Truth about CPU degradation.

[bluezero5]

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.

how about common sense from a nuclear physicist?
cause I believe I will understand how matter behave under voltage more than the common folk. If everyone understands it, then I need not write down my views to share my experiences from getting a phD in physics in college.

Common sense says you don't drive everywhere at 100MPH in a 35 zone as well. you are right, and what if you do? then you run a possibility of living shorter than the rest, a perfect example.

 

[Archer0915]

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.




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.

In theory you are correct but in application things change. I am and have always been believer in degradation but combine we (OCF) have over? of experience.

I have a box full of old yet operational procs., boards, memory and vid cards. The only thing I can say I have ever seen what I suspect is degradation from is a GPU.

So because ever CPU is different by batch we can not assign any real metrics to it other than generalizations.

 

[Archer0915]

how about common sense from a nuclear physicist?
cause I believe I will understand how matter behave under voltage more than the common folk. If everyone understands it, then I need not write down my views to share my experiences from getting a phD in physics in college.

Common sense says you don't drive everywhere at 100MPH in a 35 zone as well. you are right, and what if you do? then you run a possibility of living shorter than the rest, a perfect example.

But being a brain surgeon or a physicist really has nothing to do with it. Yes there is degradation but where and when differ.

No two processors are alike.

 

[bluezero5]

In theory you are correct but in application things change. I am and have always been believer in degradation but combine we (OCF) have over? of experience.

I have a box full of old yet operational procs., boards, memory and vid cards. The only thing I can say I have ever seen what I suspect is degradation from is a GPU.

So because ever CPU is different by batch we can not assign any real metrics to it other than generalizations.

always appreciate a share of experience.
and that is what this thread is really all about.

What if I tell you I am seeing signs of CPU degradation from my 3930K now after 11 months of 1.52V? 1 month back, my computer that has been perfectly stable, started to have some 'low CPU load crashes', took me some time to find out, but it turns out I can no longer keep my LLC at medium, I need it at High for the same CPU config to run. Now, you can totally argue that it is my VRM that are degrading instead, and I won't have an answer for that until I take the whole thing down another year down the road. However, base on my experience with electric appliances and the level these VRMs are graded for, 1.52V should really be baby work for them. (though I have no prove until I take it all down, of course.) and I also had another CPU degradation experience many years back too.

like I said, Intel chips are sturdy, much more so than many assume they be. that's why there was a time when I was arguing if 1.45V will 'cook' your chip or not. to me, anything under 1.50V is really quite ok as long as you have the cooling for it.

But being a brain surgeon or a physicist really has nothing to do with it. Yes there is degradation but where and when differ.

No two processors are alike.

and you are right. the grapgh is a generic case and fact study only. (not case study)
I am making no claims on knowing when a CPU will degrade where and when at all.
as like you said, no chips are identical, from the same mold, you can in theory have a chip that is twice as durable than his neighbour.
but that doesn't mean you running a stronger voltage doesn't 'still' shorten its intended life.

that is what I was trying to say.

 

[Archer0915]

always appreciate a share of experience.
and that is what this thread is really all about.

What if I tell you I am seeing signs of CPU degradation from my 3930K now after 11 months of 1.52V? 1 month back, my computer that has been perfectly stable, started to have some 'low CPU load crashes', took me some time to find out, but it turns out I can no longer keep my LLC at medium, I need it at High for the same CPU config to run. Now, you can totally argue that it is my VRM that are degrading instead, and I won't have an answer for that until I take the whole thing down another year down the road. However, base on my experience with electric appliances and the level these VRMs are graded for, 1.52V should really be baby work for them. (though I have no prove until I take it all down, of course.) and I also had another CPU degradation experience many years back too.

like I said, Intel chips are sturdy, much more so than many assume they be. that's why there was a time when I was arguing if 1.45V will 'cook' your chip or not. to me, anything under 1.50V is really quite ok as long as you have the cooling for it.



and you are right. the grapgh is a generic case and fact study only. (not case study)
I am making no claims on knowing when a CPU will degrade where and when at all.
as like you said, no chips are identical, from the same mold, you can in theory have a chip that is twice as durable than his neighbour.
but that doesn't mean you running a stronger voltage doesn't 'still' shorten its intended life.

that is what I was trying to say.

Wow that sucks. The design of said CPU is a tad different and I suspect the IMC would be the first thing to go.

I just lost a 2500K the other night but such is life. It got me a few golds so all is well I guess

My honest theory is look at the manufacturers specs in the white papers and work from there. I have no doubt that you could have gotten a bad chip that could not handle the long term strain but what have you tried as far as trouble shooting?

 

[Knufire]

1.52 also seems like a crazy high voltage for a Sandy Bridge 24/7 OC...isn't that the exact voltage that Intel specifies as a maximum for SB?

IMO, that's the voltage range for SB where you will see degradation over a more noticeable time span.

 

[freeagent]

Ive ran high clocks with high volts and high heat for weeks on end with no ill effects on multiple cpus from 65nm down to 32nm. If you can keep it cool, there shouldnt be any problems as long as its a stable oc to begin with

A cranky cpu will tell you when enough is enough imo

 

[bmwbaxter]

I have personally seen degradation from overclcoking.

When I first got my 3930k it could handle 4.3ghz at 1.33v it now needs 1.37v for the same speed. Temps on the CPU never went above 71C, ran it on custom water. I was running SETI and Rosetta 24/7 for 6 months before my first BSOD. Once I upped the voltage it was back to normal. So the degradation wasn't the death of the chip but it was clearly there.

 

[larrymoencurly]

Why can't I find any information like this in any legitimate engineering or scientific publication or from anybody in the chip process business?

I thought the primary limitation to CPU lifespan was any internal EEPROM losing charge, well after a decade at maximum absolute voltage and temperature, even storage temperature.

Aren't the CPU cooler and the voltage regulator going to cause all the variations?

 

[Boulard83]

Thanks for sharing !

I also beleive in good cooling equal good component lifespan. Each time i build a system for a friend, even if he dont plan on OCing, i put a better cooling than the stock one. The last rig i built was a 3570k for a friend and i installed him a Enermax ETD-T60-TB. The 3570k stay cool and the rig is quieter ! This also gives better airflow to the ram and VRMs. Everyone is happy !

 

[EarthDog]

1.52 also seems like a crazy high voltage for a Sandy Bridge 24/7 OC...isn't that the exact voltage that Intel specifies as a maximum for SB?

IMO, that's the voltage range for SB where you will see degradation over a more noticeable time span.

And there is the story to me for this example. I mean, going over VID (im aware it varies...)by .25 is a lot. Commonly held advice for 24/7 voltage is 1.4-1.45 around these parts.

I believe it exists, there isnt a doubt in my mind in fact. However if voltages are kept low and temperatures in check, that commonly held advice will keep people overclocking and happy until they feel they need an upgrade.

I find the information valuable to know, and it is common knowledge to keep things cool and voltages in check. That I agree with, but the tone, for me, appears 'alarmist'. The way its written I get the feeling a CPU is only good for 3 years. Intel warranties their chips for 3 years as you mentioned in #1, however they dont all just stop working after 3 years. So they WARRANTY it for that long, it doesnt mean it DIES. That information is based off their warranty essentially which isnt how long the chip lasts (varies with TONS of variables).

 

[Anonaru]

And there is the story to me for this example. I mean, going over VID (im aware it varies...)by .25 is a lot. Commonly held advice for 24/7 voltage is 1.4-1.45 around these parts.

I believe it exists, there isnt a doubt in my mind in fact. However if voltages are kept low and temperatures in check, that commonly held advice will keep people overclocking and happy until they feel they need an upgrade.

I find the information valuable to know, and it is common knowledge to keep things cool and voltages in check. That I agree with, but the tone, for me, appears 'alarmist'. The way its written I get the feeling a CPU is only good for 3 years. Intel warranties their chips for 3 years as you mentioned in #1, however they dont all just stop working after 3 years. So they WARRANTY it for that long, it doesnt mean it DIES. That information is based off their warranty essentially which isnt how long the chip lasts (varies with TONS of variables).

This. Cars are backed for a predestined amount of time, but they almost always last longer. If you drive 10 over the speed limit on a questionable road all the time, you're not going to jostle your stabilizer bar and front bearings any sooner than you'd expect.

If you're a moron and take those same turns going 35 over while there's sand in your chassis' joints, then sure you're going to screw it over sooner.

I'm not the least shocked blue is noticing some strange issues with a 3930k running at 1.52, that voltage is where a LOT of people start witnessing mysterious problems, failures, fries, etc on SB chips. This is why I won't even break 1.4 on my own chip; we all have to consider the risk / reward for what we do, and where the best compromise between the two is. I'd never say that you can't fry your chip too quickly by overvolting or overheating it, but I agree this graph is dubious, and a very poor representation of the idea it is supporting.

If I made a graph showing the likelihood of outcome Z due to X and Y, then set my Y and X to obscure values that serve no other purpose than to exaggerate possible results, I've made (Edit: balderdash), for lack of better. Saying, "Cooling is good because of this bogus graph representing a very incorrect version of what we already know" doesn't carry nearly as much weight as Poster A saying "My Athlon X3 fried when I let it get up to 77*c" or Blue saying "At 1.52v, my 3930k is starting to depreciate in function." Food for thought

 

[bluezero5]

I have personally seen degradation from overclcoking.

When I first got my 3930k it could handle 4.3ghz at 1.33v it now needs 1.37v for the same speed. Temps on the CPU never went above 71C, ran it on custom water. I was running SETI and Rosetta 24/7 for 6 months before my first BSOD. Once I upped the voltage it was back to normal. So the degradation wasn't the death of the chip but it was clearly there.

my experience is closest to yours here.

I got 3930k last January.
started to run a 1.52Vcore on it at 5.0Ghz.
I now need a high LLC for it to not crash on low CPU loads, and use to be able to handle Medium LLC. also I realize my 4.8Ghz now needs 1.47V instead of 1.45V where it was 11 months ago. I believe this is a sign of CPU degradation as well.

1.52 also seems like a crazy high voltage for a Sandy Bridge 24/7 OC...isn't that the exact voltage that Intel specifies as a maximum for SB?

IMO, that's the voltage range for SB where you will see degradation over a more noticeable time span.

you are 100% correct, that's the reason I used 1.52V and I reversed to get highest clock which was 5.1Ghz then, so I dial it back to 5.0Ghz.
FYI on this now, my 1.52V refuses to run 5.1Ghz now, I need at least 1.545V to get it going at that freq. this might also be evidence of degradation. all in the name of experimental science.

Wow that sucks. The design of said CPU is a tad different and I suspect the IMC would be the first thing to go.

hehe, when I started to OC, I know what I signed up for, to be honest, 11 months with such minor degradation, I should actually PRAISE this CPU for doing all the hard work for me. For me, it was worth it, as my work is calculation heavy (I am a physicists, so lots of models to run.) and I will say my computer has saved me WEEKS of time with its cal speed, when this chip dies, I will Frame it up. thank you 3930k, then it will be HELLO 3970X lol.

this is turning into a nice thread, with many vets sharing experience, keep them coming, this is exactly the database we need.

I find the information valuable to know, and it is common knowledge to keep things cool and voltages in check. That I agree with, but the tone, for me, appears 'alarmist'. The way its written I get the feeling a CPU is only good for 3 years. Intel warranties their chips for 3 years as you mentioned in #1, however they dont all just stop working after 3 years. So they WARRANTY it for that long, it doesnt mean it DIES. That information is based off their warranty essentially which isnt how long the chip lasts (varies with TONS of variables).

I feel you bro.
let me change the language a little, and add a disclaimer to the time line as 'for reference example' only.

If I made a graph showing the likelihood of outcome Z due to X and Y, then set my Y and X to obscure values that serve no other purpose than to exaggerate possible results, I've made (Edit: balderdash), for lack of better. Saying, "Cooling is good because of this bogus graph representing a very incorrect version of what we already know" doesn't carry nearly as much weight as Poster A saying "My Athlon X3 fried when I let it get up to 77*c" or Blue saying "At 1.52v, my 3930k is starting to depreciate in function." Food for thought

I also see your point bro.
I use this graph for the idea cause it is in fact a reference to decay of material, I see this chart in chemistry reference often, and also in radioactive decay models, while the gradient of the slopes might seem arbituary, but again there's no reference for what temperature they are in, and a hotter chip will decay faster than a cool chip exponentially, and I will expect that's how the freq/performance will be affected as well, that's the reason why I picked this chart from amandtech. again, the graph is just an idea, not to be taken as 'actual' data.

 

[Bobnova]

Even just your experience of degradation, especially including the stock VID and the clocks you run plus time and calculation load is very valuable.
That's how we build a model

I have no doubt they degrade, my doubt was mostly A) in the graph and B) that they all degrade at more or less the same speed.
The CPU is as strong as the weakest transistor, and who knows where that transistor starts, you know?
Experimental physics vs theoretical

Along those lines, some people's 3770k chips are standing up to hours of 1.6v+ on air cooling, while others die promptly.

 

[bluezero5]

Even just your experience of degradation, especially including the stock VID and the clocks you run plus time and calculation load is very valuable.
That's how we build a model

I have no doubt they degrade, my doubt was mostly A) in the graph and B) that they all degrade at more or less the same speed.
The CPU is as strong as the weakest transistor, and who knows where that transistor starts, you know?
Experimental physics vs theoretical

Along those lines, some people's 3770k chips are standing up to hours of 1.6v+ on air cooling, while others die promptly.

heh, I hear you bro. Let me gather more data, if I have enough to generate a graph outta it, I will do that instead. However right now there's a large gap of data, let me browse around to see can I fill that gap. Meanwhile you are right, CPU is just as strong as the weakest transistor, and tough luck if one got a dull one and all hail the golden chip that lives forever. lol. I guess I can only offer 'average' expectations only, hehehe.

This thread alone we got some valueable info, keep them coming!
meanwhile, i will see can do a better chart, so the idea can be represented better.

 

[Archer0915]

Why can't I find any information like this in any legitimate engineering or scientific publication or from anybody in the chip process business?

I thought the primary limitation to CPU lifespan was any internal EEPROM losing charge, well after a decade at maximum absolute voltage and temperature, even storage temperature.

Aren't the CPU cooler and the voltage regulator going to cause all the variations?

Could you word that a bit differently?

 

[doyll]

Great read!

 

[doz]

Well, Ive been running my 2600k now for nearly 2 years (this January). Its been at 1.488v running 4.9ghz its whole life. I have good cooling (Noctua D14) and it never goes above 70c (usually 65-68c, depending upon ambient) during daily activities (gaming stresses the most).

Since then, I have not had to up the voltage at all. Its been rock solid the entire time. Everyone said when I originally posted my voltage that it would not last a year. Well, here we are 2 years later still going strong.

Ive ALWAYS been a firm believer that heat kills faster than voltage. Some have doubted this but Ive always gone by this theory and its never served me wrong. Ive never killed a chip on air/water due to voltage but Ive always kept my temps down.

 

[wingman99]

My First 2500k at 5.0Ghz 1.5v did not run stable at stock speeds and voltage after running overclocked for a short time. The one I have now has been running great for 2 years overclocked at 5.0Ghz, you just never know.

 

[trueblack]

i am surprised to see how many people running systems at 1.5V+
maybe time to try...?

also noticing here is that SB folks are fine, but SB-e folks seems some trouble.
this is reminding me of another article I read, that first gen chips are better than second gen. or maybe hex-core chips just pack too much heat given volumn of the chip about the same.

great read.

I think saying better cooling is better for the CPU cannot be wrong.