• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

System overclock becoming progressively less stable

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

JeremyCT

Member
Joined
Feb 26, 2009
Location
CT
Weird happenings. I'd love to hear some thoughts. Full system specs are in the link in my sig, but the basics listed in the sig are current.

I started cryptomining about a month ago. For the first week everything was fine, I had an overclocked 660ti with a modded BIOS for TDP and clockspeed increases. It mined at 1250 MHz without a hitch.

A few days later I added a 5850 to the mix because I had it sitting around (hey, why not?), and everything went fine once I got the software up. After a week or so of that the 750ti was released and I got the upgrade itch, so I picked one up so I wouldn't have to hear the 5850 scream anymore. Set it up, overclocked it slightly (+50 core on a factory overclocked card). Everything was stable.

That was all humming along for about 4 days before things started going downhill. I started getting hard system lockups while mining. I tried all sorts of combinations before eventually removing the core overclocks on both GPUs. 660ti BIOS was returned to stock, no clock offset, 123% TDP setting, small memory overclock. 750ti was left at stock (factory overclocked) core speed with moderate memory overclock.

That was fine for about another week before I started getting lockups again yesterday. I toyed around with a few settings again, got nowhere, then I set both the GPUs to full stock and got no lockups. I've moved the memory offsets back up and everything seems to be running fine again, but I'm concerned about the trend I'm seeing here. There seems to be a trend where SOMETHING is degrading and my overclock ceiling keeps creeping lower and lower. I'm concerned that if this keeps up I might be unstable at stock within the next week or two.

System temps are very good. The 660ti has a universal waterblock and the radiator is mounted in the front fascia of the case blowing air into the case. Plenty of airflow. The 750ti never exceeds 62 deg C while mining, the 660ti stays under 40 deg C.

Total system power needs at peak based on back-of-the napkin math shouldn't be more than ~575 watts or so.

The PSU is my first thought, but it's a 650 watt Seasonic unit I bought from Oklahoma Wolf just last summer. It should be awfully solid, but I suppose a problem there is possible and would explain things nicely. Power comes through an APC Smart-UPS 1000 which supposedly delivers 670 watts. I've never heard of UPS power delivery degradation, but I suppose that's possible?

I'm open to any other theories people might have. Thoughts? Any thoughts on easy or sensible ways I could narrow this down would be especially appreciated.
 
Last edited:
FurMark Each GPU individually at stock settings. Watch the temps. Let it run until one of the cards brain farts. Long tests are good for long load temp readings.

If the "lockup" was better described, I'd say check the Disks too. Ram as well. Assuming the Cpu is always stable, run that stock for testing anyways. Test that separately.

Test, test and more testing. All about pin pointing.
 
I have pinpointed. I know the error originates with the GPUs. Thing is, it's not one or the other. If I increase the base clock or TDP slider on either, I get a hard lock. These are settings that were previously stable on long stress testing, and one step down from that as well. Full stock is currently stable, but I'm concerned it won't be for long. "Hard lock" consists of a solid black screen. No BSOD, no event recorded by Windows, no memory dump. The fact that reducing the power consumption of the GPUs lead to returned stability each time, plus the fact that I couldn't peg the problem to one GPU or the other specifically, is what makes me to believe that it's a power issue of some sort, but I'm not really certain how to definitively pinpoint those sorts of issues without throwing parts at it and seeing if things improve.
 
Last edited:
Well, PSU has been ruled out. 12 volt rail is stable. I'm back to suspecting the power section on the 660ti since it doesn't get a ton of air flow.
 
Back