Dangers of Overclocking?

[ryema22]

This might be a newb question, but I?ve never done a whole lot of overclocking with my previous builds:

I?ve been planning to build a Q6600 system as soon as I can get my hands on a G0 stepping chip and overclock it to the ~3.0-3.4GHz range. I?ve recently read a lot of stories about the increased temps and voltages causing all sorts of problems down the road (Electromigration, instability, inability to hold the same clock speeds, etc.) and shortening the life of the CPU. I want this system to run for 3, maybe 4 years ? so really, how dangerous is overclocking to the longevity of a chip?

 

[SickBeast]

You can kill a CPU entirely by overclocking. I did that once on a K7 chip. I volt-modded it to see how high I could clock it and it died. For the record, it hit the wall at a certain point and essentially the volt mod did nothing aside from killing the chip.

Keep your voltages within reason (usually .2 to .3 volts from stock) and you should be OK. Watch your temps.

I've had an Athlon XP-mobile 2500+ running at 2500mhz (up from 1833mhz) for probably 4 years now and it's been perfectly fine. It never 'degraded'. Basically, if you damage a CPU whatsoever it will simply stop working entirely.

With the Quad Cores, I'm thinking you should leave it at stock voltage or only give it a very minor bump. Check the forums to see what voltages other people are using.

 

[SerpentRoyal]

Limit the CPU core voltage to +10% over spec.

Limit the maximum CPU temperature under Orthos/TAT/S&M to Tjuction - 25C. Download Coretemp for Tjunction temperature. Latest Intel chips are rated at 85C or 100C Tjunction.

Enable C1E and EIST to lower Vcore when CPU is not under 100% load. Note that some MBs do not work properly with C1E and EIST.

 

[evoic]

Voltage itself is rarely the killer, as most motherboards are limited in BIOS as to how high they will go. (1.55v, 1.6v, 1.7v, etc)

The main killer is almost universally understood to be TEMPERATURE.
I've posted this before, but it's perfectly fitting here, so I'll post it again:

To my knowledge, regardless of voltage, I have yet to hear a single story of someone killing a C2D.

Excessively high temps are obviously a concern for all parts of the hardware, and of course high voltage for extended periods of time can lead to silicon and gate degradation.....

Having said that - I don't think that 1.5v (*NOTE* I was responding to someone else's post) will do anything at ALL to hurt the lifespan of these chips, other than knock some time off the end.

I would say that you're looking at the life expectancy of your CPU dropping from 12-15 years......down to maybe 7-9 years.

These are numbers that I'm completely pulling out of my ass, mind you.
But don't be surprised if they're close, and even if they're not, you get the basic idea.

Will your CPU explode after 6 months of running like that, no.
Will it possibly / probably diminish the life expectancy, yes.

More than voltage, the primary thing that you, me, or any user should be concerned with is heat.
If your system & CPU stays cool at 1.6v (35c or under idle / 55c or lower under load) then I wouldn't worry about it at all.

By the time that chip finally kicks the bucket, you will have purchased his great-great-great-great-great-great-great grandson chip.

The Intel Ocho-Quad Duo at 10000Mhz.

8x4 cores x's 2 Chips mated together in Intel proprietary packaging with each core operating at 10Ghz.

How much voltage do you think your CPU will take then?

 

[videopho]

I second the previous post's comment: HEAT is silicon's worst enemy.
For me who is just an average overclocker, a dead silicon is another excuse for a new upgrade.

 

[SerpentRoyal]

ICs are nominally rated up to 70C case temperature. C2Ds can display core temp, which is approximately 15C higher than the nominal 70C. A 20% duty cycle at 70C design temperature should not significantly degrade the life of a CPU (7 years before failure).

A quick spike in temperature up to the Tjunction limit (85C or 100C) should not permanently damage the CPU. Excessive voltage can quickly degrade the doping layers of the IC, causing permanent damage to the IC.

Want more proof? Overclocking a C2D IC from 1.8GHz to 3.0GHz will normally raise the CPU temperature by 5C with no increase in Vcore. Adding 0.1Vcore may yield another 400MHz of core speed, but the CPU temperature will go up by another 10C.

 

[thilanliyan]

Doesn't the extra voltage also lead to gate oxidation regardless of cooling? Or is the oxidation due to the extra heat put out by the increased voltage?

 

[apoppin]

Originally posted by: thilan29
Doesn't the extra voltage also lead to gate oxidation regardless of cooling? Or is the oxidation due to the extra heat put out by the increased voltage?

of course, faster electromigration occurs with just the FSB increase ... much more so with the voltage increase
-but it is certainly less significant than extra voltage plus heat

 

[KeypoX]

so what is reduced? if something is oxidized something must be reduced

 

[jjyiz28]

what you guys never heard of Sudden Overclocked Northwood Death Syndrome??

"There are a lot of explanations of the reason why Intel Pentium 4 ?Northwood? processors malfunction even though they run at lower than 45 degrees Celsius average temperature, according to thermal diodes located inside the CPU."

http://www.xbitlabs.com/news/cpu/display/news6375.html

of course im not saying this occurs in C2D CPU's, but it may be possible. me personally my rig is an athlon 1800+ (don't laugh) which was overclocked to 2.2GHz in the past but i had to run it at 1.9Ghz w/stock volts; even with increasing the voltage i can't get a higher OC. your mileage may vary of course.

 

[SickBeast]

The voltage is really the root cause of the heat, and it's the one variable that the user has full control over, which is why I say that it's the true "CPU killer".

 

[bigpow]

I've been overclocking since the Pentium MMX 166 days and I've never killed any CPU.
Use common sense, overheating & overvolting kill if in excess. Do the research, max volt/temp, measure - test -etc.

IMO, Only negative aspect of overclocking, it's addictive (and may cause non-stop bragging against Mac owners)

 

[betasub]

Originally posted by: SickBeast
The voltage is really the root cause of the heat, and it's the one variable that the user has full control over, which is why I say that it's the true "CPU killer".

Makes sense to me, but your earlier post you say:

Originally posted by: SickBeast
You can kill a CPU entirely by overclocking.

Overclocking doesn't necessarily require a higher voltage - in face many recent gen CPUs can overclock very well at default (or even lower voltage). Heat can still be a problem: it will increase if you increase the clock speed, so cooling is always important for overclocking.

As for the OP worrying about long term problems: "(Electromigration, instability, inability to hold the same clock speeds, etc.)" will no doubt become a problem if you push a CPU hard enough and long enough, but then he probably lies awake at night worrying about a meteor strike on his house.

 

[ryema22]

I don't exactly lie awake worrying about meteors, but I'm going back to school in the fall and I won't be able to afford to replace this build for the next 3-4 years, so if overclocking presented too large of a risk I would want to know that beforehand and just keep it at stock speeds.

It sounds like if I don't add any voltage and invest in a good heat sink I should be fine over that time period, correct?

 

[apoppin]

It sounds like if I don't add any voltage and invest in a good heat sink I should be fine over that time period, correct?

succinct summary

there are no guarantees, of course ... just like in "real life"

 

[Arkaign]

I've never had an overclocked processor die, personally. I've had dozens. The key is cooling, and keeping voltages sane.

I'd wager that a poorly cooled but stock-clocked processor will die much sooner than an excellently cooled overclocked processor. My examples : A P3 @ 1.66Ghz on a TUSL2-C that has seen daily use for about 6 years. Never a hiccup, same load of Win2k even, 18.2G 10k rpm IBM drive on a Adaptec card. I also have a AXP 2500 @ 2.2Ghz on an NF7-S, with a thermalright XP-120 and 900rpm fan. And a 1.6 northwood @ 2.4Ghz. All overclocked for years and still 100% stable.

 

[Dravic]

Originally posted by: ryema22
I don't exactly lie awake worrying about meteors, but I'm going back to school in the fall and I won't be able to afford to replace this build for the next 3-4 years, so if overclocking presented too large of a risk I would want to know that beforehand and just keep it at stock speeds.

It sounds like if I don't add any voltage and invest in a good heat sink I should be fine over that time period, correct?

You will be fine, all things in moderation. Your not trying to set the overclocking record.

10-15 year life cycle for the chip, and all estimates are that you are knocking off maybe 5-7 years for overvolting and the increased temps. Do you realy think you will be using this cpu in 7 years? I'm still using chips I had overclocked in my main rig in secondary servers now at stock speeds.

Overclocking is a risk/reward game. You dont buy the highend CPU and overclock it, you buy the mid range cpu and overlock it to the be like(or better then) the high end part.

If you cant afford to lose the CPU and have never overclocked before maybe spend the extra couple hundred buck for the high end cpu and call it a day. The only thing worse then losing a chip to overclocking (which i have never done) is the jacka$$ that rma's the chip he kills.

 

[evoic]

Originally posted by: SerpentRoyal
ICs are nominally rated up to 70C case temperature. C2Ds can display core temp, which is approximately 15C higher than the nominal 70C. A 20% duty cycle at 70C design temperature should not significantly degrade the life of a CPU (7 years before failure).

A quick spike in temperature up to the Tjunction limit (85C or 100C) should not permanently damage the CPU. Excessive voltage can quickly degrade the doping layers of the IC, causing permanent damage to the IC.

Want more proof? Overclocking a C2D IC from 1.8GHz to 3.0GHz will normally raise the CPU temperature by 5C with no increase in Vcore. Adding 0.1Vcore may yield another 400MHz of core speed, but the CPU temperature will go up by another 10C.

I was with you all the way until the end there.....

That's just impossible for you to claim.
First off, you would probably only see a 10c rise in temp by adding 0.1v if you were using stock cooling.......even then I don't think that would be the case, but there are so many variables to consider that it makes the argument pointless.
Cooling, ambient, TIM, air-flow, load, etc.
You CANNOT just flat out claim that raising Vcore by 0.1v makes the CPU temp go up 10c.

That's like saying, "If I press the gas - My car goes 72 miles per hour"
Uphill, downhill, how long, what surface, etc.
Too many variables.....

 

[fire400]

geeks, give him a damn break.

let him hear whatever the hell he wants to hear:

for me, here's the reasons why:

1. affects boot up process
2. instability
3. overheating

I never pay attention to:

1. life span; are you going to run a processor for ten years? for the typical enthusiast, almost never. ten years ago, the Pentium II existed, and many were still running Pentium I computers - 133MHz MMX anyone?

2. warranty; your 'usual' 30 days will be up soon enough. don't feel bad.

3. saving electricity; if you can't afford the electricity costs from your computer alone, then it's about time to rethink that career of yours or monitor your other spending habbits more closely.

This is fire400 reporting live at anandtech.com Stay tuned and listen for more updates.

BOTTOM LINE: I don't care if you overclock your motherboard if it's your own. Don't ever go into my BIOS and change my timings or voltages because that pushes me off a cliff when you change my default settings. Or when I get my timings tweaked out the way I want, and then I load up a game...

-it's slow, I update the game...

-it's still slow, I turn off all of my apps...

-I restart the computer...

-I open up my case to see if anything is overheating...

-I flash my BIOS...

-I try a different game and it's slow like the other one...

-I finally check my BIOS and all the stuff is changed...

I get fvckn' pist as hell, do you think you can make my computer run faster, when in fact I'm the one who's studied everything about it, top to bottom, played with it, tweaked it, got it right finally ... and then someone just wants to try to adjust my clocks?

That's the danger of overclockin' - I mean you wanna touch a black man's radio? Do it, see what happens. You touch a geek's computer, they forgive you. You touch a nerd's computer, they just take their glasses off, look at you real hard and laugh. Touch your mommies and daddies computer, they'll maybe chase you around the house with a broom stick and frying pan, if you're lucky your dad will just take the Mustang keys away from you.

But you wanna touch a enthusiast's computer? Huh? You think you can just flip his monitor on and flicker that power switch, pull the damn battery out from the motherboard, 'n just start tweakin' BIOS settings!? That gets my teeth grindin' on concrete.

Don't try to tell me that overclocking is dangerous because you're going to fry something. I don't care about safety settings, or Intel's safeguarded shutdown technology when your chips get too hot.
You're messin' with people when you overclock. You think Intel customer support wants to hear over and over again, that another lame joke like yourself gets into a blue screen with a bunch of numbers and just wants to test out how to overclock, and you turn off the overheating safety features of the motherboard, so your little "LGA775" chip gets fried? Your customer support representative is sick of hearin' that kind of crap, and will tell you to cry to your daddy to get you another CPU.
Yeah, you wanna do it on a GPU, don't do it on my x1900? I tweak it just the right way, I don't want pranks just opening up my ATItool to hit up the card with crazy numbers, and all of a sudden my games don't play the same because I have a permanently damaged graphics card that artifacts when I play my PC games?

I'm not playin' Starcraft on a PCI graphics card. I'm not going to play on integrated graphics, and I'm not even going to sniff AGP. I want my x1900XT tweaked to an XTX, playing any game I well please.

Don't pull CPU frenzy on me. You friend my chip? I memorize every single seriel number on my CPU's, and I mark them with my signature, I make sure that the paste is perfect. You think you can achieve that? Don't fry my shi_ and expect to replace it without me knowing. I don't care if it's the exact same replica. The fact is, it's not the same chip. Because the same chip is now in some land fill... rotting with the rest of the carcuses of old rotten tomatoes, broken Hello Kitty toys and worse of all, CRT monitors. You see my LCD? I can't even stand lookin' at CRT's, they're hideous, horrendous and hateful to my eyes.

Don't touch my computer's BIOS, and don't tweak anything. 'Cuz I will make sure you regret the very second you ever lay'd eyes on my rig. Stay home if you have to and work on your Celica, overclock your AMD fanboy X2 +5000 rig, I don't care.

Stay away from my tweaks.

 

[geokilla]

Wow. How long did it take for you to type that?? Plus, what was the point of that....it's like a story.....

 

[Arkaign]

^ you missed your meds again this morning, didn't you?

 

[evoic]

Fire400......Psyche ward on Line #4.

I'm POSITIVE they're calling for you.

 

[CTho9305]

Originally posted by: KeypoX
so what is reduced? if something is oxidized something must be reduced

I haven't heard of gate oxidation, but two effects that do exist are "hot carrier injection / impact ionization" and "negative bias temperature instability". Super-simplified explanations:
hot-e: excited electrons (or holes) jump into the (non-conducting) gate oxide (or the drain) and get stuck there
nbti: (I don't really understand it) hydrogen atoms diffuse from the transistor body / gate oxide interface and into/across the gate oxide

Both effects are temperature and voltage-dependent.