Overclocked CPU lifetime?

[I]

ummm who here is using the same CPU as they were in May 2002?

Almost everyone. Most people don't upgrade their systems every 2 years. Most people do email, web surfing, office, etc, not high end gaming or attempt to reach some new app benchmark score.

In other words, you'd be amazed at how many people have no desire other than to keep their PRE-2002 system running at the lowest replacement-cost possible if/when if eventually fails in some way. If it was built with quality parts, it hasn't yet failed in most cases.

 

[I]

To the OP,

First do yourself a favor and pretend you don't know anything (because for practical purposes, you don't).

Plenty of people who had (for example) Celeron 300 o'c to 450, or Celeron 500 o'c to 750 or beyond, still have their systems running fine. Of those who had a failure it wasn't the CPU in the vast majority of cases, it was motherboard or PSU.

Point is, if an overclocked CPU is kept cool enough, I mean reasonably so, not extremes, it is still one of the last points of failure in a system. All evidence points to this, there is no reasonable argument nor discussion about it.

On the other hand, if in order to o'c, you put a large strain on the motherboard, beyond the designer's intentions, it will effect mobo life, and if it fails in certain ways it "might" harm the CPU in the process, but usually it would not.

In summary, do not think about CPU lifespan until you have ANY indication it is relevant. Statistically, you have no reasonble justification for your post, it is just a sign of failure to do prerequisite research instead of burdening others.

 

[lexluthermiester]

Sixth, Electron Migration does NOT equal current! Get that misinformed nonsense out of your mind.

No its current and heat, and electromigration effects all ic interconnects ,they went from aluminum to copper because of the ionization of atoms that created a broken circuit after heat and current wore them down. Copper interconnects wear better but dont solve the problem.

Today we have warm room temps driving cpu's up to 70 c add overclocking to that with voltage hitting 1.5-1.6 on a mod with wattage going moderately over 100-125, and damage is pretty much assured over the long haul.

It is an industry focus to find better ways of routing such inherent detriments in chips. If you have never seen copper ionized from heat and current it get pretty brittle and pourous and finally gaps.

You've made excellent points that lend validity to the point I was making. Heat combined with to much voltage will wear a pathway thin and over time cause it to fail entirely. I have seen both aluminum and copper in this kind of failure. Gold and depleted uranium[and a few other ultra-high conducting metals, can't remember which ones...] used in combination with S.O.I. methods are the best candidates for solving the problems electron migration presents to the I.C. industry. Keeping your system cool and limiting the voltage applied is the best way ensure your system stands the test of time.

 

[chiendoux]

If i may provide an example, i have an Athlon 4200+ running at 2.80ghz at 1.55V. How long can i expect the processor to last?

My temps tend to be mid thirties idle and mid fifties load depending on room temp. If anyone has been running an overclocked x2 for a while, advice is very welcome!

 

[Grimmy]

I would guess that it would still outlive all other components when cooled properly.

Since I still run a P4, I know the northwood cores will die, if ran over 1.7 volts. The S.N.D.S or Northwood Sudden Death Syndrome.

I suppose any CPU will have its tolerance on the vcore, but if your too greedy to push it to its limits, you should expect it to die eventually.

As long as you know your vcore limits (and keep it under that limit) for that 4200+, you should expect it to last a long time.

 

[ZOldDude]

Heat and overvolting are what kill electronics.

I run Optrons OC'ed up to 50% on stock volts (MB and RAM @ stock volts as well) with Thermalright coolers and simple case mods for the Coolermasters (remove exit fan guard for 70% more airflow).

Even with the OC's everything runs -cooler- than stock and the voltages are not changed.
They should run for years after the warrenty runs out.

EDITED: The guy who said that dust and cap failure in PSU was the main cause of a systems death was correct. I have many XP systems that are now on their 3rd PSU's.
In fact OC'ed or not....the #1 cuase of failure in a otherwise healthy system is from bad caps.

 

[Mondoman]

...
First, The temperature difference between stock speed and the OC'd CPU in question is 2C. Hardly anything to be concerned over.

I agree. However, the increased current may be a bigger problem.

...
Second, Since there seems to be some confusion, let me explain the difference between "electron migration", "voltage" and "current". Current is a measurement of strength of the flow of electrons through a pathway.

If by that you mean a measure of the number of units of charge moving past a point per unit time, I agree.

... Voltage is a measurement of speed of electrons through a pathway.

No, voltage is a measurement of the electrical potential difference between two points.

...Electron Migration defines the process by which electrons "migrate" from their intended and or designed pathway to other areas of the device.

I believe you're referring to "electromigration", which involves the movement of atoms (not electrons) due to momentum transfer from the flow of electrons in a current. "Electron migration", being the movement of charge, *would* technically seem to fit the definition of "current," although I did make that statement in jest. Sorry for leaving out the smiley.

...
Third, Increasing the speed of a CPU only increases it's frequency[measured in hz], not it's current[measured in amps].

You may wish to inform Intel and AMD of this, as it would seem to contradict their strategy of reducing CPU power draw by slowing the CPU clock.

...
Fourth, Electron Migration is present in ALL electric devices regardless of type or source. The degree of persistence of electron migration depends greatly on the quality and type of materials used, manufacturing methods as well as environmental operating conditions.

Certainly electromigration exists wherever a current flows. Again, I'm not sure exactly what you're referring to by "electron migration" -- perhaps you could provide a link to explain it? If instead you do mean "electromigration", I'm not sure what you mean by "degree of persistence" -- could you clarify that?

...
Fifth, Overclocking will not cause any long-term damage to any electronic device even if it's temperature raises slightly. It may cause malfunction, but not permanent damage. As long heat is given a method by which it may "bleed" away from the device in question and voltage is carefully and properly regulated, the device will last it's expected lifetime.

OK, I'm curious. If there's no continuing accumulation of damage that eventually leads to CPU failure, what DOES cause the eventual CPU failure (at the end of its expected lifetime)?

...By way of comparison, if I were to drop the voltage of the aforementioned Pentium3 to .9 volts and run it at whatever frequency that voltage would allow, the CPU would outlive us all.

If it worked at that freq, and current draw was substantially less than normal (which I would expect), I agree.

 

[dobby]

judging by the fact you oveclock, i presume that you proberly upgrade more than most "average" users, so i would say you will replace your CPU before it dies,

and heat or overvolting a CPU is the main reason of death these days

 

[I]

Catch 22 here :lol: while they look to keeping hold of aging systems out of cost most fail to do the simplest things to ensure longevity, Like blow the dust out. A common source of pc problem is heat from dirt build up or lint build up. Psu's that are matched to a system for an expectation of 4 yrs of use usually die sooner due to heat build up in dust piles near caps.
this causes voltage irregularity and further damage.
So where a cpu could last 20 years it likely will not in the vast majority of consumers houses.

If the day comes when we try to use systems for 20 years, at that point thought must go into equipment construction that uses longer life parts, as with your example the elimination of many electrolytic caps. Presently the perceived consumer demand is not for this, so today's capacitor issues are just a matter of corner-cutting to make cheap equipment. Similarly so with fans, but if a person doesn't clean the dust out of their system for 20 years, we might say they were asking for it since that is a fairly obvious thing to maintain over time just as emptying the vacuum cleaner bag or changing a furnace filter. A user could through negligence or overt act, kill any part in a system but practically speaking we would have to assume at least a system that was otherwise viable, not some random event that does against prudent use and maintenance of a system. Similarly so with fans, if the fan is over 10 years old (even 2-3 years with some of the worst fans) it is time to replace it instead of waiting for a failure, or at least relube it if sleeve bearing type.

You could claim a dying part kills a CPU but for practical purposes we have to conclude it is not a factor because the system had already failed, it matters not whether CPU is a cascaded failure at that point. Further it would be unlikely one would source a new motherboard for that 20 year old system.

However, it is not reasonable to conclude cap failure from any cause will likely kill a CPU. I have done quite a few motherboard et al repairs of failed caps where there was no processsor damage. Further the regulation circuits on anything remotely modern (after the linear supply era ended around 1996) shut themselves down if critical thresholds aren't maintained. A very severe failure like a 'fet heating up, desoldering itself then sliding down a vertical board due to gravity might cause it, but such things are quite rare. So far the majority of the evidence is a board or PSU failing but the CPU did not.

 

[rockyjohn]

ummm who here is using the same CPU as they were in May 2002?

I am too - seems like a lot of us. I have a newer computer but use my old XP1400 for surfing. I figure if it ever gets an infection I have difficulty removing, I can toss it then. In the meantime, I seldom connect my new rig to the net except trusted sites.

I plan to always keep two rigs - one for the internet without any sensitive or significant info on it (nothing that will do indentity thieves any good) and one - the newer, faster one - for other basically secure uses with my important documents and data on it.

 

[senor_bob]

I believe you're referring to "electromigration", which involves the movement of atoms (not electrons) due to momentum transfer from the flow of electrons in a current. "Electron migration", being the movement of charge, *would* technically seem to fit the definition of "current," although I did make that statement in jest. Sorry for leaving out the smiley.

That does more than seem to meet the definition of current, that IS the definition of current. Current flow is always through the movement of electrons (- charge), if the protons (+ charge) were moving that would be electromigration instead of current, if the whole nucleus of protons was moving at once. If the protons were moving one by one, that would be a nuclear reaction and NOT something you'd want in your computer! In semiconductors, the carriers are either electrons in the n-type semiconductor, or "holes" in the p-type semiconductors. The "holes" appear to be a positively charged current carrier, but they are really missing electrons, and so the current still results from electrons moving to "fill in" the "holes".
While I would not always treat Wikipedia as a valid source, the article on current is actually correct, so everybody should read up before making incorrect statements (not directed at Mondoman whose statments have been accurate):
http://en.wikipedia.org/wiki/Electric_current

Also, the power drawn by a CPU is largely related to the charging and discharging of the capacitances of the component transistors. The power draw will therefore have an equation something like P = K * freq * voltage^2, where K is a constant dependent on the number of transistors and construction detail. So you can see that the power draw will go up approximately linearly with frequency and with the square of voltage, which is one of the reasons it's good to keep the CPU at or below default voltage. This is also why Speedstep and Cool'n'Quiet lower the operating voltage and the clock frequency together to save power.

 

[aoenate]

My mom had been using an 800mhz P3 overclocked to 900ish mhz. After about 8 years the computer repeatedly hangs on boot up. once a restart is issued the bios screen is displayed saying the system hung due to an improper CPU settings.

Maybe the processor dieing is due to the added strain of the overclock, but, then again its 8 year old technology. If your still using that old of stuff its a good reason to upgrade.

As long as you can keep the temps low and voltages at or near stock i dont think you'll have a problem for quite a while. By the time it does fail, time for a new computer anyways 😉

 

[Mobius]

Actually, heat is not what kills CPUs; it is heat CYCLES. Each CPU is born with a certain number of heat cycles. If you heat it more, you get less cycles (on and off, of course) but only if you keep turning it on and off. Leaving it running all the time improves lifetime.

However, most CPUs are rated to about 15,000 hours or thereabouts, and heat cycling an overclocked CPU can drop this number somewhat.

The issue isn't that simple though: as Ion Migration will kill a CPU too. As an example of what happens in migration, my old Athlon T-Bird ran 60% overclocked (1.6GHz OC - 1.0GHz stock) for a period of over 3 years, and constantly running in that time.

I stored my original benchmark scores from when I first got the thing, and OCed it. After 3 years I felt the CPU was slowing, and testing it again at the OC speed resulted in test results about 25% lower than when first OCed.

This is easy to explain: migration has destroyed many hundreds of thousands of transistors (millions?) but fortunately, CPUs have a lot of built in redundancy - they'll keep operating (albeit at a lower performance) even when large portions of teh core no longer operate.

I remember reading a great article years ago about some guys who started burning holes through a CPU with a laser (in one side, out the other) and they had to kill over 40% of the CPU core before the core refused to boot. Performance dropped every time they started it again.

 

[Grimmy]

Heh, I wonder if there should be a thread on Electro migration failures on OC CPUs, like I found for the S.N.D.S. Where its just a collection of what people had their settings on the vcore with idle/load temps, on a CPU that actually stopped working.

Guess time will tell when people start complaining C2D start failing from OC's... if they keep them long enough. :lol:

 

[Mondoman]

Actually, heat is not what kills CPUs; it is heat CYCLES. Each CPU is born with a certain number of heat cycles. If you heat it more, you get less cycles (on and off, of course) but only if you keep turning it on and off. Leaving it running all the time improves lifetime.

This sounds quite dubious to me. Do you have some references?

...
This is easy to explain: migration has destroyed many hundreds of thousands of transistors (millions?) but fortunately, CPUs have a lot of built in redundancy - they'll keep operating (albeit at a lower performance) even when large portions of teh core no longer operate.

Again, this is hard to believe - references?
.

 

[true_hw]

I am glad that Senor_Bob actually solved the mystery.
Senor_Bob wrote:
The power drawn by a CPU is largely related to the charging and discharging of the capacitances of the component transistors. The power draw will therefore have an equation something like P = K * freq * voltage^2, where K is a constant dependent on the number of transistors and construction detail. So you can see that the power draw will go up approximately linearly with frequency and with the square of voltage, which is one of the reasons it's good to keep the CPU at or below default voltage. This is also why Speedstep and Cool'n'Quiet lower the operating voltage and the clock frequency together to save power.

The moral of the story:
increasing the voltage = increasing the power draw
increasing the clock frequency = increasing the power draw
Finally, increasing the power draw will result in more heat.
Is any one of these more significant than the others? Yes, it is. As Senor_Bob wrote, the power draw will go up linearly with frequency and with the square of voltage. In other words, increasing the voltage is much more risky than increasing the clock frequency, but both will shorten the lifetime of the CPU.

 

[BustedSony]

First, long-term failure of semiconductor devices is not due to electromigration, there is no "Copper path breakdown,' but CONTACT migration, as VernDewd correctly pointed out. The junction between the silicon P and N layers is changed over time as a barrier is built up due to modification of the atomic structure at the boundary. In analog devices this causes a gradual change in back-conduction and gain. In digital devices the change is inconsequential until something dramatic happens, thus it would be hard to EVER find a CPU or switching diode that has failed in regular proper use. it would usually be due to considerable overvoltage rather than heat. Heat causes overall damage to the material, but does not affect contact migration.

Two, there is a very simple way to look at why faster semiconductor switching devices take more current. Even something as sophisticated as a CPU is really just a bunch of TTL gates. A Gate draws very little current when it is "on" or "Off," - a "One" or "Zero," but draws considerable current while CHANGING state. Thus a faster CPU, or one that is busy, takes more current and is running hotter simply because it has more gate transitions in a given time. And yes, extreme underclocking of serial switching circuits can also cause harm because the transitional states are lasting too long. Memory can fail in PC Ram or other digital store device if the clock is removed, because the gates get stuck in an indeterminate state drawing massive current and overheating. This is the most common destructive failure in digital video timebase correctors, the clock to the memory fails and the the chips all get fried.

I should add that gate saturation also increases current. The supply voltage is enough over transiton voltage range that the change is on the safe part of the switching cure. Since it takes time to make the change (the Slew Rate) the voltage must be increased to assure that the change occurs within the clock cycle. An increased saturation range also increases resting current and heat. It is that required increase of saturation range to assure full transition switching when overclocking that causes an overvolted CPU to heat more, not so much the increase in speed itself.

 

[Ssoulrunner]

I thought this was a forum to discuss hardware for computers. No disrespect guys, however I don't believe any of us want to read about your lovers spats. Sounds like you should both grow up. Oops, maybe this is Opra!!!!

 

[Ssoulrunner]

 

[BaldEagle]

ummm who here is using the same CPU as they were in May 2002?

How about 1997!!!

Cassini

Or 1977 8O 8O 8O

Voyager

"The Voyager spacecraft have three RCA 1802 CPUs running at 6.4 MHz. "

 

[BaldEagle]

Server double posts again

 

[warezme]

So we all know overclocking will shorten the lifetime of a CPU. But by how much? What is the non-O/Cd lifetime of a CPU anyway? What is the main factor reducing the lifetime, frequency-increase, voltage-increase, heat-increase? Is any one of these more significant than the others?

As a concrete example, while it appeared to work at 3.2GHz, I am now running my E6600 at 3.05GHz with no voltage increase, EIST enabled and junction temperatures staying under 50C at load (56C when using TAT). Will this "conservative" overclock preserve most of the lifetime of my CPU?

Of course very few of us will still use the same CPU 5 years from now. Any margin above that will probably be meaningless to most people. For the sake of this thread though I'd still be interested if a moderate overclock reduced the lifetime of the CPU from, say, 10 to 5 years for example.

pretty cool voltages man, I think yours will last longer than you will care to have it...,

Me..., If it lasts till Penryn quads come out, I'm happy, so BURN baby BURN, yeeehaaaawww!!!, kick the tires and light the fires.... :lol:

 

[I]

Many of these posts are just such utter nonsense. Many many people WILL use the same CPU 5 years from now. Today's current-gen performance levels far exceed what the average person is doing with their computer. Most people just AREN'T avid gamers or do anything in particularly that prompts them to upgrade every 2-4 years. They use the system until a failure (and some tech's advice) encourages them that it is more cost effective to upgrade/replace than just repair to keep what they had.

There is no significance to those who write about electromigration. They're just parroting what they read because there is no evidence this factor is relevant within the viable lifespan of a system. This includes overclocked and overvolted CPUs so long as two factors are present:

1) You aren't overvolting per the voltage required - giving it only the voltage it needs to run stable within the ceiling possible per the design and that while #2 is also true.

2) It was kept cool enough, like under 65C. No extreme attempt to reach some arbitrary contest-winning low temp is required. It just has to stay cool ENOUGH.

If these factors are true, you have no reasonable concern about CPU lifespan, you will have other component failures first. It's just silly how everyone focuses on the CPU as if that's the part sensitive to aging and temp. Please learn at least a bit about electronics before chasing ghosts because citing some eventual failure mechanism is pointless until there is the context of what the OTHER system failure mechanisms significant enough to abandon the system, are in relation.

 

[Fat4l1ty]

So we all know overclocking will shorten the lifetime of a CPU. But by how much? What is the non-O/Cd lifetime of a CPU anyway? What is the main factor reducing the lifetime, frequency-increase, voltage-increase, heat-increase? Is any one of these more significant than the others?

As a concrete example, while it appeared to work at 3.2GHz, I am now running my E6600 at 3.05GHz with no voltage increase, EIST enabled and junction temperatures staying under 50C at load (56C when using TAT). Will this "conservative" overclock preserve most of the lifetime of my CPU?

Of course very few of us will still use the same CPU 5 years from now. Any margin above that will probably be meaningless to most people. For the sake of this thread though I'd still be interested if a moderate overclock reduced the lifetime of the CPU from, say, 10 to 5 years for example.

I believe Heat is still the main murderer, how i wish theres Artificial Intelligence programmed into every CPU, 5mins before it dies, it will alert you and shows a message, 'I'm dying in 5 minutes time, Please save all your work and shut down to prevent loss of data.' 😀

 

[FH]

It's just silly how everyone focuses on the CPU as if that's the part sensitive to aging and temp.

There is a very simple reason: Myself and others are running the CPU out of spec, not any of the other parts. My motherboard for example is rated by Gigabyte, if not the chipset manufacturer (Intel), to run at 333 FSB and the memory at up to 400 FSB. Thus I am neither overclocking the board (well, only slightly) nor the memory. Can you therefore understand my focus on what I might be doing to the CPU? It might be naive, but I'm sure not hard to understand? That said, I appreciate your perspective as well as that of others who have made similar posts.

Please learn at least a bit about electronics before chasing ghosts...

What's wrong with trying to learn from other people in this forum?