HD4870 crashes, hardware or driver problem?

[quadra]

Hello,
I'm running win 7 HP x64, a AMD HD4870 512mb. Sometimes my screen goes blank, the fan on the gfx goes up to 100% and the red warning LED on the PCB of the gfx lights up. This has happened a couple of times, most recently it happened and I was (as always) forced to do a hard-reset, but then it crashed again just before the win7 log-in screen appeared. Another reboot and I'm up and running for now.
Now, before the two last crashes I updated AMD Catalyst to the latest version, v.13.1. This obviously did not fix the problem.

Here is what I found from the minidump, using WinDgb:

*******************************************************************************
* *
* Bugcheck Analysis *
* *
*******************************************************************************
VIDEO_TDR_FAILURE (116)
Attempt to reset the display driver and recover from timeout failed.
Arguments:
Arg1: fffffa800ae264e0, Optional pointer to internal TDR recovery context (TDR_RECOVERY_CONTEXT).
Arg2: fffff88003abde6c, The pointer into responsible device driver module (e.g. owner tag).
Arg3: ffffffffc0000001, Optional error code (NTSTATUS) of the last failed operation.
Arg4: 0000000000000003, Optional internal context dependent data.
Debugging Details:
------------------

FAULTING_IP:
atikmpag+8e6c
fffff880`03abde6c 4055 push rbp
DEFAULT_BUCKET_ID: GRAPHICS_DRIVER_TDR_FAULT
CUSTOMER_CRASH_COUNT: 1
BUGCHECK_STR: 0x116
PROCESS_NAME: System
CURRENT_IRQL: 0
STACK_TEXT:
fffff880`05d60a18 fffff880`0569b000 : 00000000`00000116 fffffa80`0ae264e0 fffff880`03abde6c ffffffff`c0000001 : nt!KeBugCheckEx
fffff880`05d60a20 fffff880`0566e569 : fffff880`03abde6c fffffa80`08dc2000 00000000`00000000 fffffa80`0825a010 : dxgkrnl!TdrBugcheckOnTimeout+0xec
fffff880`05d60a60 fffff880`0569adad : fffffa80`0ae264e0 fffffa80`0ae264e0 fffffa80`0ae264e0 fffffa80`0825a010 : dxgkrnl!DXGADAPTER:repareToReset+0x109
fffff880`05d60aa0 fffff880`05741f07 : fffffa80`0ae264e0 00000000`00000000 fffffa80`09f36af0 fffffa80`0825a010 : dxgkrnl!TdrIsRecoveryRequired+0x245
fffff880`05d60ad0 fffff880`0576fd5a : fffffa80`ffffffff 00000000`00a8af5e fffff880`05d60c30 00000000`00000002 : dxgmms1!VidSchiReportHwHang+0x40b
fffff880`05d60bb0 fffff880`0573d29e : fffffa80`0825a010 ffffffff`feced300 fffffa80`07321010 00000000`00000000 : dxgmms1!VidSchWaitForCompletionEvent+0x196
fffff880`05d60bf0 fffff880`05769e7a : 00000000`00000000 fffffa80`07321010 00000000`00000080 fffffa80`0825a010 : dxgmms1!VidSchiScheduleCommandToRun+0x1b2
fffff880`05d60d00 fffff800`0317a34a : 00000000`05650233 fffffa80`08caeb50 fffffa80`069a5840 fffffa80`08caeb50 : dxgmms1!VidSchiWorkerThread+0xba
fffff880`05d60d40 fffff800`02eca946 : fffff880`009e9180 fffffa80`08caeb50 fffff880`009f3fc0 fffff880`05d60e40 : nt!PspSystemThreadStartup+0x5a
fffff880`05d60d80 00000000`00000000 : 00000000`00000000 00000000`00000000 00000000`00000000 00000000`00000000 : nt!KiStartSystemThread+0x16

STACK_COMMAND: .bugcheck ; kb
FOLLOWUP_IP:
atikmpag+8e6c
fffff880`03abde6c 4055 push rbp
SYMBOL_NAME: atikmpag+8e6c
FOLLOWUP_NAME: MachineOwner
MODULE_NAME: atikmpag
IMAGE_NAME: atikmpag.sys
DEBUG_FLR_IMAGE_TIMESTAMP: 50a69661
FAILURE_BUCKET_ID: X64_0x116_IMAGE_atikmpag.sys
BUCKET_ID: X64_0x116_IMAGE_atikmpag.sys
Followup: MachineOwner
---------

 

[denis280]

it is the driver.sometime it is better to get the beta.

 

[quadra]

I manually removed all traces of the AMD driver, as per this guide: http://www.overclock.net/t/988215/h...l-ati-amd-software-drivers-for-graphics-cards

I then reinstalled them, however I did get this error in the installation log:

AMD APP SDK RuntimeFinal Status: Fail
Version of Item: 10.0.937.2
Size: 90 Mbytes

I have gotten this error several times before when installing the AMD drivers, but have not given it much effort. Someone pointed out that crashdump in my original post showed a lot of references to DirectX. I did launch DxDiag, but it did not report any errors.
I also ran FurMark with the standard 15 min test both with fan on auto, and manually set to 28% (to raise temperature to a max of 102 degrees celcius). Both runs completed without error.

I have yet to experience a crash again, but seeing how it varies so much how long it can take, I am unable to tell if the reinstallation of the drivers has fixed anything. I would however like to try and fix the error with the installation. Any help with that?

 

[denis280]

What are all your system spec.

 

[quadra]

Sorry, I should probably have posted the complete specs in the original post (didn't think it could be necessary):

MB: MSI G41M-P28
RAM: Corsair XMS3 DDR3 1333MHz 8GB CL9 (2x 4GB)
CPU: Intel Q6600
GFX: HIS? Radeon HD4870 512mb
PSU: Corsair 450watts (bought jan'2009)
OS HDD: WD Black 2tb
OS: Windows 7 Home Premium x64 (upgrade, but "clean install" by installing and then reinstalling)

 

[denis280]

The Radeon HD4870 512mb pull a good 193 watt idle.I would try swap psu with a 550 or 600 watt.sure wont hurt.

 

[quadra]

The Radeon HD4870 512mb pull a good 193 watt idle.I would try swap psu with a 550 or 600 watt.sure wont hurt.

No offence, but that does not sound like a good advice. If I was to spend money swapping anything, a newer, much more efficient (and better performing) gfx would make more sense? Also, if the PSU was the cause of the crashes, I find it odd that the crashes always has been happening when the gfx is pretty much ideling and that I did not ran into any problems when "torture testing" it (which puts a much greater load on the psu).

Edit:
I suppose you found the idle power consumption here: http://www.techspot.com/review/108-visiontek-radeon-4870/page11.html
193 watts is not the idle consumption of the gfx card, but the whole system...which is also running an Q6600 cpu oc'ed to 3Ghz (mine is stock).

 

[BadOmen]

Crashing and then crashing again right after boot... Have you tried placing a house fan facing your MB to see if it stops? Catalyst's PWM controls have left me in the lurch before with heat-related crashes. You may also try to increase speed (and lock it) in Catalyst, but as the problem happened while idling, perhaps the heat source (if heat is indeed the problem) is not the gpu itself.

 

[quadra]

Here is an update of the current situation:

Last crash I experienced, I looked up which LED that was lighting up (was not aware that there were 4 different ones). The code showed the following error:

D1601 - Red LED On, shows critical temperature fault
D1602 - Red LED On, shows External power connector A was removed
D1603 - Red LED On, shows External power connector B was removed
D601 - Red LED On, shows critical Core power fault

I have checked the connection and also deliberately moved them while the computer was switched on, thinking it would provoke a crash. However, the connection seemed to be good and I was not able to reproduce a crash this way. I therefore tried measuring the voltage across the two external power connectors:

"Upper connector":
Idle max: 12.18v
Load min: 12.06v

"Lower connector":
Idle max: 12.13v
Load min: 11.95v

These are well within ATX v.2.2 specs, loads were performed during FurMark stresstesting. Unfortunently I only became aware of the different fault codes recently, so I can not verify that the same code has been triggered every time I've experienced a crash. Thus I have decided to allow the computer to experience one or preferably two more crashes so I can check if the same error code shows up every time. If so, the next logical step would be to switch them around and see if the following crashes will trigger a D1602 fault code instead. If so, I'll have to consider PSU problems. If the code stays the same, I will assume it is an error with the gfx card. If it appears to be the card, I'll try to flash it with a new BIOS to see if that solves it. So for now I'll just have to wait some.

Sorry for the long post, I'll post an update when I have gathered more information. Thanks for all help so far!