is it nearly impossible to burn out a CPU anymore?

[drizek]

If overclocking results in undetectable errors so much, what's to say that CPUs aren't throwing around loads of errors at stock speeds?

Prime95, OCCT, BSOD

 

[aigomorla]

OMG...? How long have any of you been in the overclocking game...? From several years ago we learned after overclocking our old single core systems (from Celeron 300a`s to XP`s) along with the old OS`s would corrupt files just about every time we had a crash (was a cross your finger game)... The older 9X series of OS`s were alot worse but with the arrival of NT tech things got better but it still happens... Why do you think systems crash...? A virus or worm can cause havoic but a dying cpu or data corruption caused by and overclocked system can be alot harder to figure out... What do you do when you read a dmp file that has nothing but 0`s...? It happens...

i bet i went longer then you, and have OC'd on higher scales then you.

What your describing is the result of a bad oc, when the cpu does get corrupted.

However a well stressed, and well built machine with oc (meaning overclocking ready gear, and yes theres a difference), can last as long as a stock computer can.

There is ultimately 2 values one needs to worry about when overclocking a computer.

1. Heat... Heat kills, straight and simple.
2. Voltage... yes this used to not be that big of a problem until we uber shrank nodes..


Before when cpu's were tanks, id say voltag wasnt an issue.
Ive had E6600's which could swim in 1.5Vcore all day and not show any signs of degrading after 2 yrs of usages.

Then we got into smaller nodes.. like 45nm.. and omg the 32nm.
Thats when we started seeing more voltage problems when yorkies started to die at 1.45vcore, when there kentsfield big sisters were chugging along.

Then we saw a repeat with bloomfield... the first gen bloomfields were tanks, and slowly started overclocking better but at the price started becoming more fragile.


Its not impossible to fry a cpu.
Ive toasted 2 gulfies so far so i can say with fair confidence.. there is no such thing as a super man cpu.

 

[Ratman6161]

Doubt it ibex333, at best i think a unstable overclock would cause a crash or a blue screen.

Done overclocking on plenty of cpus without such a issue.

Never fried one cause im anal about temps and having the best cooling solution for the processor.

Yup, if you follow some basic rules of sanity, your chances of burning out a CPU are pretty low. Basically just turn up the speed in small increments and re-test each time + don't get insane with the voltage and watch the temperatures. You will get instability before you get to a point where you will burn out the CPU. Then again, I typically don't increase voltage more than about 1 increment above stock and go with the highest overclock I can get at that voltage.

On my just decommissioned Q6600 I could get 3.6 Ghz with increased voltage but stuck with 3.2 because I could get there without a voltage increase. On the new 2600K I'm up to 4.1 Ghz so far while leaving the voltage set to Auto.

 

[Ratman6161]

OK, true story. At one time I had an AMD K63+ 450 OC'd to 600 Mhz (yea, that's Mhz not Ghz). It had run for a long time with no problems. But one day it started crashing every hour or so. I spent days troubleshooting the problem never thinking to check temperatures (which was a lot harder to do in those days) because heat had never been a problem with this CPU. Long story short, it turned out the CPU fan was so clogged up with cat hair that it had stopped running completely. Replaced the fan and that CPU kept running fine for several more years. At the time I gave it away, it was still running fine.

 

[bryanW1995]

a mild curiosity i've been thinking about. personally, i've never had a CPU or GPU fail on me, and i've run some HOT stuff.

i started building computers in the pentium 2 era (what active cooling?). as i've built newer computers with more complex cooling setups; and helped other people with poorly seated GPU coolers, push pin woes, and whatnot, i've still yet to see a microprocessor destroyed by heat. you'd think an i7 with little to no heatsink contact would not withstand multiple bootups and crashes as the user doggedly attempts to get the computer set up. but i've seen it- just correct the issue and it seems to be no worse for wear afterwards.

am i just being silly? or have we just gotten to a level of efficiency where these things just have a hard time lighting themselves on fire?

wonder what an overclocked P4 with no cooler and no failsafes would do...it's weird, those space heaters, in many cases, have been running maintenance free in a hell of a lot of workplaces for ten freakin' years.

Aigo has a penchant for demolishing cpus, though he often has to try very hard to do the trick. As IDC mentioned, most modern cpus have so many failsafes that it takes a while for you to notice that they're failing. In many cases you'll fry the mobo/psu/ram/etc etc before getting the cpu. Part of the reason they seem to last so long is that for the dell buyer, the cpu should last 10-20 + years so it will usually be replaced before failing, while for the power user who overclocks 80% and uses 25% more vcore he is replacing it every 6 mos anyway so he doesn't see the damage yet.

 

[bryanW1995]

The 8000 series used a new solder material that earned a reputation for melting too easily when the chip overheats. The semiconductors themselves can survive much higher temps for a short duration of time, but the material below them is plastic and solder, and most likely, that was what you broke.

You should have baked the 8800gtx for 10 min or so.

 

[pantsaregood]

Prime95, OCCT, BSOD

I've always used OCCT and Linpack. The issue is that some errors are supposedly undetectable by those tests.

 

[aigomorla]

In many cases you'll fry the mobo/psu/ram/etc etc before getting the cpu.

Excellent point.

You'll get faster memory degradation / board degradation then CPU, until we get into the 45 / 32nm...

Then we are seeing despite still having room temp wise, the voltage safe net we thought we had was brought down to a much lower level.

Hence the term.. voltage becomes "crack" for the cpu.

 

[TonyB]

Fried my old Athlon XP 1700+ because I didn't seat the Heatsink properly.

 

[Mark R]

If overclocking results in undetectable errors so much, what's to say that CPUs aren't throwing around loads of errors at stock speeds?

Nothing. However, most PCs are sufficiently stable that their CPUs probably aren't throwing up many errors. However, you can never be sure. CPU manufacturers rigorously test the CPUs they sell to ensure that they work. The designers know which circuits are going to be least reliable, so they tell the test team to test these circuits extra carefully.

To some extent, modern hardware has some error checking built in - modern CPU caches use ECC RAM, to detect and correct minor corruption (which happens quite regularly), interconnect circuits, like PCI-E, include error checking - if the CPU detects a PCI-E malfunction, it will retransmit the data automatically.

However, software engineers do sometimes find weird errors (e.g. from crash dumps) that suggest some sort of CPU malfunction. I'm sure that microsoft has detailed statistics from all the crash dumps they collect, and could hazard a guess at how frequent CPU malfunctions are (although they have not published this information).

The problem is that it is extremely difficult to detect calculation errors. That's why engineers and scientists performing highly critical calculations will often run the calculations twice (or more) to double check. Distributed scientific projects like folding@home send out each work unit 2 or 3 times to different people, just to check that all the comps agree on the result.

Where CPUs and circuits need to be designed for utmost reliability (e.g. spacecraft, where a malfunction could mean the loss of a $1bn machine, and where CPU reliability is degraded by high radiation levels), then it is possible to build a CPU where each individual circuit module is built in triplicate, and connected to a voting circuit - so if one circuit malfunctions, it can be out voted by the other 2 which hopefully are giving the correct answer.

 

[piasabird]

When I have seen the most damage to a hard drive is when the power supply does not provide enough power or fluctuates.

 

[BD231]

When I have seen the most damage to a hard drive is when the power supply does not provide enough power or fluctuates.

Agreed.

 

[Bartman39]

i bet i went longer then you, and have OC'd on higher scales then you.

Are you really sure of this...? Or is this my dick is bigger than yours...? Not sure where your really going with it either...? Really strange coming from a Moderator...?

Take note I was OC`ing 386`s which I`d bet most on here dont even know what they are...? Made a Pent 90 with the floating point flaw (do you know what that was...?) run at 166mhz with soldered pin modds for voltage and what would be considered ghetto cooling (heatsink off a welding machine transformer)... I`v killed several Celeron 300a`s with voltage from hell to make them run 600mhz and yes they took out boards as well... As for Celeron 366`s all I can say is dry ice will get one to 900mhz on a stupid day just they dont last long... So with this in mind we really did not even notice what nm meant to much... Did`nt even screw with PII`s waited for PIII`s that would burn a pretty shade of dark blue or even black when they fried... Oh and AMD`s of every PCB color would smoke like a drag racer doing a burn out at the track... I even managed to cook a very nice OC`ing Opteron 170 that still has a sun spot right in the middle that would have to be ground out with a grinder to clean it... (still wont work) I skipped over C2D, C2Q had other interest at that time like OC`ing BBC`s and Merc Promax`s... But jumped in with Ph II`s and I7`s and now with SB`s but dont have the funds with a teenage daughter that will be going to med school soon...


PS... Dont care what you say I`ve seen and know what OC`ing can do which is very unpredictable no matter how stable you think it is or just how long it will last... One other little question do you even know what causes heat in the first place...? Resistance of the componet which causes a VOLTAGE drop which creates heat (so which happens first...?) Basic shit dood...