Does OCing really shorten CPU lifespan?

[Ol'Tunzafun]

I am considering buying an XP 1600+ that I know has been seriously overclocked (high voltages with good cooling). My question is, if it works fine now and I am going to be running it in the throttled back area of 1400 - 1600 (it'll do 1800 at default voltage), can I not reasonably expect it to live a long time?
I have never had a processor die of old age, even some that I have run 40% overclocked for long periods of time. The thing is, I don't play with extreme voltages and I don't know how that is going to play out.

[Benssax]

I think that upping the voltage to much is what shortens the life.

[Rugor]

I wouldn't want an overvolted chip myself. It does lead to things like electromigration, and it means the chip was run hot..

[causticVapor]

When I posted a theory about CPU lifetime here, people made a huge joke about it.

Generally, running a chip at a higher speed, no matter what voltage, will lead to faster electromigration because its transistors are switching more rapidly (hence working harder). Overvolting can lower MTBF by an exponential value, and as transistors are made smaller they become more sensitive to overvolting.

+.4v seems to be what .25-micron chips will handle, +.25v is OK for .18-micron chips (Like your palomino), and you don't really want to go over .2v with a .13-micron chip. Athlons distribute heat more evenly than P4s do (same-speed everything), so are less vulnerable to overvolting. However, their idle -> active heatspike is much sharper than that of the p4, and thus they require a heatsink with a thick enough bottom to buffer that burden.

And yes, for every ten degrees celsius higher a chip is run, its average transistor reliability is cut in half. This implies that the chip's useful lifetime is cut in half, say, from 10 years to 5 years, 40 degrees to 50 degrees. Inversely proportionate. :t

[tony_j15]

But as we pointed out before, who keeps a chip for longer than 5 years? By that time you will be ready to upgrade. I dont have any problems with OC'ed chips, but would reccomend buying a new one anyway.

[Rugor]

I know overvolting seems to play a role in Sudden Northwood Death Syndrome, and this would certainly help explain that.

And In this household we do use chips until they die. We've still got free ports, let's add another box to the lan. Get the kids up to one system per not two between 3 kids.

[AllGamer]

By the LAW of Physics: YES
By the LAW of OC-ers: NO :p

here's the run down

the more you OC-ed, the HOTTER it gets

the more heat, the faster the electronics and MATTER in simple terms, expands and designtegrade faster

but by applying Proper Cooling

then it "SHOULD" slow down the premature aging due heat

:t

[Rugor]

Yes, proper cooling will help mitigate the effects of heat. There's no doubts about that. BUT, proper cooling won't help with all the effects of running more current through the chip than it's designed for.

That's what seems to be doing the killing, running more current through transistors that are burning out under the switching load.

I'm not surprised that newer processors are more subject to overvolting. They run on much lower voltages to begin with, so the smaller absolute jump is as large or larger proportionally.

[Ol'Tunzafun]

Are there any tell-tale signs, performance-wise, of a chip that has been over-volted?
Are they slower or unstable or are they just more fragile than a chip that has never been OC'd?

[BFlurie]

My old AMD K6-2 300 2.2V ran @ 375, 2.5V for 4 yrs w/o a hiccup. Benchmarks never varied. It's retired now 'cause of upgrades, but still works. Always had good cooling.

[BiG MiKeY]

they dont make chips lyke they used to :rolleyes:

[causticVapor]

No, the effect of the same current on smaller transistor sizes is exponential. Ask any EE, they will tell you the same thing.

The smaller they get, the more careful you have to be with them. They were already invisible at .35-micron. Now we are at .13 and migrating over to .09. Only a few diameters larger than your average bacterium.

Also, people are expecting more of an overclock with modern chips. One could get about 100-300MHz more out of a .25-micron chip in the old days if they were lucky. Now we consistently attempt to milk 500-700MHz out of select chips and lambast them if they can't do it. The same people who ran around, jumpy and scared about putting 2.8v into a K6-2 seem to have no qaulms about running T-breds at 1.9v.

1.5->1.9 26.7% overvolt
2.2->2.8 27.3% overvolt
:rolleyes:

But as we pointed out before, who keeps a chip for longer than 5 years?

As I pointed out before, the vast majority of people. :rolleyes: :p ;)

[tony_j15]

One could get about 100-300MHz more out of a .25-micron chip in the old days if they were lucky. Now we consistently attempt to milk 500-700MHz out of select chips and lambast them if they can't do it.

That is because they are perfectly capable of doing so. With the .25 process, you would only get 100-300 because of the large micron size. The smaller we make them, the faster they go, and the more OC capable they are.

One of the major points of OC'ing is living on the edge! If you cant take the pressure, you have no business doing it!

[morpheus kain]

You know if we worked out some formulas and such for electron migration and the like we'd be millionares.... In the end there are a dozen factors that determine the lifespan of a chip so oc'ing my take off a week or it may half the lifespan. OC'ing has risks that are managable so if it's down well then why not eh?

[causticVapor]

Originally posted by tony_j15


That is because they are perfectly capable of doing so. With the .25 process, you would only get 100-300 because of the large micron size. The smaller we make them, the faster they go, and the more OC capable they are.

One of the major points of OC'ing is living on the edge! If you cant take the pressure, you have no business doing it!

We're putting more and more current through ever more sensitive transistors. People expect higher and higher overclocks, and might damage the chips in the process. The same people who say something is wrong with those who can't get 2GHz out of a palomino are saying that all T-bred B 2100+s should reach 2.4GHz. Not that you are one of them, but how good do the folks who can't get 2GHz out of their AIUHB W49 feel?

[tony_j15]

Stop being so squemish. We can get higher OC's because of the smaller micron process. For someone who has the math figured out so well on overvolting, I'm suprised you haven't figured this out.

[causticVapor]

Stop being so squemish. We can get higher OC's because of the smaller micron process. For someone who has the math figured out so well on overvolting, I'm suprised you haven't figured this out.

That's not the point I'm trying to make. Yes, speeds are proportionate, since it is an exponential curve, BUT -- we can do much more with chips nowadays than we could back then, when only a few motherboards would allow a 40-50% overvolt, and chips could handle it for quite a long time. Now, we are faced with a dilemma: Chips can go at much higher speeds, but are more sensitive to voltage changes due to the smaller "micron sizes" (trace lengths) that you keep bringing up. And more and mroe people are finding out that their chips burn out in less than a year. Think about it. ;)

[BiG MiKeY]

ok how bout this xp2100 slightly overvolted (1.7 for stability, i can prob lower it now becuase i think the crashes were do to the overclocked video card in UT2k3, just got my 9500pro)@2700 speeds... cost 90$ xp2700 cost $280 .... guess which one i brought.. even if it last one year ill be happy cuz by then ill be able to buy a chip better then the xp2700 for the prices that the xp2100 cost me. even if i do buy the xp2700 in a year i have saved $100 total from not buying it at the start.

but my whole argument is useless becuase who knows wut the market will be like in a year, and if we still have the same sockets by then.

[Picard]

Electromigration requires a tad more than a fraction of a volt to amont to anything, like closer 12. Nasa may have lost proce4ssor due to electromigration, perhaps in one of there probes sent out during the seventies. I doubt anybody on Earth has had a processor in use long enough for the effects to ruin a cpu, even overclockers.

This figure of 5 years for a CPU to waer out due to electromigration is baseless. There is no eveidence whatsoever to back that up.

By listening to some of the people here, gate switching speed is a big factor. One could assume, based on this notion, that intels 3.06Ghz processor are gonna have a short lifespan and we should all be buying the slowest processor we can find. There is, of course, no truth to this, and no evidence to support it. Who wants to use there processors for decades anyways.