Electromigration, or Motherboard Wear?

spynoodle

spynoodle

Active Member
Over the past few months, my previously rock stable 4GHz overclock at 1.4v (Vdrooping to about 1.375-1.385v) on my Pentium E5200 has become increasingly unstable, and I can't figure out why. I've always kept the chip cool, and I've barely ever seen temps above 70C, even when stress testing with Prime95. At first I didn't think much of this possible degradation, but after seeing Prime95 fail often in Linux, I decided to install a copy of Windows 8 that a friend gave me on another drive to see if I would find the same results; in fact, it's actually much worse. In Windows, even moderate load seems to lead to a bluescreen, although I know that this was not the case about a year and a half ago when Windows 8 Developer Preview was my daily OS.

I've tried messing with the RAM frequency and timings, as well as swapping sticks in and out, but it seems to make no difference, so I've pretty much confirmed that this is a CPU issue. I've also tried moderately raising some of the P5K's slew of crazy motherboard voltages, and I've tried using Load Line Calibration, both to no avail.

I just refuse to believe that what is supposedly an entirely safe voltage for a Wolfdale chip has managed to noticeably degrade the processor in a year and a half of mostly idle usage. The only other possibility I can think of is that the motherboard VRMs could be giving me more noise than they previously did, as is suggested by this thread. Does anyone have any suggestions as to what could be happening here?
 
The first thing I would look for is bulging capacitors. Be surprised the issues they cause.
 
I don't think I see any bulging ones, but I'll have to look behind my heatsink next time I shut down. One thing I tried today is manually setting my PCI-e frequency to 100 MHz, which actually seems to have helped a lot; I've currently been running a Small FFT Prime95 torture test for a couple of hours, and it hasn't failed yet, although I did still fail a blended length FFT test after about 40 minutes. Could I maybe have accidentally overclocked my PCI-e bus along with my FSB? Perhaps it only recently became a problem when I upgraded from my old Radeon x800XT to a Geforce GT 545.
 
Okay, so it did eventually fail, and it also failed again with my Northbridge, Southbridge, FSB Termination, CPU PLL, and Clock Overcharge voltages all slightly raised. :angry:

Here's a question: is there any chance that wear on the motherboard's VRM circuitry could simply cause it to deliver less voltage to the CPU than it once did? I have it set to 1.4v, but I'm reading 1.384v idle and 1.376v load in CPU-Z. Sadly, I cannot remember if this is the same as it was a year and a half ago, or if it has actually decreased.
 
Yes to pretty much everything you are thinking.

Have you tried to completely remove all overclocks, update bios and see if that helps?

Any overclock comes with risk, although I agree with you 1.4V shouldn't be an issue.

Couple of questions:

Do you have a multimeter?
Is it warm where you live/computer runs right now?

I would be testing the 12V rail under load as this looks like voltage droop issues to me. That PSU isn't the worst, but it certainly isnt great. I would guess that under extreme load and overclock, its going out of spec, thus the bsod.
 
Yeah, I wouldn't doubt that this PSU could be getting a little flaky, especially under the load from my new GPU. Maybe that would explain why it seems like my seemingly random and infrequent instability began soon after I upgraded. Where should I attach my multimeter to measure the +12v output on my CPU rail while the system is running Prime95?

Also, as an additional note, my system seems to be running stable with all my MB voltages at auto and my CPU voltage bumped up to 1.4125v in the BIOS and ~1.39v actual. However, I still want to get to the bottom of why it seems to be excessively drooping even at idle, which seems to maybe be my problem. Oh, and in response to Okedokey: I'm running the newest BIOS revision for my board.
 
Last edited:
Connect the positive probe to a molex connector (yellow) and negative to black.

Game for around 30 minutes (to get up to temp etc). Most multimeters will read out the min and max 12V DC readings. Report those to us.
 
Okedokey said:
Connect the positive probe to a molex connector (yellow) and negative to black.

Game for around 30 minutes (to get up to temp etc). Most multimeters will read out the min and max 12V DC readings. Report those to us.
Click to expand...
Since my PSU is multi-rail, though, will that tell me if the CPU is being underpowered? I guess what I'm asking is if the CPU or GPU rail is overloading the PSU, will it be reflected on the other +12v rails?
 
Well, running a Prime95 blended test, I seem to be getting a somewhat stable 12.23-12.28v off of a 6-pin PCI-e connector, which was easier for me to attach my multimeter to than a molex connector. Is that actually too much? I'm going to try reading the voltage while running 3Dmark as well, but I have to wait for it to download.

EDIT: Just to show you what I mean by stable:
2013-08-13_13-30-00_234.jpg


EDIT #2: While running 3Dmark, I saw my voltage dip to 12.21v at times.

Since I seem to have a low amount of voltage variation, I might try to use an oscilloscope to measure ripple later today, although it seems pretty difficult.
 
Last edited:
StrangleHold said:
Long shot, but do you have another video card you could try.
Click to expand...
I could put my old x800 back in and see what happens, I guess. Although now that I think about it, maybe it actually has to do with now using the second pci-e slot; the main one broke about the same time. Still, I seem to be 100% stable with a mild CPU Vcore bump and all my MB voltages at auto, so now I'm thinking more along the lines of VRM or Power Supply degradation.
 
You must log in or register to reply here.
Back
Top