Possible PSU issue

[brav0hax]

I have been trying to diagnose this issue for a week or so now. I am fairly certain its the PSU, though it is a brand new unit so I didn't want to blame it immediately.

The PSU is 875w with one 7990 at stock. Everything works just fine when its not under heavy load. However when I say execute a bf of 8 chars, the operation lasts between 17-20 mins and then my system reboots.

I have watched all the sensors, the GPU is not overheating (around 80 C) and all my other sensors are well below temp ranges (water cooled CPU) So there is no temperature issue.

I've tested the fans on the cards and they respond properly and the fans in the case are fine.

I believe 875w is fine for a single 7990, but I was wondering if anyone else has run into this. I checked the forums for previous posts and found nothing directly relating to this (the googles too)

Again, running masks and rules and normal operations I have no issues, just under a heavy load it simply reboots consistently.

Any other tips on troubleshooting is greatly appreciated.

Thanks in Advance,
Eric

[epixoip]

do you have supplemental power supplied to the pci bus?

it sounds like what's happening is your gpu is trying to draw too much power through the pci slot, and either the motherboard is proactively shutting things down, or you've already melted down the 12V on the ATX connector and it's tripping the surge protector in the PSU.

this is actually a fairly common problem with dual-gpu cards, and the solution is to feed supplemental power to the pci bus.

[brav0hax]

Melted down the 12V? Wouldn't it stop working all together then?

Its a single 7990 card with a 875w PSU. I assume I want a 1000W but was told 875 was enough power for the card.

To answer your question, about the supplemental power, is most likely no. I have a single PSU that supports the entire system.

This is a n00b answer so I apologize. Other than the pci-e slot and the 2 6+2's connected there is no additional power (the 6+2's are from the same PSU)

So I assume if I get a 1000W I should be ok and if I want to at some point add a second card, 2 1000W should be used? That just seems like an awful lot of power for 2 7990's.

do you have supplemental power supplied to the pci bus?

it sounds like what's happening is your gpu is trying to draw too much power through the pci slot, and either the motherboard is proactively shutting things down, or you've already melted down the 12V on the ATX connector and it's tripping the surge protector in the PSU.

this is actually a fairly common problem with dual-gpu cards, and the solution is to feed supplemental power to the pci bus.

[epixoip]

no, it won't stop working all together if you melt the 12V. i had a board with 4x 6990 that i ran for several months with a melted 12V. very unsafe, but the point is it will still run. that board did exactly what you are describing though, it would reboot or power off under heavy load.

no, you do not need a bigger power supply -- you simply need to supply more power to the pci bus. if your motherboard does not have any molex connectors on the board to supply extra power to the pci slots, then you can get something like the EVGA Power Boost cable, or solder a molex connector directly onto the GPU.

there's more information about this problem here: http://forums.evga.com/tm.aspx?m=749507

[brav0hax]

Thanks again for the great info. Sounds like either way the PSU is the culprit and will need to be replaced.

After looking at the link you sent, I guess if I pull the plug from the board I should be able to see actual melted plastic. (Like the pic on the site)

Am I understanding correctly?

no, it won't stop working all together if you melt the 12V. i had a board with 4x 6990 that i ran for several months with a melted 12V. very unsafe, but the point is it will still run. that board did exactly what you are describing though, it would reboot or power off under heavy load.

no, you do not need a bigger power supply -- you simply need to supply more power to the pci bus. if your motherboard does not have any molex connectors on the board to supply extra power to the pci slots, then you can get something like the EVGA Power Boost cable, or solder a molex connector directly onto the GPU.

there's more information about this problem here: http://forums.evga.com/tm.aspx?m=749507

[epixoip]

you may or may not see melted plastic. if you do see melted plastic, then yes you will need to replace the motherboard and psu.

the psu is not the culprit! you only need to replace the psu if the atx connector is melted.

the problem is the gpu is drawing too much power through the pci slot, which is causing the board to draw too much power through the 12V, which is tripping the surge protection in the PSU. this is very common with dual-gpu cards. the solution is to replace the motherboard with a high-end board that has molex connectors specifically to supply extra power to the pci slots, or pick up an EVGA Power Boost.

[brav0hax]

werd. Thanks for taking me to school. I am truly learning a ton here.

You know now that you mention it, the first day or so, there was a faint smell or melting plastic. I thought it was just the new card "burning in" or something. But I assume you are going to tell me that is not normal. It wasn't a strong smell, but it was noticeable for sure.

Looks like I'll be going back to square one here...

Should have just built a box from scratch like I originally planned. At least I have the card which is the most expensive piece...

I'll search the forums for specs and such with regards to boards.

I've always learned the "hard way"

Thanks again,
Eric

[brav0hax]

Had the motherboard replaced...it wasn't the issue. The 12v rail was fine everything was in good condition when they looked at it. Issue still exists with new MB. Was told the MB should have no problems handling the card and the amount of power required, even under load.

Still assuming its PSU issue, that's next to be replaced.

Are there any other temp sensors I can look at to see if the heat is causing issues. Everything in sensord looks just fine, well below anything that would trigger a thermal shutdown.

Any tools for testing the PSU?

Thanks,
Eric

[proinside]

Any tools for testing the PSU?

Thanks,
Eric

Don't know if it is useful, didn't try it yet, but found this one that eventually might help:

OCCT "Overclock Checking Tool"