Strange issue

[Vezina]

I have made this rig for a neighbor , almost 3 years ago if i recon correctly.
It consists of :

Gigabyte 770T-UD3 v1.5 (F10)
HIS AMD 6870 1Ghz with a nice 5 heat-pipe cooler
Corsair CX 600 PSU
4Gigs of Corsair RAM
AMD Phenom2 X4 980BE + Cooler Master TX3
2 normal HDD-s.

Now the guy has started reporting a strange reset issues while playing some games ,with no BSOD code.He was also reporting strange noises.Only extremely stressful games seem to trigger this.

So took it for analyze and discovered dust and some temps getting out of range a little ,so i said this may be the issue.Video card hitting only 72 Celsius on the GPU and the CPU a bad 68 Celsius.After clean up i had some nice temps maximum 62 Celsius for CPU and barely 68 for the GPU.Running Kombustor or Intel burn test did no show up anything even with the rig full of dust though.

After playing around cleaning it up , i have reinstalled Windows to rulel out video drivers or OS corruption and started the stress tests.
Right after the Kombustor run started a strange electrical noise was generated in the PSU judging from my ears.Stopping all fans one by one did not clear it up.It would pass away though after some time of running the stress test.Temps looking fine no more than 68 Celsius.

Seeing this turned on along Kombustor ,the Intel burn test and a 3DMark 11 demo instance .So with Kombustor in the background i had this 2 running.

Now here comes the interesting stuff.
If at first run of all this it s all okay at the second run ,when the the Fire Strike test starts ( after the Intel Burn test finishes it s 10 -th run and stops ,the CPU feeding only the benchmark)  the PC lock ups with a black screen ,everything running , fans an so on .Power draw at that time is like 135Watts when sitting lock-ed up.This happens always at the second benchmark run during the FireStrike demo.First run finishes successfully always

During this FireStrike test the power meter shows a 240Watts consumption ,being much lower than in other suite tests when it reaches easily 280Watts.
After the black screen happens it either stays as it is ,either the PC resets itself after few seconds ,no BSOD and no error in the Event Viewer.Only the Windows start-up shows the options to start normally or in other modes.

If i force the reset by the reset switch it may continue to stay this way ,black screened ,when the boot-up routine starts ,but beeping a strange code until i turn the PC off totally and restart it from cold.The beeping it s not continuous though, a long beep finishing short and higher tonality is done once ,then nothing happens sound stops ,until i reset again from the switch ,before i turn the PC off.The switches of reset and power work during this.

I have no idea if it s the PSU ,the video card or motherboard chipset .The fact that no errors are recorded in the event manager makes me think it s a hard reset ,either the motherboard chipset lock s up either the CPU does something.If CPU would overheat it should shut down though , its  nowhere more than 62 Celsius which is perfectly fine for a 980BE.
The only temperature getting higher is the one reported by HW Monitor as 80-84 Celsius which belongs either to the chipset either to the VRM (most probable).
PSU not even half loaded so w ..
RAM no problems in Memtest.
Both CPU and GPU coolers are ok at touch ,no excessive heat.

Any hint before starting the swaps ?!
Revision 1.5 of the board takes the CPU officially so no problem should happen either.
Never seen something like this without the complete automatic shutdown of the PC ,thats why i am puzzled.

Thanks for reading the "wall" of text

[autotech]

I have had this happen and it was a video error. Had this happen on my last set of 5890's actually rma then to etailer and traded in for a 5950 and it didnt happen anymore. The thing i find strange is the sound from the PSU.  Makes me wonder if a capicator is going out in it.

If you have a few parts you can swap out i would start with the video card and use Driver sweeper to get rid of any old drivers before you install the new ones.  If it still happens that is when i would leave the video card in and swap out the PSU.  I know it is a pain but i build alot also and sometimes you have to swap parts to find the real culprit.

Out of all my gigabyte motherboard i only had one that 2 ram slots died in so that is the last part i would try.

[Vezina]

Thanks for the feedback man.

I went and continued with troubleshooting and it looks like video card related ,but i am not sure it s the video card itself or something else related to it like the motherboard way of delivering power tru the PCI-E slot or the PSU.

After stress testing i have added a fan to blow into the power supply and it looked like the issues would go away ,but then after repetitive testing , boom it started again.Changed the fan in the power supply and then by adding a 12 cm fan over the video card back  it looked steady ,but occasionally would give some video errors at benchmarking.

Then I ve set the video card to lower clocks ,the stock 6870 clocks ,it s a factory overclocked 6870 , instead of 920/4200 clocks it has 920/4600, and set the fan to 70% and again it looks like steady.

I don t understand if it some over-current protection kicking in on the video card ,motherboard or strangely the PSU rail (though in this case the PC should of shut down all together) or if it s temperature related as the card it s extremely cool at touch in load.

On the other hand i took the card and installed it in the FM 2 system where i am running XP.Tested by using both 3DMark 03 and 3Dmark 06 (weak benchmarks by now as stress) along Kombustor and in spite of getting hotter as under the card was a sound card ,so lower air flow , it looked steady as a rock ,but of course i am not able to run the 3D mark 11 in there.

Voltages on 12V rail look also steady
The card is also almost non dusty.

Could it be some hidden incompatibility between the mobo and video card that gets triggered during heavy usage ,like the amps coming from the slot being much lower than the card expects ,or the voltage supply on that slot going into thermal protection ?!

Or maybe the GPU on the video card ,has an over-current issues ,as we know that since the ATI 5xxx series they have placed inside the chip protections against "thermal viruses" as they call them referring to Kombustor and such.

Anyway i ll keep testing the lower clocks and obtain stability and advice the guy to get a new video card ,it s time for upgrade anyway
I hope it s not some motherboard issue particular to this card.

A new Windows has been installed ,prior to testing after the full dust PC clean-up ,with the latest 13.10 beta driver in there.I wouldn t go lower than 13.1 with the drivers as AMD have issues with UT3 engine based games with drivers prior ,back to Catalyst 10.3 i think.

Thanks again.

[autotech]

On mine something on the 5850's wasnt kosher with the motherboard. Drove me nuts for about a week before i ordered the 5950 and tried it. It was at that point i knew it was something in either drivers, bios though it was fully updated or just in the cards themselves.  I kept the 5950 as it cost a little more but was stable as all get out.

Was worried when i went up to the 6950 i really wanted the 5990 but didnt want the same problem so setteled for less with lots more stability. I have a few coarsair PSU in other computers i have built and not has failed yet and the oldest one is a 750 and it is at about 4 years ols and have 2 i built with the tx-650 and they are also.

I have one i built with a builers cx 500 coarsair and it has a restart problem when left sitting idle. The person that owns it said they are cool with it as it doesnt whine and it only happens once every week or two. I was wanting him to bring it to me to troubleshoot but as i gave him a really good price he isnt worried about it.

[Vezina]

I may have nailed the issue.Set the card to 900/4200 clocks as per stock of ATI 6870 reference.
Dismantled the cooler of the card even if it looked clean and discovered a strange thing ,the memory modules have indeed a small aluminium heatsink on all of them a simple plane aluminium stuff.

While the GPU has the beefie 5 heat-pipe cooler there is no contact of the main heat-sink with the memory heat-sink.I have no idea if it s overheating or something else ,or where it occurs ,motherboard ,PSU or video card though.

The power supply on the other hand had a bad fan and it has at cold boot a strange electric noise after at start i  turn on of the Kombustor stress test.After 4 minutes or so of Kombustor the noise goes away though.The 12V line looks steady though ,maybe the PSU delivers some high frequency voltage generating the crashes in specific conditions ,i have no idea.

Oh and shocking is that the 5600K APU with 4 Ghz for the CPU side and 500Mhz for the GPU side (lower than stock) gives me more 3D Mark 06 points than the Phenom 2 X4 980BE + 6870 1G combination.21000~ points for the APU compared to 19700 for the 6870.

Thanks again for the feedback.