-
We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.
Is "degradation" real or a myth?
Page 4 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
There is no hard number. All chips are different. etc.
It also generally takes so long that there is no way in the world anyone could answer that now anyway (even for SB, maybe not even for Nehalem).
Any voltage will cause degradation, since the degradation is caused by electrons moving through the transistor. Higher voltages degrade the performance at a higher rate. This increases exponentially so that degradation increases at a higher rate than the increase in voltage. I believe that Intel designs their processors to run at spec to last 10 years under full load (although I don't know this for certain, but most companies I have worked for designed around the 10 year life cycle under worst case conditions for critical electronic devices.)
The equation to determine collector current is simple: Ic=Is[e^(eVbe/kT)-1] where Vbe is the base emitter voltage, Is is the saturation current, and Ic is the collector current. As you can see, even a small increase in voltage will increase current by a relatively large amount, which will in turn increase the rate of degradation. (Hopefully that makes sense, I typed this on my phone and equations are hard to type that way.)
That's the Ebers-Moll equation... and it's for bipolar junction transistors (BJTs)... CPU's use metal-oxide field effect transistors (MOSFETs) which use a totally different equation. From Wikipedia: http://en.wikipedia.org/wiki/MOSFET
As far as degradation, yes, it's definitely real.
A fairly straightforward explanation: http://www.anandtech.com/show/2468/6
More in-depth explanations:
http://en.wikipedia.org/wiki/Hot-carrier_injection
http://www.chironholdings.com/chirontechnology/references/App Notes/TDDB.pdf
http://en.wikipedia.org/wiki/Negative_bias_temperature_instability
And I could post some really good articles if people have access to IEEE Xplore - or you can do it yourself by searching for articles on "time dependent dieletric breakdown (TDDB)", "bias temperature instability (AKA "PMOS BTI")", "hot electron threshold reduction (aka "NMOS hot-E") which are the three main degradation mechanisms nowadays.
As far as calculations, they won't work because you don't know the cells that are prone to be affected on a given CPU - so knowing the current values or whatever doesn't help because you don't know the structures that will fail first. In simpler terms, you have a chip with a weak spot, but you don't know what it is. The only guys who could make an educated guess are the reliability engineers at Intel, and even then it's a matter of statistics. Meaning, even an expert at Intel couldn't look at an individual chip and say anything specific.
I will say that, like MrDudeMan, I work as a CPU designer at Intel and I look at some of the voltages that people say are "safe" and I think that I'd be worried about doing that with a chip that I am playing with on a tester (and thus costs me nothing if I nuke it... I walk over to inventory control and say "hey, can I have another?"). I'm an engineer and so maybe I'm a bit conservative, but my "safe" voltage for my i7-2600K on my home computer is a lot lower than most of what I read on here.
* Not an Intel spokesperson *
Most degraded chip I've seen is the e8400. Some statistics of e8400 owners at XS :
http://www.xtremesystems.org/forums...5-E8400-8500-degradation-myth-possibly-busted
Heat did not necessarily play a part in some of the degradations as many of the chips were well cooled by their owners, incl some with H2O.
My own e8400 in a second rig atm can barely go above stock speeds unless I crank up volts to more than before. As a consequence, I end to stay away from aggressive OC'ing.
http://www.xtremesystems.org/forums...5-E8400-8500-degradation-myth-possibly-busted
Heat did not necessarily play a part in some of the degradations as many of the chips were well cooled by their owners, incl some with H2O.
My own e8400 in a second rig atm can barely go above stock speeds unless I crank up volts to more than before. As a consequence, I end to stay away from aggressive OC'ing.
Yeah, I did that on purpose. I don't need some high-up guy at Intel showing up in my cube and saying "hey, we read that you told everyone on a super popular website that it's ok to take their CPU's up to <insert voltage>. How about we walk over to that conference room over there and explain to us how your paycheck is going to cover the returns from your great advice". And while I mean that in a lighthearted way, I can't imagine very many good things coming from posting my own educated guess on here. 🙂 I notice that MrDudeMan wasn't very specific either. You can email me if you want. My email should be visible in my profile.
Patrick Mahoney
Circuit Design Engineer
Intel Corp.
Fort Collins, Colorado
* Still not an Intel spokesperson. *
MrDudeMan
Lifer
Good post, pm.
I'd also like to point out that the equations are drastically different depending on the operating and bias conditions. This is far more complicated of an issue than most people seem to think. Consequently, there is a lot of misinformation floating around the internet because it's very difficult to say anything definitive without knowledge of the process, which is obviously confidential.
When you increase the voltage, you are changing much more than you realize. Pretty much everything is non-linear including resistors, capacitors, inductors, amplifiers, channel admittance, temperature effects, and the list goes on and on (textbooks actually do a pretty good job discussing these things, surprisingly). Transistors aren't the only things that fail, either. Increasing voltage may drastically change a bias condition that draws exponentially more static current through a metal segment or via. Current is a vector and both components (direction and magnitude) are important.
<removed some stuff - probably not sensitive information, but I'm not going to risk it>
- Not speaking for Intel -
Last edited:
That's the Ebers-Moll equation... and it's for bipolar junction transistors (BJTs)... CPU's use metal-oxide field effect transistors (MOSFETs) which use a totally different equation. From Wikipedia: http://en.wikipedia.org/wiki/MOSFET
As far as degradation, yes, it's definitely real.
A fairly straightforward explanation: http://www.anandtech.com/show/2468/6
More in-depth explanations:
http://en.wikipedia.org/wiki/Hot-carrier_injection
http://www.chironholdings.com/chirontechnology/references/App Notes/TDDB.pdf
http://en.wikipedia.org/wiki/Negative_bias_temperature_instability
And I could post some really good articles if people have access to IEEE Xplore - or you can do it yourself by searching for articles on "time dependent dieletric breakdown (TDDB)", "bias temperature instability (AKA "PMOS BTI")", "hot electron threshold reduction (aka "NMOS hot-E") which are the three main degradation mechanisms nowadays.
As far as calculations, they won't work because you don't know the cells that are prone to be affected on a given CPU - so knowing the current values or whatever doesn't help because you don't know the structures that will fail first. In simpler terms, you have a chip with a weak spot, but you don't know what it is. The only guys who could make an educated guess are the reliability engineers at Intel, and even then it's a matter of statistics. Meaning, even an expert at Intel couldn't look at an individual chip and say anything specific.
I will say that, like MrDudeMan, I work as a CPU designer at Intel and I look at some of the voltages that people say are "safe" and I think that I'd be worried about doing that with a chip that I am playing with on a tester (and thus costs me nothing if I nuke it... I walk over to inventory control and say "hey, can I have another?"). I'm an engineer and so maybe I'm a bit conservative, but my "safe" voltage for my i7-2600K on my home computer is a lot lower than most of what I read on here.
* Not an Intel spokesperson *
Thanks. It has been quite a while since I worked in circuit design (8 years ago this July) and I have forgotten nearly every equation other than the most basic in that time. I do remember seeing permanent changes in the transistors I tested (power gates, which would have large amounts of current rush through them to fire a squib and deploy an airbag. And also similar circuits but used to power a three-phase motor and those coukd draw north of 150 amps making testing difficult to test on a bench, since I had to design a circuit to generate that much current with a powersupply that could only supply 100 amps.)
I remember pumping so much current though the transistor that its ceramic shell would crack from the heat, and the traces would leave brown lines on the circuit boards (complete with that unmistakable burning silicon smell), and the transistor would still function. However the voltage needed to cause it to switch would always go up, sometimes by up to 30%. This wasn't a big deal because it was well within the margin, and It only had to go off once (for the airbag). I never once burned one so bad that the circuit acted like an open circuit.
I apologize for stepping in here, as I am far from an expert in CPU circuitry, but I really enjoy going back to my circuit design days and it was fun to try and relive some if that.
MrDudeMan
Lifer
You try playing with a 2 year old climbing all over you.
40-45c
MrDudeMan
Lifer
My daughter turned off the xbox as I was about to shoot. /derail
I'd never heard of it until you mentioned it. I'm reading up on it now. At home, I personally use IC Diamond. But this Indigo Extreme compound is interesting.
For what it's worth, though, I'm not much of an overclocking expert. I know why circuits don't work and I have a lot of experience with testing and debugging CPU's but the guys who are really good at overclocking seem to have a bit of a knack for setting things up. It's almost a bit like they have an instinct for it that I seem to lack. I see a sea of variables - there's so many possibility combinations of memory timings and votlages - and I start trying to be methodical and then there's just too many possibilities.
At home, I like to overclock because I'm a bit of a geek and I take personal satisfaction from running my CPU overclocked, but I also like it to be quiet and totally reliable so this limits how far I'm willing to push things. I'm running my 2600K at a fairly safe and boring 4.2GHz.
I will gladly tip my hat to you for remembering the Ebers-Moll equation... I couldn't hope to remember it. I memorized it for an exam way back when... but it's long since been forgotten. About all that I could remember of any of this is that the equation you posted didn't look right for MOSFETs. An airbag control circuit sounds like a really cool project... that must have been a lot of fun to get working... and to test.
Last edited:
I appreciate the compliment, but I looked up the equation myself. I just didn't notice it was for a BJT when I did.
As for working with squibs and three phase motors, I will admit that I had fun then. I spent most of my time in the lab building test circuits and testing components. I liked doing it myself rather than have the technicians do it, but I got to be good friends with them and they showed me some great tricks for alot of things. I once took a 6800uF electrolytic capacitor and hooked it up backwards to see what it would do. Luckily I took the precaution of putting it between a bench and the concrete wall becaue it sounded like a shotgun when it blew up and left shrapnel in the concrete wall!
Sp12
Senior member
Fair enough, but I'm curious what you run your specific cpu at, not what you feel the safe limit. My car 'can' go 120mph 'safely', and further than that 'unsafely', but what do I actually drive it at? As someone who doesn't work with circuits I'd be curious to hear what an expert would actually use.
If you're still not comfortable (which I understand) I can always PM.
taltamir
Lifer
Funny you would call it theoretical. The word theory means "the acceptable scientific explanation that accurately explains all existing evidence and experimentation and can be used to make accurate predictions".
You probably meant "hypothetical", which means "a guess as to what the result of an experiment will be"
The article in question does bring proof, those data graphs are not pulled out of the nether, but rather they are based on collected evidence by intel. Furthermore they form the basis of intel's warranty.
Since you value anectodal evidence so much. I personally witnessed degradation, every system OCed to the max will eventually require more voltage or few mhz to remain stable. On one system I have seen it go to the point of no longer being stable at stock.
Furthermore degradation occurs even at stock speed, in fact it occurs even if you undervolt, underclock, and overcool a processor. All three of those will greatly reduce degradation but as long as a current runs through it, degradation will occur.
red454
Senior member
So let's assume that OC'ing accelerates degradation. CPU manufacturers offer K series chips, unlocked for the purpose of... overclocking.
And it appears that the manufacturers are allusive with respect to stating the specific voltage / parameter limits with respect to OC'ing.
Sounds like the degradation must not be much of a factor (at least for mild overclocks) until past the warranty period.
And it appears that the manufacturers are allusive with respect to stating the specific voltage / parameter limits with respect to OC'ing.
Sounds like the degradation must not be much of a factor (at least for mild overclocks) until past the warranty period.
TRENDING THREADS
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 25K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes + WCL Discussion Threads
- Started by Tigerick
- Replies: 25K
-
-
-
