Constant crashing. Overheating power section?

[JeremyCT]

What sort of symptoms do you get when the power deliver on your card overheats? I've been using my 660ti for mining for a few weeks now, and recently swapped out an air cooled 5850 for an air cooled 750ti in the second slot. Problem is, my 660ti now crashes constantly. Black screen, blue screens, green screens (yes, green), other times the driver just quits. Takes about 8-10 minutes of mining before it dies.

I do have a 120 fan inside the case blowing air towards/over the cards, but the 660ti power section is way back towards the monitor connectors and I'm not sure how much airflow there actually is back there so I'm wondering if that's the issue. I completely removed all nVidia software/drivers and reinstalled so I don't think that's the issue.

Does this sound like the type of issues one would have if the power delivery is overheating? It's just the stock aluminum heatsink on it presently since I have a universal block on the GPU.

 

[JeremyCT]

I set the "power target" to 76% in Precision X. It's been mining happily for an hour now. I think I'm on the right track, but I'd be happy to hear feedback.

 

[GTXJackBauer]

Give it say 10+ hours at the reduced rate in precision. If you don't get a sympton than I would think its a overheating issue since you're watercooling it via only universal block. Doesn't the 750 Ti use less power than the 5850? If it does I would count out the PSU obviously as I believe you have. I believe you're on the right track.

 

[JeremyCT]

Yea, the 750ti uses WAY less juice. The 5850 ran at like 90% fan speed to keep itself cool and had a cooler design than recirculated air inside the case, the 750ti is perfectly happy at 50% and exhausts out the back. I'm thinking the high fan speed and cooler design on the 5850 moved enough air around to keep the power delivery section happy on the 660ti and now it's a little starved for airflow.

I'm at 4 1/2 hour of constant mining since I changed the power target and no hiccups or problems. I even upped it to 85% a few hours ago.

The problem is that I can't see an easy way to move more air over the power section of the 660ti. I'm going to pull the case up onto the desk in a few minutes and see what I can finagle.

 

[JeremyCT]

Went full stock. I had previously flashed an uprated TDP bios and clocks that had proven stable with one card in the system to the 660ti. I flashed it back to stock and so far stability results seem promising. I might have gotten a little too aggressive with the clocks/power on that card. Weird that it didn't really show until the 750ti was on the scene, but I might have just been right on the hairy edge and the small change pushed it over the edge.

 

[GTXJackBauer]

Glad it all worked out. Rock those puppies out.

 

[JeremyCT]

Rock those puppies out.

Dude. That's what created the whole problem to begin with!!

I'm really trying to be cautious this time. 12 hours stable at full stock, I've started easing back into it. I set some parameters based on memory, and well, that didn't work out so well. So, nice and easy back up to the (apparently new, lower) limit it is.

 

[GTXJackBauer]

Its funny, I shutdown the folding rig after foldling for almost a week for Adak. Started it up this morning and it wouldn't go passed the welcome screen. Get into safe mode after a bunch of tries and checked my dumps which was stating that I was getting BCcode 116 which is GPU related. Switched each GPU and finally got it to work individually. Than put both back in SLI (GTX 480s) and is working now. I've always had issues with the 2 480s in the past with the same 950w PSU. I have to figure out what keeps giving. I know 116 is GPU related. Could be the PSU not giving enough juice or one of the GPUs is going bad. Idk but that is driving me nuts. Don't feel like sending out hardware to the manufacturer before their warranties run out which is soon only to have them send it back to me and say it works fine. Sorry to go off topic a bid on a personal matter but it was frustrating waking up this morning. lol

 

[JeremyCT]

480s, Jack? I'd find a reason to RMA and hope for the best if one or both is going wonky. My GTX570 SC went wonky (wasn't stable at stock) and that's how I got the 660ti.

It IS seriously annoying where what worked previously suddenly stops working for no apparent reason. AMHIK!

 

[RnRollie]

with the BCCode 116 its more likely that the MB cant route the 75Watts per PCI slot vs the PSU not being able to deliver. After all the power from PSU direct to GPU is fairly simpel & stupid. So it'll rather be a seating problem or running one too many things off the MB... not running a couple of Delta's off the MB fan header are you?

Maybe at full tilt the MB is asking for more as the PSU can give on that rail.
Or the VRMs are on their way out...

Or.. i could be completely wrong ...

.

 

[JeremyCT]

Interesting that you bring up power. The power supply is a 650W Seasonic unit, should be fine. The motherboard power to the PCIe slots though, that might be a bit suspect.

EVGA sells this dohickey: http://www.evga.com/Products/Product.aspx?pn=100-MB-PB01-BR

If the problem was slot power delivery, that would probably (maybe) solve it unless it was a simple contact issue in the pins on the slot. Also interesting is that GPU PCB seems to be ready for a 6-pin connector. The pins are there, and the traces appear to be intact as well, there's just no power connector there. In theory, an ambitious person could probably add one to this card. I am not that person. The system is stable now with a lower overclock. I'm going to leave the damn thing alone and just let it mine.

I should've just waited for the stupid FTW to come back in stock when I bought this, but I got impatient. Lesson learned.

 

[GTXJackBauer]

I am a little confused. We're you referring to me Rollie? If you were maybe I should plug in the extra PCI-E Power header next to the PCI-E slots.

 

[JeremyCT]

If your board has that available, Jack, I would use it.

 

[EarthDog]

Interesting that you bring up power. The power supply is a 650W Seasonic unit, should be fine. The motherboard power to the PCIe slots though, that might be a bit suspect.

EVGA sells this dohickey: http://www.evga.com/Products/Product.aspx?pn=100-MB-PB01-BR

If the problem was slot power delivery, that would probably (maybe) solve it unless it was a simple contact issue in the pins on the slot. Also interesting is that GPU PCB seems to be ready for a 6-pin connector. The pins are there, and the traces appear to be intact as well, there's just no power connector there. In theory, an ambitious person could probably add one to this card. I am not that person. The system is stable now with a lower overclock. I'm going to leave the damn thing alone and just let it mine.

I should've just waited for the stupid FTW to come back in stock when I bought this, but I got impatient. Lesson learned.

Some boards REQUIRE the use of the extra power in SLI/CFx, others, it is optional. Check out the manual on how it should be used to confirm.

For mining, plug it in as it sucks any juice it can!

 

[JeremyCT]

Maybe I was unclear. My 750ti doesn't have the 6-pin power connector. The PCB appears to have a spot where it could be mounted, however.



I'd need to do some testing with the multimeter to ensure the traces were in fact intact behind those solder points, but I already looked and found that a six-pin connector with pre-tinned legs is available. It's kinda tempting, but not tempting enough. I've never hard-modded a video card.

 

[EarthDog]

I was more referring to jack and talking about supplemental power on the motherboard for the pcie slots.

 

[GTXJackBauer]

The board is a Asus X58 Sabertooth. It doesn't have a extra PCI-E power head as I originally thought since my main rig's intel X58SO MB does. Sorry for the confusion. I think I've had the same errors for years on both MBs and different rams used with the same CPU which is working just fine on my main for almost 5 years now. So thats why I thought its possible the same GPUs or PSU were the culprits.

(Sorry for the hijack Jeremy.)

 

[JeremyCT]

I was more referring to jack and talking about supplemental power on the motherboard for the pcie slots.

That does make more sense, lol.

(Sorry for the hijack Jeremy.)

It's cool, no worries.

 

[JeremyCT]

Still having issues. How could I cool the power section on the 660ti better? A full cover block would help, obviously, but this isn't the type of card where that sort of cash outlay makes any sense.

Would a backplate help? Could I just get a few more RAM heatsinks and stick them on the back of the card where it's overheating?

For reference, this section on backside gets wicked hot:



And this is where the power section heatsink is and why I'm having trouble getting good airflow over it. Despite the fan almost right in front of it, it's difficult to get airflow into that specific area.



Should I give up and just run stock clocks? That doesn't seem to be the overclockers way, but I'm having trouble coming up with a solution that doesn't cost more than I really want to spend. Thoughts and ideas appreciated.

 

[caddi daddi]

cable ties make very good fan mounts if you case side is not made to have one mounted on it.