1.6v safe for e6600?

[Solarfall]

got thinking about this issue "is xxx amount of voltage safe.
can some one really say for sure that it is the high voltage that kills chips. what about heat?
usually high volts means high temperature at the same time. so the question is which one actually is the killer.
high voltage or the heat?
i would say that is has to be a combination of both of them that is the cpu killer? what do you gues think is my logic faulty? personally i would never increase the voltage over 1.6v with conroe, mayby if i would have some serious cooling at my disposal, singe stage or LN.

[nealh]

the issue is the combination of heat and voltage...

heat alone is bad..add extra voltage and you have reall issues

there was an intel engineer at Anandtech who wrote about the issue..voltage alone, heat alone can be damaging..both together can be a disaster...

[sreedx2]

the issue is the combination of heat and voltage...

heat alone is bad..add extra voltage and you have reall issues

there was an intel engineer at Anandtech who wrote about the issue..voltage alone, heat alone can be damaging..both together can be a disaster...

How long have you run that voltage in you sig for ?

I am running 1.55 bios 1.52 loaded at 3.6ghz

[nealh]

How long have you run that voltage in you sig for ?

I am running 1.55 bios 1.52 loaded at 3.6ghz

my setup has been up initially for 2-3 weeks ..then I had to RMA becasue I tried a sync mode and my board gave an unrecoverable(and I tried everything) C1 error

New board has been running 2 weeks...so total of 1 mo..non 24/7(weekdays typically from 5-6 hrs..weekends, 16- 48hrs

I think 1.52(1.51875 bios setting..bios reports 1.53v...load droops to 1.48/1.5v per SG) is probably safe on water for the amount of time I run the box

[GNU]

And if you cooled your heat CPU/NB (my NB has 1,72 24/7) with water that's maybe not bad or ?

[nealh]

And if you cooled your heat CPU/NB (my NB has 1,72 24/7) with water that's maybe not bad or ?

GNU..I have read from Tony/RGone...over 1.5v on the NB will shorten its like 24/7....becareful

[DerekT]

The effects of electrostatic migration are not transitory in systems that are running 24/7. Simply restarting the system, or turning it off for a few hours will not lessen the effects. The effects are cumulative. Electrostatic discharge is the characteristic (affect) and electrostatic migration is the effect. Given a system running 24/7 with highly volted CMOS substrates to the point of highly cumulative electrostatic discharge will bring forth the effect of electrostatic migration in the longer term and this is what permanently damages the CPU.

The characteristic (affect) will lessen in time, however the effect will not. Since the effect is cumulative, there WILL (in time) be a lessening of processor validity in data streaming and more adverse effects on the silicon. Very highly overvolted CMOS substrates will find the effect far more quickly than those who moderately operate the voltages.

You will see the effect before the affect. In other words, if you run a highly volted CMOS substrate you will see issues result. The system runs fine for days and then for some strange reason, it reboots. The games run fine and then for some strange reason (without software or hardware change) the FPS lower, etc. You shut your system down for a day or two and when you start it back up, it runs smoother and faster FPS etc. There are many more effects but I am sure this is clear.

Regarding life expectancies of the highly volted silicon substrate? One cannot, with any certainty gauge the life of such parts. Silicon Substrates do not have a life such as humans and animals etc. have. In their physical nature (with regards to electrostatic discharge and many other atomic level migrations) they are endowed with a half life, so you can consider their span as radioactive decay for purposes of determining possible length of minimal error use.

For example:

Given:

10,000 New CMOS's from the same fabrication silicon wafer. Their half life is the time that would pass before 5,000 of them fail. So, based on how long people have working CMOS's of the same fabrication silicon wafer, the half life of your CMOS could easily be 10 or more years (non overclocked).

This is why maximal voltages are given by the manufacturer. Be clear that the above does not mean that you will damage the CMOS in the shorter term. The problem is that one can not differentiate which is the silicon substrate that will last the long term and the one that will not. Highly overvolt the one that will and you very likely will still have the CPU for longer than you will want it but then you have the ones that are on the short end of the standard deviation (short end of the stick if you will) and the high overvolt will return errors and finally issue more frequent hardware interrupts.

It's the luck of the draw

[stealthbomber]

My E4300 is currently running under load @77C~ with 1.55v~ with stock cooler, it's covered by Intel's 3 year limited warranty so I'm not worried

[dekruyter]

"Their half life is the time that would pass before 5,000 [out of 10,000] of them fail"

Sorry to disagree, but the average 1/2 life of a cpu would be the same whether there are 10,000 or 5,000 or 1 of them. The units of decay are in "time" not "number of units". I think you just miss-stated a bit. I.E. (per your example), the half life of Plutonium is the same, whether you have one pound (or say a Mole) or whether you have 10 pounds (or say10 Moles). It is a subatomic property of the material. However, I do not know if a cpu degrades according to an approximate, mathematical exponential formula?

[Pete@X]

My E4300 is currently running under load @77C~ with 1.55v~ with stock cooler, it's covered by Intel's 3 year limited warranty so I'm not worried

Nice joke - but it's not -> http://cache-www.intel.com/cd/00/00/...0372_50372.pdf

1.55VCore voids the warranty - as it´s outside of Intel`s Specs

[Bail_w]

Nice joke - but it's not -> http://cache-www.intel.com/cd/00/00/...0372_50372.pdf

1.55VCore voids the warranty - as it´s outside of Intel`s Specs

Ya, unless you fried your chip then they will find out that you use more voltages that its limit. otherwise, they cant tell.

[stealthbomber]

How would they (Intel) know the CPU has been damaged over time due to heavily overvolted/overclocked usage when someone RMA's the 2-3 year old chip?

[DerekT]

"Their half life is the time that would pass before 5,000 [out of 10,000] of them fail"

Sorry to disagree, but the average 1/2 life of a cpu would be the same whether there are 10,000 or 5,000 or 1 of them. The units of decay are in "time" not "number of units". I think you just miss-stated a bit. I.E. (per your example), the half life of Plutonium is the same, whether you have one pound (or say a Mole) or whether you have 10 pounds (or say10 Moles). It is a subatomic property of the material. However, I do not know if a cpu degrades according to an approximate, mathematical exponential formula?

I am speaking towards the decay from the effect of electrostatic migration due to discharge, not the entire CPU life. Substrates of the silicon, not the entire CPU. The half life of such components that are subjected to the discharge and their error returns. The half-life is of a standard population of testing, not a single CPU. And no, a CPU does in point of fact, not degrade according to an mathematical formula.

If you read my words, you will note that I say that you may "consider their span as radioactive decay for purposes of determining possible length of minimal error use". Clearly my words show that this is a comparative analysis, not a discrete one. It is for purposes of understanding, not specifying discretely.

[H[]R3NDUS]

some times you just have to show yourself :-P

vdroop mod bios 1.7 ...windows 1.69 idle ....1.68v orthos loaded..

[dekruyter]

I am speaking towards the decay from the effect of electrostatic migration due to discharge, not the entire CPU life. Substrates of the silicon, not the entire CPU. The half life of such components that are subjected to the discharge and their error returns. The half-life is of a standard population of testing, not a single CPU. And no, a CPU does in point of fact, not degrade according to an mathematical formula.

If you read my words, you will note that I say that you may "consider their span as radioactive decay for purposes of determining possible length of minimal error use". Clearly my words show that this is a comparative analysis, not a discrete one. It is for purposes of understanding, not specifying discretely.

Regarding: "The half-life is of a standard population of testing", and the rest of your post.

Thanks for the info, and I think I understand what you are saying. But you seem to be couching it in terms of production quality control of thousands of units, not the life of a single CPU. And that's what we, or at least I, am discussing, or want to learn about. We are discussing the affect of voltage on our very own CPU, not a company's ability to predict the failure or error rate on "X" number of units produced, given (provided) a specified voltage specification. Further, it seems the voltage is "fine" tuned to the CPU (after production), rather than the cpu being designed/produced to the tight tolerance of a voltage spec.

So the questions continue regarding the life expectancy and affects of voltage increases beyond spec. of a single cpu; such as the affects of electromigration, as you say.

[DerekT]

So the questions continue regarding the life expectancy and affects of voltage increases beyond spec. of a single cpu; such as the affects of electromigration, as you say.

You have a valid request dekruyter. Unfortunately, the best we can do with a single CPU is hope it will respond well. There is no method to fundamentally ascertain the degree of CPU declination with regards to a singular sample. One can not apply statistical determination at that level.

One can only speculate. I was attempting to show the method that is used (by CPU design engineers) to bring about CPU degradation in a standard population of testing. Raising the voltages and setting the environment for failure beyond a point of error.

The simple request that is sought in this thread is not (in my view) possible. One can only theorize maximal voltages. You will find some who are able to run 1.6V (on air) without any mid-long-term effects (5 or more years) and others with the same fab/batch experiencing issues with lesser settings.

In the industry, the prevailing view has usually been a +.2V on .65nm products with decent core thermal extraction via air/medium water. +.4V on .65nm products with high end water and +.5 on LN/High end thermal removal with freeze attributes holding discharge in a more marginalized environment.

It's still a crap shoot though.

[IluvIntel]

I've been running 9 x 400Mhz (3.6Ghz) @ 1.52vcore set in bios (droop = 1.46-1.47v) for the last 6 -7 months on my L626 E6600 with only good air cooling in a closed up coolermaster case and I'm still not having any problems, in fact its doing 1.58vcore @ 3.75Ghz still on good air cooling with the system in my signature and its stable enough for my usage - PCMark05 entire test suite is enough proof for me cause' I never have a need to run it at full load for 24/7 usage anyway. That cpu test suite in PCMark05 pro version is better at peak stability testing than orthos imo with 4 threads running concurrently.

If the overclocked system is performing to your daily requirements then its good enough for the individual user irrespective of what any expert dictates to you.