XFX 7970 DD 2012 model power surge issue

[goxed]

Hi guys,

I have a XFX 7970 DD 2012 model (it's a bit different from the original XFX 7970 DD), and used it for mining since november 2012.
 What's happening is that the card is using so much power at its default clock / voltage (> 300W) that the PSU (tested with Seasonic 750W 80Plus Gold and Sparkle 600W 80Plus Platinum) is turning off the computer.  
The power surge happens only at the default clock / voltage (925MHz / 1.13V) but not at (700MHz, 0.95V).

Any ideas / suggestions. Has someone else faced this issue?

Hi Guys,

The problem is repeatable in both Linux and Windows. Its repeatable on a computer which can successfully mine with an overclocked Radeon 7990 (=2x7970).
I removed the 7990 and inserted this problem 7970 and the system power consumption surged within a few seconds of mining and the computer turned itself off. I have never seen such behavior before.

[1714890660]

1714890660

[1714890660]

1714890660

[1714890660]

1714890660

[1714890660]

1714890660

[1714890660]

1714890660

[qbits]

I faced similar issue (same symptoms) with 6990 and a 750W power supply using windows/guiminer/phoenix. I could start one instance of phonenix loading one GPU, but not two as the computer would shut down in about 10 secs unless I dropped the gpu clock to about 700 MHz.

Surely enough hash rate suffered as best I could hope for was 650 MH/s using both GPUs.

Switching to linux(ubuntu) and bfg/cgminer using the same hardware proved to be a solution once I was able to properly install Ati drivers with appropriate xorg.conf file that initialized both GPUs.

I was therefore able to increase hash GPU frequency to 900 MHz and consequently hash rate to about 760 MH/s give or take a few percent.

+ the whole thing can boot from 4GB usb key and needs no disk
+ reboot time is quite fast
+ linux provides for such niceties like "byobu". if you don't know what that is you are indeed missing out.
+ no need for vnc/remote desktop

I've stayed with 900 MHz as I feel comfortable running these cards in the 80°C range, not going over 85°C.

That's all fine and well, however recently I began to suspect that the real reason for instant poweroffs was related to sudden increase in temperature which apparently phoenix could not manage properly. so I began to experiment with cgminer on windows. sure enough I was able to get the card running at say 850 MHz range. Going to 900 was still a problem as when I did that I would see in Afterburner GPU frequency jumping between 800 and 900 MHz and not staying stable at some predefined frequency.

as I was not happy with that and because I was worried about the temperature spikes, well not really spikes, but rather slow 10 minute temperature cycles where temp would rise to almost 90°C, so I returned to ubuntu. I did get slightly better hash rate in windows though (777 MH/s)

one more thing: when you say 750W PSU, that does not say a lot. for example 750W PSU from say Corsair is able to absorb quite a load and system will be stable. 750W noname PSU can have a theoretical peak power rate of 750W with 4x 12V lines that should not be loaded with more than 16A, and even approaching 16A can cause problems.

I had one cheap noname 500W PSU and I tried it with 5850 which with the whole system should not use more than 380W (CPU was not loaded and I measured power usage at the wall). So plenty of headroom right? Wrong: PSU was dead in about 1h.

for what its worth...
BR

[scifimike12]

@goxed, Is this a single card you're using in your rig?  What are your temps like (GPU, VRM)?

[ssateneth]

I had a similar problem at one point. I was mining about 5 GPU's on an evga nforce 790i sli ultra mobo, and it would turn off at random (not shut down, power would just go away and stop). I thought it was my power supply (PC power & cooling 1kw-sr 1000 watts) so I RMA'd it. The guys said there was nothin wrong with it. I got back a reconditioned version and the issue continued to happen, -especially- if there was any sort of significant CPU load (it was a single core celeron 420 which was weird). When I upgraded to a newer msi z77a-g45 motherboard, the problem immediately went away.

tl;dr it's possible its your motherboard and not your power supply causing shutdowns on heavy load.

[goxed]

Hi Guys,

The problem is repeatable in both Linux and Windows. Its repeatable on a computer which can successfully mine with an overclocked Radeon 7990 (=2x7970).
I removed the 7990 and inserted this problem 7970 and the system power consumption surged within a few seconds of mining and the computer turned itself off. I have never seen such behavior before.

[qbits]

@goxed, Is this a single card you're using in your rig?  What are your temps like (GPU, VRM)?

at this point in time this is the only card in the system. reason was lack of proper psu. recently i got reasonably priced corsair 1000w psu and will consolidate this card with 5850 and 2x 5830 depending on what slots mobo will support.
br

[crashoveride54902]

Hi Guys,

The problem is repeatable in both Linux and Windows. Its repeatable on a computer which can successfully mine with an overclocked Radeon 7990 (=2x7970).
I removed the 7990 and inserted this problem 7970 and the system power consumption surged within a few seconds of mining and the computer turned itself off. I have never seen such behavior before.

sounds like something is bad/going bad in the card...either the vrm or the gpu itself might be on the way out...would be my guess...just rma it