BSODs on GPU jobs

[Vester]

I have had three BSODs on GPU jobs in the last two days related to fah.exe. I have re-imaged my HD to one week ago, removed F@H, and tested my computer and found no problems on my end. I am running an HPE-h9-1135 Phoenix with AMD FX-8350 processor, AMD HD 7790 video card, 16 GB RAM, and Windows 7 Home Premium 64b. I am reinstalling F@h in a few minutes and see how it goes.

[edit]Driver Packaging Version 14.10.1006-140417a-171099C
Catalyst Version 14.4
Provider Advanced Micro Devices, Inc.
2D Driver Version 8.01.01.1390
Direct3D Version 9.14.10.01029
OpenGL Version 6.14.10.12874
AMD Catalyst Control Center Version 2014.0417.2226.38446
AMD Audio Driver Version 7.12.0.7718[/edit]

[JimF]

My experience with BSODs usually implicates hardware problems, which can include over-heating. Are you overclocking either the GPU or the CPU? Warmer weather could bring out latent problems. Less likely, but also possible, are memory errors (either in the DDR3 main memory, or the video card memory).

[bruce]

(Even without overclocking,) when you added the GPU card, did you do anything to make sure that the fans would be able handle the increased heat without an increase in internal temperatures?

[PantherX]

I do hope that your PSU is able to power the system once loaded. BTW, have you considered running other stability benchmarks to see if this is a F@H only issue or a stability issue in general? Finally, did you choose the updated version of 14.4 WHQL drivers or the initial release?

[Vester]

I have an 800 watt power supply. This CPU has never been overclocked. CPU temp (liquid cooler) is 52C and GPU temp is 60C. I run without the side cover. This morning I uninstalled the video drivers and Catalyst Control Center and reinstalled using a new download of 14-4-win7-win8-win8.1-64-dd-ccc-whql.exe. It has been running satisfactorily since. This is an error I found in Event Viewer | Applications:

Faulting application name: FahCore_a3.exe, version: 0.0.0.0, time stamp: 0x4d4720af
Faulting module name: FahCore_a3.exe, version: 0.0.0.0, time stamp: 0x4d4720af
Exception code: 0xc0000005
Fault offset: 0x00261bf6
Faulting process id: 0xb2c
Faulting application start time: 0x01cf8db6bf51d914
Faulting application path: C:\Users\Vester\AppData\Roaming\FAHClient\cores\web.stanford.edu\~pande\Win32\AMD64\Core_a3.fah\FahCore_a3.exe
Faulting module path: C:\Users\Vester\AppData\Roaming\FAHClient\cores\web.stanford.edu\~pande\Win32\AMD64\Core_a3.fah\FahCore_a3.exe
Report Id: ca1dcce0-fa2b-11e3-a5d1-9cb70d9c72c4

The message could be related to the recovery using a six day old image of the drive.

After the BSOD early today, I restarted (before computer testing/reinstalling) and had a popup when the desktop loaded that said Fah had faulted. We'll see how it goes. I was a moderator on another project and I often said, "The client won't break your computer but it will reveal computer problems." I've been sorting this computer for 20 months. Thanks.

[JimF]

I don't run CPU jobs, but that error could be a result of the crash rather than the cause. I practically never see errors in the GPU jobs and don't even think of them as a possible cause, though it could happen of course. Let us know if you discover any hardware problems (not due to temperature though!).

[P5-133XL]

The 0xc00000005 is a conventional memory access error and the FAHCore_A3 is a CPU folding core so I'm not sure why the BSOD would have anything to do with the GPU unless we are dealing with power issues where the video card is pulling so much current that it is dropping the voltage for the CPU/MB/RAM. May I suggest that you check your PS rails to make sure that the video card is on a separate rail from everything else.

[bruce]

I have had three BSODs on GPU jobs in the last two days related to fah.exe. I have re-imaged my HD to one week ago, removed F@H, and tested my computer and found no problems on my end.

You do realize that there's a false assumption there. You didn't test the computer under identical conditions. Put simply, a computer running FAH is drawing more power and generating more heat and undergoing more double-checking of results than a computer that's not running FAH.

[davidcoton]

Why do you think the BSODs were related to GPU jobs? The one Win log extract relates to a CPU job (Core_A3). Do you have any other error logs?

Memory access errors are often the result of failing RAM, or sometimes the motherboard circuitry. Run a memory check, preferably for several hours. Check memory is correctly seated. If errors persist, remove or swap memory. I have seen machines fail with two identical sticks, but work with either one alone. Otherwise, check power distribution and voltages on load as P5-133XL suggests above. Could also be a local hotspot near the memory, but this is unlikely with the temp you noted.

However, if the machine is now working properly after driver update, and if that was the only change since faults were appearing, there could have been a problem in the previous driver version that FAH revealed. I don't think the Win log records enough information to determine the actual faulty software module -- at least not in a form mere mortals can interpret.

[Vester]

OK, accept this statement: "This computer crashed three times while crunching seven CPU jobs and one GPU job. I am done. I don't have time for it anymore. I have a new boat. Bye."

[DeeGee]

Although this is bit late, all of my BSODs have happened on 14.xx series drivers. Even the newest 14.6's cause BSOD together with folding and firefox/flash. 13.12 drivers have been bluescreen free for me, so I'll probably will be going back to them.

[JimF]

I have been Folding on two HD 7790s on the same motherboard with the 14.4 drivers (Win7 64-bit) since they came out with no problems. A couple of days ago I switched the cards from a Haswell machine to an Ivy Bridge machine. These are the Power Color cards, and they run hotter than most cards (typically 70 to 80 C), but with a side case fan on them that has not been a problem thus far.