Death by OC, share your experiences

[Ho72]

So as not to hijack this thread, I thought I'd start a new one... sort of a poll only without the multiple choice.

In short, have you ever significantly degraded or killed a CPU by overclocking and, if so, what were the circumstances?* I'm not really looking for examples of flagrant idiocy but go ahead and share those too (for comic relief if nothing else).

I'm more interested in how common it is for a stable, 24/7 overclock to have issues down the road, say past the 3 year mark. Back in the day, I ran a OC'd Celeron 300A for nearly 8 years and it never missed a beat. Just wonder how feasible that is today or if anyone even keeps a rig that long (other than me)?

*By circumstances, we'll take as a given that temps were under control. I'm curious about things like C-states, Speedstep, esoteric voltage parameters and timings, load line calibration, etc. and how you think they may or may not have contributed to diminishing your processor and bringing an end to your fun.

 

[dave_the_nerd]

Not as such, no. I have fried things while playing with electricity, but not death by OC.

 

[ratjacket]

when i was just new to overclocking i killed acouple of motherboards instantly from to high oc

but as for cpus never had one go bad

and i always run 24/7 overclocks

my last system was a core2duo at 4.0GHz for 5years always on
no cpu power saving aswell so it was always at 4.0GHz

i keep them cool though.

 

[tolis626]

Well, I have only fried 1 piece of electronics ( My defective NVidia 6800 Ultra doesn't count! :sneaky: ), and it was an AMD Athlon XP 2200+ I think (I was almost certain that it was a 3000+, but I'm 100% sure that its stock clock was 1.8GHz, while the 3000+'s was 2167MHz, so who knows? I was too young then anyway, building using spare parts from my father mostly).

The thing got fried well into 2009, or maybe even 2010. I hadn't touched that PC for quite a while (It was a 2003 machine after all, all I did on it was mess with various Linux distros and Hackintosh stuff- yes, I managed to get a Hackintosh installation working on it), and suddenly one day I got the desire to overclock the damn thing. I'd been running it at stock clocks for its whole life, so why not? So I jump in the BIOS and choose the 10% overclock option (It was just there, not tweaking involved at this point) as a starting point I could work my way from. Then disaster strikes. I had forgotten that I was using the stock AMD HSF instead of the aftermarket heatsink I had been using before, which I had given to a friend. The PC made it to the Windows XP welcome screen before it went dark. Shortly after that, there was the stench of a fried CPU, which was that chip's "Goodbye". RIP. :'(

On a more serious note, I've been overclocking stuff, not just PC CPUs, for quite a few years. Nothing has failed me, except for that poor, old and tired AMD CPU. Only thing that has degraded, but is still alive and kicking, is the CPU of my old HTC Desire HD. It was a second-gen Qualcomm Snapdragon at 1GHz that I ran for about a year and a half at 1.85GHz. It just stopped being stable at that frequency with time and now it almost can't hit it. It's still running along happily at 1.5GHz though, no frying has occured.

 

[VirtualLarry]

I upgraded a PC for a friend. His new PC had an E5200 in an IP35-E. Not a bad little combo, and it was inexpensive.

Anyways, I OCed the E5200 to 3.75Ghz at 1.4v (CPU-Z under load) / 1.425v (BIOS).

After a few years (and the heatsink had to be cleaned out, was getting BIOS temp warning alarms during gaming from the PC speaker), I did some testing, and it failed Prime95 small FFT fairly quickly at settings that had been stable. Had to crank the OC down to 3.5Ghz, where it stayed, until upgrading to a Q8200 at stock speeds.

 

[SlowSpyder]

I killed a GeForce 6800 (non-GT/non-Ultra) overclocking it and unlocking it. Thought I had a close call at 1.7+v on my FX 9370, but all is well. (LLC on my motherboard is tricky )

 

[WhoBeDaPlaya]

I knew I shouldn't have read this thread earlier - got jinxed.
Just put 1.4V through a 3770. Was about to load test it in Prime95 when I saw the Vcore reading in CPU-Z and couldn't hit the kill switch fast enough.

 

[dmens]

None at home in my own machines, but I've probably destroyed three dozen chip over 10 years at work in a variety of testers and probes. Thermal protection is disabled so it's easier to kill them.

For example, typo on the voltage settings, 2.1V Vccore into a Westmere instead of 1.1V. Done.

Also on the very first batch of Nehalem B0 dies coming from the fab, I killed one within 10 minutes when I didn't tighten the heatsink properly.

 

[ehume]

i7 870K ran on variable Voltage and clocks, up to 4GHz @1.36v (software reading, so don't take the number to the bank; it was set far lower, but read 1.36v) . . . for a while. After 2y, It quit being able to do 4GHz. Had to scale it back to 3.6GHz with less Voltage.

 

[blackened23]

i7 870K ran on variable Voltage and clocks, up to 4GHz @1.36v (software reading, so don't take the number to the bank; it was set far lower, but read 1.36v) . . . for a while. After 2y, It quit being able to do 4GHz. Had to scale it back to 3.6GHz with less Voltage.

Had 2 i7-870s run for about 1.5-2 years in separate systems give out the same way. They were OC'ed with manual voltage 24/7 and eventually both systems started acting up at OC'ed settings. Then was normal at stock. Then stock started acting up. Then every clockspeed started acting up.

Death by over-voltage electromigration. Although in hindsight I should have used variable voltage, but that's no guarantee of a shorter lifespan.

Oh and, I know the CPUs gave up from over-volting too long because I replaced both with a drop in replacement and both systems worked fine afterwards with a new CPU. Nowadays I *never* use manual voltage for over-volt OC'ing for this very reason, it takes a few years but will increase the rate of electromigration if you're over-volting. Lesson learned. I use manual voltage to dial it in, but after that I always use adaptive or offset voltage when over-volting a CPU. Not a guarantee but definitely helps increase the lifespan. Now i'm aware that some have over-volted for 5-10 years with NP, so have I with older CPUs, but I wasn't so lucky in these two situations.

 

[DrMrLordX]

The only chip I completely flatlined was my Athlon II x4 635 which got nickel/copper shavings on the pins during a completely-superfluous lap job intended to remove CLU (turns out, you don't have to remove the stained portion of metal from the HSF or IHS unless you're OCD about that sort of thing). I put it into the board like that and it died instantly.

I volt-stressed the IMC on my x2-3600+ Brisbane chip to the point that it stopped "liking" certain memory ratios. It is now a (probably) dead chip after many years of service, and I'm guessing its "early" death was due to overclocking.

I volt-stressed my Sempron 140 on silly SuperPi runs to the point that its max clock dropped a few days after the runs. Later on - much later on - it started acting weird, and now requires 1.38v vcore just to run stable at stock speeds. It has since been retired, though it still works. Sort of.

 

[jihe]

killed an old thunderbird.

 

[skipsneeky2]

Pumped enough voltage into a 478 pentium 4 2.4Ghz 533fsb that i clocked to 3.5Ghz on a asus p4c800-e deluxe i guess 1.5v could have killed it.

Without that tower sitting directly into the flow of my window ac unit,it would overheat and shut down within a minute of a good stress test.. Big beast of a cpu cooler on that puppy too.

Q6600 g0 i clocked to 3Ghz on stock voltage on a g33 ecs board popped all the usb ports and the ps2 ports,about as close to death as it came for me.

 

[sm625]

I had been running an i5-750 for about two years. I've had to reduce my OC several times. It started out fine at 3.6GHz. Then in a few months I had to drop it to 3.5. Just recently I started having trouble at 3.2 and that's when I decided to go with the much much faster G3258. To be fair, this could easily be the motherboard. I bought it off ebay for $45.

 

[Bubbleawsome]

I knew I shouldn't have read this thread earlier - got jinxed.
Just put 1.4V through a 3770. Was about to load test it in Prime95 when I saw the Vcore reading in CPU-Z and couldn't hit the kill switch fast enough.

Wait, did it die? I've seen some haswell chips at over 1.4v and they survived for a little while.

The only thing I've ever killed was an hd 7770 because I overclocked it to 1300 on the core and higher on the memory. It ended up unstable at stock and eve down to 800Mhz.

 

[Alatar]

Edit: yes I know this is more along the lines of those extreme cases but why not, someone might find the pics and info interesting and/or funny.

I've had a 3930K and a Rampage IV Extreme die after 5.3GHz (~1.54v) 24/7 use for months on phase change cooling.

I've had an 8320 die from too much voltage under LN2:

It did work perfectly fine after the pic above but another 2.3V+ try months later finally killed it. Both my 8350 and 9370 have lasted over 2.2V without dying and without any apparent degradation.

My 4770K on the other hand has been sitting in its greased up socket for over a year and its been through DICE runs, LN2 runs, 1.9v+ and lots and lots of 5.4ghz benching under phase and it's still fine.

 

[SlowSpyder]

And in comes the pro to show us how death by overclocking is really done.

 

[WhoBeDaPlaya]

Wait, did it die? I've seen some haswell chips at over 1.4v and they survived for a little while.

Thankfully no, but I did not put it under load at that voltage.

 

[rbk123]

None for me. Have a Q6600 running at 3.2G for over 5 years now as my main rig. Also a 2140 at 3.2G for my HTPC. Both are slightly overvolted but I don't remember how much (nothing wack).

I also have a 2500k running at 4.5G for 2 years or so, but it's not heavily used. It does all my handbrake encoding but that is infrequent. Encoding at the stock speed is agonizingly slow.

 

[bradly1101]

I used to overclock, but only mildly and never had a CPU problem, but even so I just don't have the nerve anymore. Everything runs so well and fast at stock, I wonder, why take the risk? I guess I'm older now too though...

 

[Ho72]

I used to overclock, but only mildly and never had a CPU problem, but even so I just don't have the nerve anymore. Everything runs so well and fast at stock, I wonder, why take the risk? I guess I'm older now too though...

I kind of fell into the same camp, except that after a year of running my 3770K at stock speeds, I opted to try it at 4.4 GHz to see if I liked the result (I do).

That's partially what prompted this thread, wondering about the long-term viability of today's OC'd systems.

 

[blackened23]

I'll re-iterate if you over-volt, adaptive or offset voltage will significantly reduce any electromigration your CPU may receive. At the end of the day electromigration is what kills over-volted CPUs.

I'll also point out that this process usually takes 1.5 years or longer, but still, who wants to take the risk? Note that even adaptive or offset voltage is not risk free, but you can almost guarantee a longer lifespan of your over-volted CPU by using offset or adaptive voltage. Most CPUs spend the *vast majority* of their time in an idle state, especially while doing desktop tasks. I spend a lot of time on my desktop screwing around with browsing or email. And my CPU is at 800mhz and sub 1V voltage levels while doing menial tasks or idling/sleeping. I'd much rather take that over running 1.4V+ 24/7......You do not want 1.3 or 1.4V being pumped through your CPU while at an idle state, it's a waste of power and will speed any electromigration that may occur exponentially.

Just my advice!

 

[VirtualLarry]

That reminds me, I sold a friend a Q6600 rig a few years back. It was OCed to 3.6.

Maybe I should ask him to bring it back to me for a day for an OC checkup.

 

[Hans Gruber]

That reminds me, I sold a friend a Q6600 rig a few years back. It was OCed to 3.6.

Maybe I should ask him to bring it back to me for a day for an OC checkup.

My Q6600 will still do 3.6ghz on air. It just requires more voltage than back in 2008. The professional overclocker and power waster above who has insane clocks at 2.3volts. I like to put that in writing so they know that is the CPU voltage and not DD2 ram Voltage.

Who does that? I have an i5 3570k at 1.28v and it was running at 1.23v at 4.5ghz but decided for power gaming I needed to crank up the party to avoid blue screens after 6+ hour gaming sessions.

If you see the inside of a Dell computer or other manufacturers. I am amazed with their stock cooling solutions how they have not killed off chips in the past. I bought one Dell and after that, I built my own machines. I have yet to have a computer die on me since 2003. All have been retired or I donated the parts to others as hand me downs.

I was going to say if properly cooled you cannot kill a chip but the phase changer guy showed up with his story.

 

[ehume]

I'll re-iterate if you over-volt, adaptive or offset voltage will significantly reduce any electromigration your CPU may receive. At the end of the day electromigration is what kills over-volted CPUs.

I'll also point out that this process usually takes 1.5 years or longer, but still, who wants to take the risk? Note that even adaptive or offset voltage is not risk free, but you can almost guarantee a longer lifespan of your over-volted CPU by using offset or adaptive voltage. Most CPUs spend the *vast majority* of their time in an idle state, especially while doing desktop tasks. I spend a lot of time on my desktop screwing around with browsing or email. And my CPU is at 800mhz and sub 1V voltage levels while doing menial tasks or idling/sleeping. I'd much rather take that over running 1.4V+ 24/7......You do not want 1.3 or 1.4V being pumped through your CPU while at an idle state, it's a waste of power and will speed any electromigration that may occur exponentially.

Just my advice!

But I had my 17 875K on the predecessor of adaptive Voltage -- the Offset Voltage. It still lost its OC.