OC'd system deterioration: Revisiting an old issue with some specific questions . . .

[BonzaiDuck]

I built a Wolfdale E8600 system with an nVidia 780 mobo -- almost two years ago. Mobo: rated for FSB 1333 -- the OC upped the CPU to 4.25 Ghz and FSB 1700.

This has been a very stable system -- running within the thermal spec as seen at load -- with voltage pushed to between .04 and .02 volts below the upper limit of the spec "safe range."

Occasionally, I "re-certify" these settings with LinPack and Prime95. I've noticed that over time, the voltage no longer seems sufficient to sustain stability for the OC setting under the load tests. It means that I either need to drop the OC setting, increase the voltage (which I don't want to do at its current level) -- or retire the system to "default" settings.

I've also noticed (for this motherboard) that adjusting the OC settings doesn't seem to "take." It takes a couple reboots and re-adjustments to get the system to boot into the more conservative settings. I shut the machine down this morning with the lower OC setting, started it up again this afternoon, and it booted to the higher setting!

What is causing the BIOS/CMOS to hold the previous setting?

Is the deterioration I noticed under this latest Prime95-run attributable to the motherboard (North Bridge?), or the processor itself? Or -- both? I had thought to replace the Wolfdale with a Q9650 Yorkie -- just to keep the machine "alive" and "useful" -- but I didn't want to move too quickly as long as the Wolfie seems reliable. Depending on how this phenomenon derives from which pieces of hardware, it would determine whether I just put the change together to move up to an I7/I5 configuration -- as opposed to buying the last-gen quad-core. . . .

 

[cbn]

I have been under the impression that the mainboard is usually the culprit in these circumstances.

 

[BTRY B 529th FA BN]

Try a new RAM kit. After hrs of passing memory tests, a newer kit proved to be where the issue was and brought my OCing back to life.

 

[BonzaiDuck]

Try a new RAM kit. After hrs of passing memory tests, a newer kit proved to be where the issue was and brought my OCing back to life.

Interesting, and I won't discount it. Even so, the RAM in this case was a stellar "Black-Pi" kit from G.SKILL rated at DDR=900Mhz. In this sense, the RAM is actually under-clocked. But the latency-setting spec is 4-4-4-12 @900, and I have them set at 4-3-3-8.

And one more thing: To check the TCase temperature under load (as well as the individual cores) I ran LinPack "IntelBurnTest" about a week ago for five iterations -- not really enough to prove stability -- but supposedly more reliable than Prime95. This discovery about Prime95 instability occurred yesterday, and it failed variously in three to ten minutes. Also -- the test I ran was "small-FFTs" -- focused on the processor -- not the memory.

Any additional thoughts? If it's the motherboard -- I can salvage the processor, or just reduce the clock and voltage. If it's the memory, I can either loosen the timings, down-clock the OC, or both -- and I have some spare memory kits -- both of them G.SKILL rated at DDR=800 or above.

The mobo was the lesser of the two eVGA 780i models produced. Supposedly, the better of the two was the FTW model, which had solid-state capacitors in contrast to my 132-CK-NF78-A1(?).

 

[BTRY B 529th FA BN]

Funny I have 8g of those same 2x2 kits

What I was referring to was on a Q9550 and my UD3 setup. I had a DDR2 1000MHz kit and my UD3 was more than capable of running 500fsb or 1:1 In less than a year the kit servely degraded. It wouldn't run 400fsb @ 3.4Ghz. I then bought a 2x1G kit of PNY DDR2 800 and my 3.4Ghz clock was stable again.

 

[VirtualLarry]

My only experience with "deterioration", is with my friend's IP35-E and E5200 that I put together and overclocked for him. When I built it, I did some extensive Prime95 testing, and ran the CPU at FSB 300, 3.75Ghz. About a year later, the computer started making some beeping noises while running in games, and it was found to be overheating slightly. Cleaned the dust out and maybe re-did the TIM (I don't remember exactly), and the temp problems got taken care of. But when I did some more testing with Prime95, 3.75Ghz errored within 10 minutes. Not good. So I dropped the freq to FSB 290, which comes out to 3.625Ghz. Still a decent overclock, but something wasn't running as good as it used to be.

The BIOS vcore setting was 1.425v, which came out to around 1.4v under load. The computer used to be under load all the time, as I had my friend running SoB for me. But he no longer wanted to run that, so instead, the idle voltage was higher, above 1.4v. I have heard that 1.4v is around the max safe vcore for 45nm CPUs.

So, I have no idea whether the CPU degraded, from the voltage, and/or temps, or whether the motherboard is slowly giving out.

 

[Vesku]

I'm with "Computer Bottleneck" on this, it's usually the motherboard aging. Often as a motherboard ages it starts showing more Voltage droop or is unable to deliver clean power at the wattage demanded. This is especially true when an OC brings the board close to it's maximum deliverable wattage.

 

[mav451]

I agree - I wouldn't discount testing the RAM too. RAM can die suddenly and unexpectedly...but some signs are a flakey OC. As for mobo "aging" - that's just depressing haha.
Also as for a quad on a nVidia chipset...I probably wouldn't bother.

 

[Gillbot]

as soon as I read the title, then saw this:

nVidia 780 mobo

I figured mobo or Ram. Of all the nv based boards I used, they all tended to EAT ram or "degrade" but i've never had a CPU "degrade" unless it was truly abused.

 

[DominionSeraph]

Goddamn why are so many people here addicted to useless tests?

If your computer doesn't flake out or lock up, guess what? It's stable. Making your CPU draw max current for 24 hours isn't going to make it more so.

And if you wanted long-term stable what were you doing at 4.25GHz?

 

[BonzaiDuck]

Funny I have 8g of those same 2x2 kits

What I was referring to was on a Q9550 and my UD3 setup. I had a DDR2 1000MHz kit and my UD3 was more than capable of running 500fsb or 1:1 In less than a year the kit servely degraded. It wouldn't run 400fsb @ 3.4Ghz. I then bought a 2x1G kit of PNY DDR2 800 and my 3.4Ghz clock was stable again.

Did you get an idea of voltage differences for the memory? I'm wondering if the real degradation was on the motherboard as it supplies power to memory and CPU components. Then again, we may be looking at variations on the problem and phenomenon. In your case, you're saying there was a 100 Mhz drop in RAM performance at your RAM's original spec. In my case, we're talking about 10 Mhz . . .

I might even have some wiggle-room on the RAM voltage setting, and I could see what happens there. But again I point out that my observation occurs with the small-FFTs test.

This is at least the second time I've seen this degradation. There was more room to push up the VCORE in the first case -- RAM performance didn't seem to be a problem.

I just finished a 15-hour-run of Prime95 small-FFTs on the Wolfdale system -- 0 warns, 0 errors. As I mentioned earlier, I chose to leave the VCORE alone and drop the timing -- or 1,700 FSB/850 RAM/425 [CPU]FSB to 1,680-840-420. I lost 50 Mhz on the CPU overclock speed.

I've heard of [conventional] capacitor aging on components like the power-supply, and turn again to the difference between the two eVGA 780i motherboards. But I wouldn't know for sure how capacitor components would affect motherboard features. I'm in the dark on these electrical issues or motherboard-design.

Another forum member had made an observation just after I reported my OC'ing results when I first built the machine -- noting that he had to keep upping the VCORE setting to keep his E8600 Wolfdale stable. So this is not some isolated phenomenon: it is something OC'ers would need to periodically assess and watch out for. That's just my guess . . .

How the newer gen of I7/I5 hardware behaves -- I've yet to find out. It's been a bad two years on the money angle. I've postponed "keeping up" to the now-established Bloomfield/Lynnfield/Gulftown technology just out of caution for "more serious priorities." But -- it's gonna happen -- even as there may be something newer on the horizon.

A 5 Mhz drop in my CPU FSB is not much of a disaster -- so far . . .

 

[BonzaiDuck]

Goddamn why are so many people here addicted to useless tests?

If your computer doesn't flake out or lock up, guess what? It's stable. Making your CPU draw max current for 24 hours isn't going to make it more so.

And if you wanted long-term stable what were you doing at 4.25GHz?

I suppose you have a point, depending on what the machine is being used for. This one was a total luxury for the fun of it. Nor would I have known that something changed unless I'd just attempted to run Prime95 for only ten minutes.

Fact is -- Prime is mild compared to Linpack tests.

The 4.25 OC had been a 28% OC -- not an excessive expectation for these chips. CPU Magazine had once given a tongue-in-cheek guideline of "totally safe" at 10 to 13%, but that was a previous generation of chips.

The more I think about this, given the responses of our brethern here, I'm eyeing the motherboard more and more. At 1,333, the mobo is at spec. There are more components that can age. Neither the thermal spec nor the voltage spec has been exceeded on the CPU. Both components are over the speed spec, but the mobo has "more parts" -- and parts that can age for supplying power.

 

[BTRY B 529th FA BN]

Have you ran Memtest HCI?

 

[aigomorla]

everything degrades.

The only thing that doesnt degrade is a diamond.

Each time current is fed though a curcuit it gets degraded.

Over time it builds up and you lose oc or whatever u had.

This is the reason why intel bins much lower then what the cpu is capable of.
So by the time it hits a critical degrading point, it will still function on stock settings.

Once u play with overclocking, you accelerate that process.


The things in order in how they die from my experience:

1. Ram... 99.99999% of the time your ram will die on you from usage.
2. Board
3. everything else
4. cpu - unless ur a 13 yr old kid who enjoys torching cpu's at 1.5vcore. (not applicable on AMD... im talking 1.5vcore on intel)

 

[mbevolution]

ya with my e6300, was able to run 2.6 fine, then it just pooped that one day and I tried to mess around with it for a while and eventually gave up and ran it at stock.

 

[BonzaiDuck]

Thanks, Aigo . . . It's been a while . . .

I've got about six machines to track and service in the "extended family."

With the delay in my "new build" plans, it's important to inventory and re-evaluate reliability.

Like ??revolution said, if one plans to use a machine like mine as a fallback for more serious computing -- it's probably wise to reset the BIOS at stock. But also-- this seems to be a matter of 5 Mhz on the CPU-FSB -- from 425 to 420. By the time I need to worry about this, I'll be building a Nehalem . . . .

Anyway -- on Aigo's list in order of likelihood. It would seem that this depends on which components are run out of spec to greatest extreme. In my case, it may be the motherboard. This doesn't seem to be a memory problem.

At the moment -- it's just an "adjustment" problem. . . . . Er, except for fact that the BIOS keeps going back to the higher OC setting everytime system is started from cold boot.

 

[BonzaiDuck]

THE "STICKY" BIOS SETTINGS SOLVED

Of course we all know that nVidia boards are last-gen-things-of-the-past, so this would seem totally useless for those interested in up-to-date forum information.

I'd installed nVidia's control panel and system monitor -- the software that supports that Silverstone automatic fan-controller -- which I have.

Apparently, profiles saved under the control panel cause this sucker to "hold" the old OC. Even if you change the BIOS, those profiles are loaded with the OS and over-ride the manually-set BIOS.

Anticipating advice responses to this, I'd be chided for not uninstalling the nVidia software. These things had been noted, discussed and concluded over the years. But apparently once you sync the software settings with the desired BIOS settings, the mystery of the failure of new settings to "take" disappears. And I'd run this puppy with the same OC settings for almost two years with the software installed.

So other than an "aging" of the motherboard -- what Gillbot called "depressing" in a pun -- it's still "holding up" with a 50 Mhz downward adjustment to the CPU speed, or 4.25 revised to 4.20 Ghz.

Fun and games; fun and games . . . .

 

[lopri]

everything degrades.

The only thing that doesnt degrade is a diamond.

Each time current is fed though a curcuit it gets degraded.

Over time it builds up and you lose oc or whatever u had.

This is the reason why intel bins much lower then what the cpu is capable of.
So by the time it hits a critical degrading point, it will still function on stock settings.

Once u play with overclocking, you accelerate that process.


The things in order in how they die from my experience:

1. Ram... 99.99999% of the time your ram will die on you from usage.
2. Board
3. Everything else
4. Cpu - unless ur a 13 yr old kid who enjoys torching cpu's at 1.5vcore. (not applicable on amd... Im talking 1.5vcore on intel)

qft

 

[mutz]

yeah, but it depends how fast it would degrade,
ElectroMigration degradation should be the issue with CPU's being overclocked, even though the interconnects are made out of Copper and not as been with aluminum, processors are not build exactly for OCing,
they're expected life frame is determined at the factory and accordingly they're clock and voltage.
it should depend on the CPU itself and you can never know how long it would last or what did actually fail unless each individual piece is being replaced and retested.
any OC is a certain risk, either under volt it, clean it or change parts .

 

[nyker96]

I don't know if one can say OC is the cause for these deterioration. It may just happen anyways without OC. Also maybe the case and cooling system also plays a factor here.

 

[mutz]

heck, this response was made without reading the whole thread,
i would really start with cleaning the system and replacing the thermal paste, these are small things to do which cause no harm and are always welcome, especially if it has been a long period of not doing it,
the TIM might get dry and dust can add up everywhere burdening the system.

2. the MB seems like a long shot, you can start with the RAM if you have replacements,
it might not bother losing only 5MHz yet it is always a benefit to lay aside speculation and getting as close as possible to finding the real cause.

3rd point, if the system has deteriorated, it might mean it will deteriorate even further, faster.
i.e if the CPU was holding fine 4.25 and would have hold even 4.5 without any issue, having it reach instability means it is losing it's resilience and you might find yourself having to underclock it further each time sooner.
that's just a thought though.

yet this case doesn't necessarily means that the CPU has deteriorated,
it can mean many things and worth being tested.
finding the real cause might give you true peace .