Truth about CPU degradation.

[bluezero5]

Alright, another write up.
this time I want to write about the truth about CPU degradation.

The commonly asked question I have seen time and again on this forum is: whether a CPU would 'degrade' if you overclock pass 1.XX Volts?
Now, before I answer that, lets go over some basic underlying:

Heat is known to accelerate the depreciation of CPU life. Vice-versa.
As heat increases electric resistance hence increases wear and tear.
(think of heat as increasing Friction in an engine, so gears grinds out faster.)

Now knowing this, lets take a look at the following chart:
credits to Anandtech:

this is a chart of required voltage against time.
3 cpu are charted. stressed/normal/cooled



edit:
* voltage editted per bobnova's comment to reflect a more generic situation.
* time line reference editted per ED's comment to avoid misconception of chip life.


While this chart's original intention is to show you why Intel has a 3 year warranty for chips, it tells us a few other things as well.

1, Intel's chip lifetime of 3 years, is calculated on 'stock speed, stock cooling'.
2, cooler temperatures will increase life. vice versa (as we pointed out above)
3, higher voltage = higher temps = shorter life.

So when you come down to it:
There's no such thing as a 'safe' overclocking voltage, as by definition, Overclocking IS decreasing a chip's life.
As whatever voltage increase you give, you are shortening the life by x amount, and there's no hiding from it.
So do know what you signed up for before you overclock.

NOW, moving on to more positive thinking:
Investing in a decent cooling system WILL increase your chip's life.
And this is a fact. Whether or not you plan to have a chip for 9 years is another story, but this is why if you have invested in a decent cooling, air or water, then you probably have already by definition, already extended your chip's life by X amount of months already. (as Intel's test are base on stock speed and stock cooling.)

So here comes the obvious Conclusion:
With a better cooling system, you will be able to maintain a suitable CPU health despite overclocking to moderation. Now, Intel chips have different test temperatures they use for each generation of chips, and you need to look it up, but as long as your system is running under this temperature on regular loads, then your chip's lifetime 'probably' won't wander 'too far off' from the original intended lifetime of a chip. Exceptions will be if you stray FAR from intended voltages, then your chip might suffer transitional damage, most obvious from the powering up of the system. Any chip will degrade given time, overclocked or not.

REMEMBER:
- CPU degradation simply means your CPU wants MORE voltage to reach a certain frequency than before, and this is BOUND to happen, one way the other. Fear not. This is why I encourage moderate overclocking if you invested in a decent cooling device, as 3 years down the road, you probably will want a different chip anyway.

HELPFUL HINTS TO MINIMIZE DEGRADATION:
1, power down your system when not in use. really. why power it up if you are one of those that use it for 2 hours and let it standby for 22 hours.
2, consider using some voltage regulating tech, Intel has plenty, putting the computer to sleep for example is one, speedstep another.
3, invest in a decent cooling system, I don't mean water necessarily, but a decent AIR cooler will easily add some months to your chip.
4, STRESS TESTS by definition is taking life away from your computer, do in moderation only.

good luck all.

 

[azanimefan]

good writeup about a phenomenon I've never seen on a CPU. Maybe I've been lucky, or maybe I just know how to cool them... who knows?

I have seen this on a GPU... several of them in fact (my current one for example).

 

[Archer0915]

Well my 2500K died last night Degradation? Hell no it was cooked! I did it! My fault!

I have never lost a CPU due to any type of degradation and I have run 20-50% OCs on things for ten years+.

I killed the thing on a memory review to boot!

But yes there is such a thing as degradation and if you want to run your CPU for more than its useful life (3-5 years) then don't OC much past the fastest speed of the flagship on the given process!

 

[trueblack]

can't help to feel that I am being a sales target of a better cooling system...

 

[Archer0915]

can't help to feel that I am being a sales target of a better cooling system...

Well better cooling helps and the cooler you keep it generally the less voltage you need to overcome the extra resistance created by the heat.

 

[bluezero5]

I personally believe in investing in a decent cooling unit is a very wise choice,
not only are most of them recyclable for your next built, when you use it,
you are basically extending your CPU 'by definition'.

while this extension won't be that necessary for those running on near stock speeds,
for an overclocker, a good cooling device is the saving grace for the sins of overclocking that we are guilty of.

 

[doyll]

Nice writeup. Thanks!

I use better cooling to have quieter system as much as for overclocking or life. Quiet is nice!.

 

[eskimochaos]

Could you argue that the thermal cycling of on and off usage is worse than leaving it constantly on?

 

[EarthDog]

Interesting article... from 2008. I have only experienced degradation at extreme voltages not used for air/water. That was one CPU. With any other, I haven't had that problem. So long as temps/voltages are under control, I wouldnt worry about it...

Could you argue that the thermal cycling of on and off usage is worse than leaving it constantly on?

Last I heard this was when I was in HS in the early/mid 90s... and that was because of the switches used... but was told to keep the PC's on. No idea on that one.

 

[Bobnova]

What Anand doesn't note is that the chart in question is made up. It is not the result of study. It is not from intel. It's something that they made up for their site.
Also, this is the actual chart, the one in the OP appears to have been edited for different voltages.

Here's the original, from the original article:

What the curves are supposed to represent is beyond me. The "cooler" temps are specified, the tcase measurement he talks about has squat to do with the core temps or the maximum rated temp for the silicon, too.
In my own very personal opinion, it's a junk article and best ignored.

Note that the article is about 45nm CPUs.
You know, those CPUs that a LOT of us ran at 1.4v 24/7 for years without issues.

This is much like Anands old LLC article with that made up graph, it got anandtech an awful lot of hits and started the whole "OMG OMG OMG OMG LLC IS BAD FOR YOU" craze. The craze that has yet to prove true, at all, ever.

EDIT:
Oh and my first 45nm Intel CPU had a VID of 1.225v, that's what it ran at 100% stock. Just a touch over where that graph says. By that graph, it should have died promptly even at stock, let alone at 1.4v and OC'd.

 

[doyll]

There may be a timing where staying on 24/7 is better than cycling off for a few hours and back on a few hours later and back off, etc.

 

[hokiealumnus]

I've never had a CPU degrade, from C2D chips to today's Ivy Bridge / Piledriver chips. There has been one chip that died (Bulldozer), which was a result of putting 2.3V through it because it wouldn't get to 8GHz and made me angry.

That's not to say degradation doesn't happen, I'm sure it does in some form somewhere, but it is not a large problem and as I was told when a n00b starting out worried about the same thing: The CPU will be obsolete years before degradation will kill it. Keep temperatures reasonable and your CPU will last long past the accompanying parts.

 

[Anonaru]

I've never had a CPU degrade, from C2D chips to today's Ivy Bridge / Piledriver chips. There has been one chip that died (Bulldozer), which was a result of putting 2.3V through it because it wouldn't get to 8GHz and made me angry.

That's not to say degradation doesn't happen, I'm sure it does in some form somewhere, but it is not a large problem and as I was told when a n00b starting out worried about the same thing: The CPU will be obsolete years before degradation will kill it. Keep temperatures reasonable and your CPU will last long past the accompanying parts.

Hahahaha, I actually laughed out loud at that little bit about blowing up your BD chip

I am keen on keeping the parts as cool as possible-- this is common sense. I also have had older parts on / off OC'd from moderation to LN2 levels for YEARS.. some even close to as long as these graphs run for.. The only noticiable degradation I've seen in computing has been in motherboards, PSUs, and graphics cards. Never experienced anything like this in a CPU.

Edit: I was wondering about those insane VID values on that graph. Thanks for posting the corrected graph, "Bobnova"

 

[bluezero5]

What Anand doesn't note is that the chart in question is made up. It is not the result of study. It is not from intel. It's something that they made up for their site.
Also, this is the actual chart, the one in the OP appears to have been edited for different voltages.

Here's the original, from the original article:
View attachment 121177
What the curves are supposed to represent is beyond me. The "cooler" temps are specified, the tcase measurement he talks about has squat to do with the core temps or the maximum rated temp for the silicon, too.
In my own very personal opinion, it's a junk article and best ignored.

Note that the article is about 45nm CPUs.
You know, those CPUs that a LOT of us ran at 1.4v 24/7 for years without issues.

This is much like Anands old LLC article with that made up graph, it got anandtech an awful lot of hits and started the whole "OMG OMG OMG OMG LLC IS BAD FOR YOU" craze. The craze that has yet to prove true, at all, ever.

EDIT:
Oh and my first 45nm Intel CPU had a VID of 1.225v, that's what it ran at 100% stock. Just a touch over where that graph says. By that graph, it should have died promptly even at stock, let alone at 1.4v and OC'd.

thanks for pointing that out.
however, the reason why I do not think the graph is bad, is because it represents an idea, not data. And the idea is the first-order decaying of silicon based transistors, which is a physical/chemical reaction, and common sense to us in the scientific field. (I am a nuclear physicist) That graph simply put voltage on the Y-axis instead of percent remaining, and that's why the chart goes up instead of down. That is why though the graph is 'made up', it has very good credit, it is not from Intel, but it is from common sense of how a silicon based product reacts under electricity.

I haven't read the LLC article you pointed out, but I will google it and read it a bit and see what it says.

CPU degradation wise, this graph is a Good model, and it represents an idea, not actual data. but guess what, if you collect a sample of 3x 100 chips, runs them through different voltages and environment, guess what, I will wager you will get a graph very similar to the above, thus how science works. However as leader of the benching team your experience will be priceless input as well, do let us know if you have had any signs of CPU degradation for running extremely high -temps- for an exceeding long period of time.

truth remains as:
- adding more voltage = shortening life of chip as more heat = more resistance as well
- better cooling = a good way to keep a chip's life extended.

edit:
I had a CPU degrading badly on me, but that was YEARS ago before OC was even that mainstream (makes me feel old), and of course I was running WAAAAY above the designed voltage, and what I had left after 3 months, was a CPU that needs 1.3XV (sorry forgot exact.. but you get the idea) to run at stock, so CPU clearly will degrade, even older chips as the underlying science has not changed at all. However, I DO WILL SAY, Intel chips are actually sturdier than what most people 'assume' of them. I have been torturing my 3930k for 11months now and still maintaining 90% of it's function, (notable is that I need LLC on high instead of medium now, else I can get some low CPU load crash that never happened before.), and for my chip most of the time running 24/7 at 1.52Vs, but that's also cause I keep the temp under 65'C most of the time too. I am quite OK if the chip burns out, if it does I will just replace it with something better.

 

[Archer0915]

thanks for pointing that out.
however, the reason why I do not think the graph is bad, is because it represents an idea, not data. And the idea is the first-order decaying of silicon based transistors, which is a physical/chemical reaction, and common sense to us in the scientific field. (I am a nuclear physicist) That graph simply put voltage on the Y-axis instead of percent remaining, and that's why the chart goes up instead of down. That is why though the graph is 'made up', it has very good credit, it is not from Intel, but it is from common sense of how a silicon based product reacts under electricity.

I haven't read the LLC article you pointed out, but I will google it and read it a bit and see what it says.

CPU degradation wise, this graph is a Good model, and it represents an idea, not actual data. but guess what, if you collect a sample of 3x 100 chips, runs them through different voltages and environment, guess what, I will wager you will get a graph very similar to the above, thus how science works.

Sadly the representation is not backed up with tangible data.

 

[Bobnova]

Without known scales, the graph is of dubious use. Maybe the curve is perfect if you take the current X axis values and multiply by 10. Maybe it takes 1.5vcore for any degradation to start at all.

Why'd you edit the vcore labels?


From what I've seen it's been the averaged experience of benchers that the level that will cause degradation in a meaningful time frame is just a touch under what kills it outright, rather than it degrading slowly even at stock clocks.
Tell that to the chips that have been in service for a decade at stock clocks on the stock cooler and are still in perfect condition.

 

[doyll]

Many times we don't need tangible data to apply common sense... like it's not a good idea to step in cow pies, especially wet ones.

Happy New Year to you all!

 

[azanimefan]

Sadly the representation is not backed up with tangible data.

I'm gonna have to agree with you. I've beaten chips bloody. Really the abuse i've put a few through is the reason why these companies say OCing voids the warrentees.

I've blown every part of a computer, including a cpu. Yet i've never seen performance degridation of any type in a CPU. And i've used some old and beaten eq.

I have seen it in GPUs... but then they run at much higher TDP, and usually are inadequately cooled stock. Combine that with factory made oc software, and I've come across a few that have been thoroughly destroyed by the power driven through them.

Perhaps the science is correct. I won't argue it... i mean I've seen this phenom with GPUs, so there is something to it. I've just never seen this with CPUs.

 

[bluezero5]

Sadly the representation is not backed up with tangible data.

we all know you shouldn't eat roadkills, you don't need a study with data to back up what is common sense.
but yes, the chart represents an idea, but a trail.

Without known scales, the graph is of dubious use. Maybe the curve is perfect if you take the current X axis values and multiply by 10. Maybe it takes 1.5vcore for any degradation to start at all.

Why'd you edit the vcore labels?


From what I've seen it's been the averaged experience of benchers that the level that will cause degradation in a meaningful time frame is just a touch under what kills it outright, rather than it degrading slowly even at stock clocks.
Tell that to the chips that have been in service for a decade at stock clocks on the stock cooler and are still in perfect condition.

the graph is meant for an idea. maybe I should have made the voltage labels as 'low' 'medium' 'high' instead. I only relabelled it so it represent a voltage more recognized today, I see what you mean though. (I will amend that in a bit.)

for your above CPU study case, are we talking about a 24/7 station? or a benching station? At benching levels I too will expect any 'observable' changes to happen only at what can 'kill a chip', as the time we are measuring is in weeks instead of years. Will your experience predict a 24/7 system to actually run 'fine' with voltage 'under levels that will outright kill it' though? I do value your opinion here and I am sure the forumer wants to hear the leader speak too.



edit:
Bobnova, I editted the labels above, cause your point is correct indeed, without 'actual' data, it is best to use theoretical axis instead.

 

[Archer0915]

Many times we don't need tangible data to apply common sense... like it's not a good idea to step in cow pies, especially wet ones.

Happy New Year to you all!

CPU clocking is about many things and common sense as the normal person sees it is not applicable.

If you have to worry about degradation you don't need to OC.

Common sense says you don't drive everywhere at 100MPH in a 35 zone as well.