Has anyone ever killed a Intel chip via OC/high temp/high V?

[CB5000]

Has anyone killed a recent Intel processor or even a recent AMD processor from very high temps/ voltages or insane overclocks? There are so many people saying that having a 32nm or a 45nm processor at 75 or even 80+C running 24/7 will kill the processor....and yet I haven't heard of anyone actually killing 32 or 45nm processors by temps that were too high, or high voltages. At worst the processor ran unstable and then reverting to a lower clockrate or voltage made it perfectly stable.
And then there is the shortened lifespan problem but... in my experience, just about everything else dies before the CPU... and I've never had a CPU die even over a period of 8+ years running at 100%, 24/7, 365.25 days a year at temps 70-80+... and i usually retire a CPU after 5-7 years anyway.

[Starfalcon]

I had a P3 back int he day the socket tab broke on one side and the heatsink fell off. It ran for a couple minutes before I could get it shut down and was perfectly fine after that...still runs fine in one of my older retro systems without an issue over 10 years later.

[canmnanone]

i dont think you can kill a processor by over volt, oc temp. it will throttle down or just freeze. people say that an oc will kill a processor over time but i have my oc q6600 3.6gig running 24/7 for over 3 years now its hasnt died on me yet. maybe time will tell if it does or not. but its been running strong from the start.

[TurtlePerson2]

The part of the processor that breaks is not the transistors. The integrated circuit elements are not destroyed by 3 volts. A few hundred volts might destroy a gate oxide, but a minor change in voltage shouldn't cause any transistor problems.

It's the things surrounding the CPU that I would be worried about. Bondwires on the chip will get hotter and may break sooner because of overvolting.

[just brew it!]

The fan on my Phenom 9550 seized up a few months back. The system never became unstable; one day I just happened to notice the core temp was 100C! I'm not sure how long it was running like that... probably not more than a couple of days. It still runs fine (with a new fan, of course).

[TurtlePerson2]

I was under the impression that the system would blue screen if the temperature got too high. At least that's always what happened with my chips when I did overclocking. There would be a certain temperature at which they would BSOD.

As a side note, 100 C is pretty impressive. Most chips are not tested at that temperature. I'm surprised that it was still working correctly.

[just brew it!]

As a side note, 100 C is pretty impressive. Most chips are not tested at that temperature. I'm surprised that it was still working correctly.

Me too. Either the 1st gen Phenoms are pretty tough, or my CPU temp sensor is mis-calibrated!

Edit: And for what it's worth (which may not be much...?), whoever programmed the default temperature thresholds in the Linux "lm_sensors" package seems to think that 95C is the critical alarm temperature for AMD CPUs...

[Dposcorp]

I think I had a P4 1.6a overclocked, and the heatsink failed while I was away. That was all she wrote.

[CB5000]

The funny thing I forgot to mention is that computers that weren't set up to do Boinc 24/7 or folding 24/7 on average had a lower life span due to motherboard failures, PSU failures, HD failures, etc etc. You'd think that heat generated from a 24/7 number crunching machine would have mobos that die earlier but for me that doesn't seem to be the case so far. I have had 5 mobos die on me over 15 years from blown caps to mysterious instability, 2 dead PSU's, 3 dead HD's... and they were all from low-use workstations for internet/email/wordprocessing and other light office work. As I said before, no dead CPU's... just plopped the chip into a new mobo and it worked perfectly. On the other hand... I've never had any of my gaming rigs that double as 24/7 BOINC/F@H machines fail on me.

For some reason, something tells me that computers like to be "ON" all the time... haha...

[kamikaziechameleon]

i dont think you can kill a processor by over volt, oc temp. it will throttle down or just freeze. people say that an oc will kill a processor over time but i have my oc q6600 3.6gig running 24/7 for over 3 years now its hasnt died on me yet. maybe time will tell if it does or not. but its been running strong from the start.

Mine was at 3.0 even for over 3 years no problem. Still running 3.0 ghz in my sisters system. No issue.

[StuffMaster]

I seem to recall that the P3 added the ability to shut off when overheating and the P4 could throttle down as is done today. Perhaps common wisdom about killing processors dates back to earlier days.

[Firestarter]

I seem to recall that the P3 added the ability to shut off when overheating and the P4 could throttle down as is done today. Perhaps common wisdom about killing processors dates back to earlier days.

The OP is not worried about that kind of damage, he's wondering about the effects of overclocking on the chip's lifespan. Heat is another factor, but at stock speeds/voltages, CPUs should last very long even when constantly run near their thermal limit.

[FireGryphon]

Back in the day a site called Tom's Hardware (a false and unreliable site) took some videos of what happened to various chips if they overheated. I remember some chips getting up to 700*-800* F or so with smoke coming out and the chips consequently dying. Nowadays CPUs regulate themselves so they don't die horrible deaths like that. You can definitely shorten their lives by pushing them too hard, but they'll still last for years, and it'd be rare to outright destroy them.

[just brew it!]

Back in the day a site called Tom's Hardware (a false and unreliable site) took some videos of what happened to various chips if they overheated.

While I'm not a fan of Tom's either, considering the poor design of early Socket A HSF retention mechanisms it wasn't *totally* unreasonable to ask the question "What happens to the CPU if the HSF comes loose while the system is running?"

[bthylafh]

I had a Celeron-533A which I'd mistakenly bought without a heatsink. I was so impatient and used to old 486 chips that didn't need HSFs that I just assembled the computer and booted it to DOS without a heatsink. Worked for 20 minutes before it shut itself off. Then I did it again and it lasted for a couple minutes. Found an old K6-200 heatsink (no fan) and that worked until the real HSF got there.

The whole thing worked for another three years before it was retired.

[FireGryphon]

Back in the day a site called Tom's Hardware (a false and unreliable site) took some videos of what happened to various chips if they overheated.

While I'm not a fan of Tom's either, considering the poor design of early Socket A HSF retention mechanisms it wasn't *totally* unreasonable to ask the question "What happens to the CPU if the HSF comes loose while the system is running?"

Actually, that video remains one of my fondest memories of THG, since it was pretty eye-opening and their methodology seemed legit. I guess my wording didn't come across like that.

[clone]

more often than not it's the motherboard that fails.

friend's heatsink fell off his Nvidia 64mb video card one day and it got hot enough to melt the chip contacts, he tossed it into the garbage bin and another friend pulled it soldered the connections put a new fan heatsink on it and it worked fine after that.

cpu's in general are quite robust and have some protection built into them so failures are less common, by 2004 I and some friends began pushing stuff pretty far with no concern to longevity, we'd started back in 1997 with overclocking and hadn't killed much so by 2003/4 we'd decided we didn't care and wanted to see how far and how long things would last.

combined we killed a few socket A's, killed a few P4 Prescotts and a cppl early core cpu's then lost interest in dealing with the stability issues, by far the failure rates for motherboards was triple the rate of cpu failure.

all of our group have since lost interest computers and have moved on with life to other things, really happy with my SSD purchase though.

[ludi]

The part of the processor that breaks is not the transistors. The integrated circuit elements are not destroyed by 3 volts. A few hundred volts might destroy a gate oxide, but a minor change in voltage shouldn't cause any transistor problems.

Partially correct -- a high voltage is what it would take to punch through a gate oxide instantaneously. Which is why we use ESD precautions.

However there are more gradual mechanisms at work. Back in the old days, everyone in the overclocking community used to know a thing or two about electromigration, in which the semiconductor doping compounds can gradually be moved between two regions of potential (voltage) difference until a kind of short-circuit path is created. A lesser-known problem in MOSFET devices is hot carrier injection, in which charge carriers having sufficient energy can tunnel out of the silicon and become trapped in a FET's gate oxide, increasing the leakage current. Both of these phenomena are aggravated by increases in operating temperature and operating voltage. Over time, the silicon device's stable switching frequency may be reduced, and its operating current and therefore its heat dissipation may increase.

I don't know how much of a problem this is on modern devices, especially given the complex power and clock management technologies that prevent the user from doing whatever they want. In the old days, the chip simply had to accept whatever input voltage and clock it was given, and either it worked at those parameters, or it didn't. Or in the case of a K6-2/400 I experimented on, it worked at 504MHz for a while but became increasingly unstable, and then it wasn't stable at 450MHz, and then it wasn't stable at 400MHz. Then it sat in a workbench drawer two years, and when I gave it to a relative with fair warning, it started working at 400MHz again, and did so for a couple more years. Go figure.

[CB5000]

I don't know how much of a problem this is on modern devices, especially given the complex power and clock management technologies that prevent the user from doing whatever they want. In the old days, the chip simply had to accept whatever input voltage and clock it was given, and either it worked at those parameters, or it didn't. Or in the case of a K6-2/400 I experimented on, it worked at 504MHz for a while but became increasingly unstable, and then it wasn't stable at 450MHz, and then it wasn't stable at 400MHz. Then it sat in a workbench drawer two years, and when I gave it to a relative with fair warning, it started working at 400MHz again, and did so for a couple more years. Go figure.

Now I remember having a cyrix 6x86 processor doing the same thing. I think electromigration used to be a bigger problem when there was more aluminum used as interconnects. Maybe the more modern electromigration resistant design has made it so that the chip doesn't need to be kept as cool as it used to?

[Sunburn74]

I'm pretty sure I killed an 875k with voltage. Not so much the voltage actually. The chip shortly died after turning on loadline calibration after like 2 months even though the voltage was in spec. Since then I never overclock with load line. Lots of people do; I think I've learned my lesson.