PC Switches off - Kernel-Power

[sgtbarton]

Could it be faulty kettle lead their end . Gets kicked etc .

Sent from my Desire HD using Tapatalk

[evilsatan]

I replaced the kettle lead when I replaced the PSU mate

[sgtbarton]

Hmm I would prob also run with 1 stick of ram n 1 Hdd no cd drive or any other power using items . Minimise causes . Also if u do get it to replicate try safe mode to remove drivers from possibly being the cause . Looking at past posts u prob already tried all that mind you .
Is there office small / hot etc . Bloody pcs , u should have got em dell lmao

Sent from my Desire HD using Tapatalk

[evilsatan]

Office is pretty small but not overly hot. I have checked the core temps remotely using speedfan and cores are always below 40 degrees. She only uses it for office applications, internet and email so the cooling and spec I have in place is overkill but they wanted a fast machine.

[pctuner1984]

Nothing is impossible, the APC should rule it out, and you should be able to monitor power through the powershute software to check for overload or spikes. I hate problems like this lol

[evilsatan]

I loaded that software on there and just got back. All setup with the APC so will monitor that!

[evilsatan]

Not heard back from the client yet, I called them today so I could check the powerschute logs but they were out. I will check these when they get back to me.

[evilsatan]

Jesus christ, just heard back and it is still playing up, if anything more frequently than ever.

Powershute has recorded no problems so it should not be a problem with their mains.

I am not sure if this is coincidence but it seems to occur when the machine is idle more than when under any 'load'. I had it running BurnInTest fir 21 hours on full CPU and GPU load with no issues. It shouldn't be a power setting as it always starts with 'Windows did not shut down correctly) and the same Kernel Power 41 (63) is in the event logs.

Troubleshooting to date:

Spoiler:

[No minidump file to analyse, set PC to not restart in case of system failure. Now an error screen may be displayed if the fault occurs again. System is set to Kernel memory dump.

Thermals checked, all cores running at same temperature as other PC, well below chip maximum.

PSU already replaced, sufficient overhead so not caused by faulty PSU or lack of power being supplied by PSU.

PC set to never sleep; screen will turn off after 10 minutes. Hard Drives will power down to save electricity after 20 mins idle.

Memtest found no issues.

Virus scans passed.

Machine disassembled and rebuilt from scratch, new thermal paste, all connections to mobo checked.

HDD formatted and software reloaded from scratch.

BurnInTest ran at 100% load for 21 hours in my office, no issues. Issues seem to occur when idle at client office.

Connected UPS to mains to test for power out, no faults detected in Powershute software, problem persists.

The only part of the PC I didn't bring to my office is the LCD, but I did use the same HDMI cable. I can only think it is the mobo, CPU or HDD at fault. Can't see the ODD, PCI WLAN card or memory card reader causing this.

I have insisted I take the client a loan PC tomorrow morning and I will have their machine on my bench over the weekend at least. I have the same mobo in my machine so I will swap them over to test.

Any more advice or comments on my troubleshooting to date and going forward would be very much appreciated, intermittent faults are so annoying because I can't replicate them to test the repair has been successful.

[muttleymacclad]

You seem to have done everything possible evil.

Could there be a fault with the CMOS battery? Long shot and not usually a cause of intermittent faults but about the only removable item you haven't checked.
Mml

[DejaVu]

Stop feeling yourself up (Touching Wood....! Tut, Tut!)

I'm guessing you've looked into this already?
http://support.microsoft.com/kb/2028504

This explains about the Microsoft-Windows-Kernel-Power getting shutting off without cleaning shuutting down first.

Other source also suggest the Realtek Audio Driver installation is corrupted.
Try uninstalling that completely and/or disabling it in the BIOS.
If you disable it and it stops it, either the driver is corrupt or the device is borked completely - Mobo replacement/BIOS Update (Check the details first, there maybe a fix).

Me personally, I would not have gone to the amount of hassle you have with this. I'd have replaced the mobo from the first instance - you maybe too late to send it back for a replacement now too, depending on where you got it from. eBuyer do not tend to check as much as they say they do.

[evilsatan]

Thanks for the replies. I have already tried disabling the audio driver and that didn't help After further reading the general consensus was that didn't help many people with this error. The strange thing is the machine is a dupe of another machine in the same office which has had absolutely no problems at all!

I haven't replaced the mobo yet as I wanted to make sure that's the fault, if it isn't then I get charged to get it back from Scan and would be out of pocket for another mobo as well as the customer being in the same boat they are in now. That said it has got to a point where replacing the mobo out of my own pocket is the sensible choice so I can test it in my machine and return if necessary to the retailer. That will also rule out the CMOS battery as muttley suggested but I think the machine has been holding time ok so that battery should be fine.

[evilsatan]

Found this thread, not finished reading it yet but have a list of possible problems:
http://social.technet.microsoft.com/Forums/en-US/w7itproperf/thread/9e71f600-7c62-4869-8236-964e93d17936/

Could be:

Audio drivers
Windows sidebar
RAM voltage
Other BIOS settings
Case (bad earthing)
HDD FW
Mobo
WLAN card
Drivers
Hibernation file
Windows Aero desktop
Faulty fan

So pretty much could be anything. I think I am going to have to build them a new PC from scratch out of my own pocket and then keep this one running 24/7 until I fix it and then sell it on. Problem is by then the warranty of some parts would have expired so the resulting machine would decrease in value. Not very happy about this but I need to provide the client with a machine which works! The only components I will reuse will be the disk drives, memory card reader, kb+mouse, CPU and PSU. New HDD, new case, new mobo, new WLAN card and new cables.

I have the machine on my workbench atm so if I somehow manage to get it to power off I will try just the mobo first but the above thread doesn't provide me with much confidence Windows 7 seems to be the only common factor, at least one user reports the problem with 32-bit W7 but I suppose almost everyone uses 64-bit now.

[muttleymacclad]

Come on Evil, you have to beat this b*stard.

mml

[evilsatan]

I would like to repair it but not at the expense of a client. I have now spent more time and money trying to repair this machine than I have earned from this client in total

I am leaving it on and idling now to try and cause it to power off, if I manage to do that then I will replace the mobo and test. There is a Edimax PCI WLAN card in there so I can remove that and test but I just want to replicate the fault before I try any of these steps.

It has been happening very regularly at the clients office, I will try to post event logs for reference.

I will also reflash the BIOS. Apparently some have fixed it by upping the RAM voltages and the FSB etc. but overclocking isn't my forte as I don't game at all on the PC.

[evilsatan]

Attached is a custom view of events to show all warnings, errors and criticals, as you can see there is nothing happening before the critical Kernel-power failures and the bugcheck codes are blank.

http://www.megaupload.com/?d=WNZCKEPB

[tombott]

Well the lesson to be learned here is don't build machines. Buy branded kit like HP or Dell. It comes with a warranty and they sort out any problems.
As I said before the first thing you should have done is put a loan PC in place, doesn't matter if the client was not keen on this you should insist.
As you have already stated this has now cost you more than you've billed, so put it down to experience and move on.

[evilsatan]

Yes tombott, lesson learned. I love building machines but from now on will reserve it for personal customers rather than businesses wherever possible. A member on here has been kind enough to PM me contact details of some Dell system suppliers.

The PC just restarted, it happened whilst left idle so it looks like it is a power setting causing it. I can't find reference to C-state in the BIOS. I am about to reflash the BIOS and leave it on the default settings to see if it makes any difference.

[sgtbarton]

Oh man . I keep coming back here . I need to know what s causing it . It's worse than a Friday eastenders cliff hanger !!!!

I cannot remember if you covered this but you did clear the cmos at some point right? ( not just popping battery out but resetting with jumper) I have had many builds need this so I just do it now .
Come on d00d you can solve this. It's just another lesson to learn you just gotta find your way to it .

Sent from my Desire HD using Tapatalk

[evilsatan]

Don't want to count my chickens yet but I think I have cracked it. I started by turning off all C-State technology in the BIOS which didn't help, I then switched the RAM voltage from auto to manual and bumped it up from 1.50v to 1.64v (the RAM can take up to 1.65 but the mobo doesn't have 1.65 in the BIOS) and the machine has remained on for the past 9.5 hours.

I am going t re-enable all the C-State stuff and just alter the RAM voltage then leave it running 24/7 for the next few days to test and will report back.

EDIT: Left all C-State stuff off to be safe and increased RAM voltage to 1.660 to ensure it had the 1.650 it needed. Will leave on to test until Tuesday.

[evilsatan]

Well it had a good run, lasted far longer than usual but it happened again last night. Just looking at other options in the BIOS which might need adjusting.