[PAID .1 BTC] Problem getting all GPU's working

[wedge]

.1BTC bounty to anyone that can identify the problem.


I've got two identical rigs.  Here's the specs on both:

AMD Athlon II dual core
Kingston 4GB ram
MSI 990FXA-GD80
4x Gigabyte GV-R795WF3-3GD
Powered 20cm pci-e risers from hashratestore.com
Seasonic X-1050 (SS-1050XM Active PFC F3)

They are running BAMT 0.5C off an 8GB usb stick.


One of them is working perfectly, running 24/7 for almost two weeks with no issues.

The other one works perfectly, if I only install 3 cards.  If I install all 4 cards, they all get detected, and they start mining for a few minutes.  But in about 5 minutes the machine will just shut down.  The power button on the motherboard is unresponsive, I have to switch it off and on at the power supply to bring it back up.

I've swapped the psu from the bad rig to the good rig, and the problem did follow the psu.  So replaced it today, but the replacement is having the same problem!  I know this model of PSU can handle the load, because I have one that is handling it perfectly well.  But why would I get two that aren't working at full load?  What am I missing here?

This has been going on for 2 weeks now, and I can't figure it out.  Any help?

[derr777]

Have you tried testing the other 4 cards from the other rig in this one?

I once had a card that when it failed it would 'lose' its voltage regulation and just pull more and more power at the wall until the PSU was overloaded and it would shut down.  I witnessed it with a killawatt or I wouldnt have believed it.

I've also had cards that would do this to a machine if they were clocked a bit too high.

worth testing anyway..

[wedge]

Have you tried testing the other 4 cards from the other rig in this one?

I once had a card that when it failed it would 'lose' its voltage regulation and just pull more and more power at the wall until the PSU was overloaded and it would shut down.  I witnessed it with a killawatt or I wouldnt have believed it.

I've also had cards that would do this to a machine if they were clocked a bit too high.

worth testing anyway..

I've tried every combination of 3, and even swapped cards between rigs.  It happens when any 4 are installed.  So it's not any one of the cards causing it.

The cards are overclocked, but not too high.  I haven't tried running at stock yet, so that is worth a try.  I'll try all 4 at stock and see what happens.


Edit: Remember that the problem did follow the power supply when I swapped that between rigs.  So it feels like a power supply issue.  But it's very odd that I got two that both didn't work on this rig, and one that does.

[wedge]

Okay, it's been mining on all 4 at stock speed for 15 minutes and counting.  Already better than before.  If it lasts 24 hours without a shutdown/freeze/lockup, then I will declare it stable and the bounty is yours.


Edit: Even if this works, something doesn't feel right to me.  It should be able run with the settings I had...  I was running the overclock that's pre-packaged in BAMT, 1085MHz core and 1375MHz memory. Which is a very safe/sane overclock for any half decent card.  Each of the cards has proven they can do it easily, and so has the psu.  That psu should be strong enough to handle a 5th card with room to spare.  There's got to be more to this...

I'm going to let this go as-is for a day and make sure it's stable.  Then at least I finally know what area I need to be working on.  If this lasts then I'll start playing with the clocks to see what it can tolerate.  Then I may end up offering another bounty to anyone who can figure out why it's unstable with such a moderate overclock.