KWSN Orbiting Fortress

lvanst

I'm getting platform failures under heavy load. No failures if the platform is idle. The screen goes black, followed by a BIOS restore/recovery, so it's a pretty hard crash.

I removed the Radeon card, and it still crashes running only Nvidia
I removed the Nvidia card, and it still crashes running only Radeon
So it doesn't look like the video cards, or drivers, are the problem...

I use hwinfo64 to monitor/log platform sensors, and the temps are stable right up to the crash:

CPU 60c (AMD FX-8350)
Motherboard 50c (Gigabyte 990FXA-UD3)
Northbridge 46c
Radeon 78c (R9 270x)
Nvidia 75c (GTX 760)

CPU and both GPUs have been rethermaled within the past six months.

I also ran a Windows memory test, to confirm my Corsair memory was healthy.

I'm out of ideas. It's definitely load related, as it will idle for a day without failures, but detonates within a few hours under heavy load.

Any help would be much appreciated...
#ni-1

_________________

lvanst · Posted: Tue Nov 22, 2016 9:23 pm Post subject:

33 views and no ideas or suggestions?

Help me out here my fellow knights. Rolling Eyes

_________________

JumpinJohnny · Posted: Tue Nov 22, 2016 10:58 pm Post subject:

PhastPhred · Posted: Wed Nov 23, 2016 3:06 pm Post subject:

I assume you already did the Vacuum Cleaner stuff, and have a clean CPU and Power Supply - Vent holes in the case, etc? - Sorry, I have nothing...
_________________

lvanst · Posted: Wed Nov 23, 2016 7:06 pm Post subject:

Thanks guys. I too was thinking mb or power supply. I'll check for a bios update.
_________________

JumpinJohnny · Posted: Wed Nov 23, 2016 7:54 pm Post subject:

KWSN-HoC · Posted: Fri Nov 25, 2016 7:43 am Post subject:

Sounds to me the power supply has aged capacitors and max load isn't stable anymore ....

Cheers

HoC
_________________
Housekeeper of Camelot

Mastrumistulo de Kameloto

(also a Member of the Migratory Coconuts)

'My name is Homer from Borg. Resistance is fu ..... Oh doughnuts!'

sir spuddly buddly · Posted: Sat Nov 26, 2016 1:47 pm Post subject:

Yeah, agree with start with getting BIOS right then slowly cranking it up.
_________________
a monster can be excused for his behaviour . . . The problem is not how a monster could do it, but how a human being did it.
Scaring 11 year olds since 2005
"There is only one path to Ni!-dom - through the shubbery"
Click every day at http://naturarvet.se/en/ to save a forest!

lvanst · Posted: Tue Nov 29, 2016 2:46 pm Post subject:

BIOS flashed and still failing.

Trying a new power supply next...
_________________

lvanst · Posted: Tue Dec 13, 2016 5:20 pm Post subject:

This stopped happening when I stopped running the Einstein project. It seems to have been CPU centric, as I took turns removing the nvidia and radeon cards to no avail. The platform has been failure free for two weeks now, whereas it had been failing multiple times per day...
_________________

JumpinJohnny · Posted: Tue Dec 13, 2016 6:11 pm Post subject:

Soooo...
We were right!
Gravitational waves from Einstien project were overloading PS capacitors and causing the BIOS to cause CPU errors when communicationg with the GPU's.
It was probably that newly discovered double pulsar that threw a whammy on the CPU. Very tricky, those gravitational waves... popping capicators, unflashing BIOS and throwing GPU's into interdimensional quanta.
Rolling Eyes

If it's possibly related:
I get excessive 'hardware errors' from my U3 Bitminer whilst running the Einstien "Multi-Directed Continious Gravitational Wave Search" WU's and also long running Prime Grid WU's like the PSP (LLR), even when I limit the consecutive WU's to half of the cores.
If I give the CPU a rest for a day or reboot and put something different on the CPU like Astroids, then the Bitmain U3 Asics miner gets almost no errors and hums along just fine.
I think some of these programs write their applications to 'hold open' memory without actually using it.
Just an observation. It would be more coherent if I knew what I was talking about.

_________________

sir spuddly buddly · Posted: Thu Dec 15, 2016 12:41 am Post subject:

Glad to hear you found the source of the problem - hope you've told the people at Einstien.

#ni-1

_________________
a monster can be excused for his behaviour . . . The problem is not how a monster could do it, but how a human being did it.
Scaring 11 year olds since 2005
"There is only one path to Ni!-dom - through the shubbery"
Click every day at http://naturarvet.se/en/ to save a forest!