driver from Nvidia. https://www.geforce.com/drivers/results/132838

It will include other software which you may want for other uses, all you need for F@H is the video driver from Nvidia.

if that is the driver you have, then I would suspect an over clock.

[DefiDJ]

Thanks for the quick answer. The driver was a bit outdated, but the problem still remains.
The Error occurs everytime, when I try to start a WU with the GPU.
I tried to downclock the GPU with the MSI Afterburner, but that does not seem to affect the situation at the moment.
If not paused, after a delay the error repeats.
As I'm new to this, I don't know which logs are relevant.

The log from Unit 01 reads:

23:10:36:WU01:FS01:0x21:*********************** Log Started 2018-06-18T23:10:36Z ***********************
23:10:36:WU01:FS01:0x21:Project: 11713 (Run 14, Clone 142, Gen 154)
23:10:36:WU01:FS01:0x21:Unit: 0x000000c58ca304e75a5cfe1f5d9298cb
23:10:36:WU01:FS01:0x21:CPU: 0x00000000000000000000000000000000
23:10:36:WU01:FS01:0x21:Machine: 1
23:10:36:WU01:FS01:0x21:Reading tar file core.xml
23:10:36:WU01:FS01:0x21:Reading tar file integrator.xml
23:10:36:WU01:FS01:0x21:Reading tar file state.xml
23:10:36:WU01:FS01:0x21:Reading tar file system.xml
23:10:36:WU01:FS01:0x21:Digital signatures verified
23:10:36:WU01:FS01:0x21:Folding@home GPU Core21 Folding@home Core
23:10:36:WU01:FS01:0x21:Version 0.0.18
23:10:40:WARNING:WU01:FS01:FahCore returned an unknown error code which probably indicates that it crashed
23:10:40:WARNING:WU01:FS01:FahCore returned: UNKNOWN_ENUM (-1073740940 = 0xc0000374)

The Windows error reads:

Problem Event Name: APPCRASH
Application Name: FahCore_21.exe
Application Version: 0.0.0.0
Application Timestamp: 588257cc
Fault Module Name: StackHash_c9a4
Fault Module Version: 6.1.7601.24024
Fault Module Timestamp: 5a58e571
Exception Code: c0000374
Exception Offset: 00000000000bf6b2
OS Version: 6.1.7601.2.1.0.256.1
Locale ID: 1031
Additional Information 1: c9a4
Additional Information 2: c9a4bdbd5228c2041c8c9b401d8ca542
Additional Information 3: 5f9f
Additional Information 4: 5f9fbe9b9b86b10e2c728a934b8443cb

[bruce]

Factory Overclocks are GENERALLY safe, but it's impossible to be sure.

Have you applied all the Win7-64 updates?

[DefiDJ]

All Windows Updates installed. ok, so the GPU is returned to standard settings.

Now I noticed sth new: when starting the client, the first run results also in an appcrash error, but the second time works o.O

This may be unimportant, but the Webcontrol client seems to continually reload (thus failing to login), which does not affect the Advanced Control Client - which is working fine except the original problem.

Also I noticed, that the index of the GPU slot has the ID "01", but the allocated WU is labeled with "02". Whereas the IDs of the CPU slot and -WU match.

Yesterday I also tried deleting the local client and creating a new one, but with no success.

[bruce]

Now I noticed sth new: when starting the client, the first run results also in an appcrash error, but the second time works o.O

Is your system overheating? Does your RAM pass extended memory diagnostics?

This may be unimportant, but the Webcontrol client seems to continually reload (thus failing to login), which does not affect the Advanced Control Client - which is working fine except the original problem.

I don't understand. Both WebControl and FAHControl should be monitoring a single instance of FAHClient and should see the same WU running in each slot. Are you thinking of the NaCl client in Chrome?

Also I noticed, that the index of the GPU slot has the ID "01", but the allocated WU is labeled with "02". Whereas the IDs of the CPU slot and -WU match.

The queue number of the active WU in each slot is not related to the slot number. Each WU receives the number of the next available queue position at the moment it is assigned. This is NOT associated with the Slot number.

[DefiDJ]

Is the Windows Memory Diag sufficient? If so, there were no problems found.

Are you thinking of the NaCl client in Chrome?

Yes. When first started, both the desktop app and the web app are opened.
Also, the initial appcrash is now inconsistent.
Since the last two reboots, I'm unable to connect to the webclient. May this be partially relevant?

If I read the log correctly, the connection is not the problem, as the project has a number

23:10:36:WU01:FS01:0x21:*********************** Log Started 2018-06-18T23:10:36Z ***********************
23:10:36:WU01:FS01:0x21:Project: 11713 (Run 14, Clone 142, Gen 154)

I have removed the GPU slot and readded it. Further, I think the WU fetching is no problem, because at first the project was "unkown" then the entries were updated, and at that time the error occured.
If the slot is stopped and then started again, it fails immediately.

[JimboPalmer]

If you have updated Microsoft Updates, you need to reapply Nvidia Drivers, the MS drivers cause failure.

[DefiDJ]

Thanks for the hint. Done and now back to crashing at every 1st time after bootup, but the 2nd time it works just fine o.O
I can connect to the WebControl, but it still jumps between three different frames around. I captured two of them, but i doubt that this will be of any help.
https://imgur.com/0mdLdc5
https://imgur.com/CEsRq8q

Anyway, if I can't solve this in the near future, the CPU will have to suffice

[bollix47]

Okay let's try to get folding itself working and forget about Web Control for now.
First of all use FAHControl to Pause all folding.
If you told the installer to start Folding automatically then there should be a shortcut in your startup folder and possibly one on your Desktop too. You need to edit whichever one that's being used to start folding.
"%appdata%\Microsoft\Windows\Start Menu\Programs\Startup"
Right click on the folding shortcut, select properties and remove everything after FahClient.exe" which should be --open-web-control then click on Apply & OK.
Now reboot your system. Web Control should not start. Folding should start in a paused state if you told it to start automatically. Now open FAHControl, right click on the slot for your GPU and select Fold. Don't start your CPU slot for now. We just want to concentrate on GPU until later.

[DefiDJ]

Thanks, I did all of that. It worked up to the point, when the GPU slot is activated. The old problem still remains: a Windows appcrash error.
Some points are awarded, but as I understand, this is only an estimate. May be nothing.

[bruce]

This error is a stack/heap corruption error. The most probable source of the probelm has something to do with the drivers. Although you've apparently already visited this issue, I would still recommend reinstalling the latest drivers from NVidia. Since all programs share access to the heap, it can be very difficult to diagnose. There are many possibilities including something simple like using the OpenCL code from Microsoft instead of the OpenCL code from nVidia.