how many people here have legitimately had a cpu fail? What's the longest time one has ever lasted? It might just be because im so young (21) but i've never had it happen. my old e6750 core 2 duo racked up about 80,000 power on hours before i retired it, and i know there's still people running phenoms and shit that have been overclocked since they popped them in. That and RAM are the only 2 components i have never had to replace, even burned through a monitor. Seems these things pretty much last forever if you treat them right.
>>52479361
Yes on a laptop twice.
CPUs generally don't die on their own. To kill one you have to significantly overvolt it, bend its pins (provided it's a PGA), or just smash it.
i've only seen it once where a cpu failed post b/c dead cache
Aside from cases where someone had overvolted their CPU or physically destroyed them I have never heard of a CPU going bad.
80k hours would be over 9 years. did you have it in a server?
>>52479361
My gpu failed once. Most likely because i broke the heatsink on my old laptop
>>52479476
Was it new? possibly DOA. I'd think it's way more common for a chip to die really early than it is after it stands the test of time.
>>52479411
you ended up replacing the cpu on a laptop twice? jeez, you must have had horrible luck or it was ancient...
>>52479418
So it's fair to say that the main reason cpu's even die after they have clearly been proven to be in working order by extended usage, involves the VRM module or even the PSU going haywire?
I had an AMD CPU once, so that was automatically a failure.
>>52479586
GPUs should be pretty resilient too but they're pretty much all BGA packages so if for some reason they become unseated for some reason it's a complete pain in the ass to put them back on. Still GPU memory is the more likely component to fail on a graphics card
Has one of my 2 2500ks die after 3 years of 24/7 use, mildly overclocked. Intel were happy to take it back and sent me out a brand new boxed replacement a week later.
Only one I've ever killed and I've been building systems since 1995.
>>52479554
Yes, it only dealt in very light workloads for the longest time, so i just let it be. if it ain't broke don't fix it, you know? ended up grabbing a i3-4150T recently to replace it because I wanted to use the equipment for something slightly more demanding.
>>52479590
Turned out the motherboard was frying them, Nasty business.
Lucky for me they where e-school laptops with shitty pentiums so I just kept getting those for free until the motherboard finally crapped out and I threw it in the garbage.
i cranked my 4790k up to 2.4v and suddenly my display went up to 300fps. i also grt headshots in cs:go every time now. brolen cpu? hah, i don't thinl so, faggot.
>>52479361
I believe intel and AMD both fully test their CPUs, and all units they send away are fully functional.
More than 90% of dead CPUs are likely heat damaged, through improper cooling or overclocking.
>>52479659
they still have 2500k's in stock?
>>52479811
they don't fully test them, it's too costly I believe
>>52479811
If a cpu were heat damaged by say the retarded former owner ocing it too much on stock cooling how long do you think it would take to manifest through a failure if the second owner treated it properly? A few days? weeks? months? a year later? is there any point after acquiring one second hand that you can be reasonably assured it won't fail from that.
>>52479811
Overclocking is pretty much a non-factor. If the CPU cannot run at a clock speed it will crash or the thermal cutoff will kick in almost immediately and shut things down. If you enter the wrong settings and overclock to like 1THz or something your BIOS/UEFI will actually prevent your CPU from running in the first place.
The issue isn't overclocking or heat it's almost always the voltage. If the voltage is too high then you'll short the CPU out and fry it and that will be it. It might even be instantaneous. Usually if you do it gradually the increase in voltage wont really do anything too destructive but you might shorten the expected life time of the part.
>>52479361
>how many people here have legitimately had a cpu fail?
I destroyed one on my first PC build. I think it was a 486. You had to manually set voltage levels using jumper switches. I was in a hurry to see if everything worked and for some reason concluded that the default settings would work. They didn't. I think there might have been a fizzle noise. No smoke. No burning smell. Just blackness.
If I were into overclocking, then I imagine I might have roasted a CPU or two by now.
>>52479872
They do have ways to sort the bad dies out with machines though. The good dies are cut off of the wafer and the rest of the wafer is reclaimed. I think both Intel and AMD actually picks through the semi-bad dies as well though if there are enough of them. They use lasers to cut out bad cores or other dysfunctional parts and repackage them for sale on the lower end.
I chipped the die on an Athlon XP when trying to remount the heatsnk with the case still vertical. Besides that, I'm pretty sure my failing Core 2 systems were motherboard failures not CPU (RAM slots were dying then eventually would refuse to POST).
>>52479992
>They use lasers to cut out bad cores or other dysfunctional parts and repackage them for sale on the lower end.
It's my understanding that the chip design itself has redundancy to it. They scan every chip and zap whichever bits are bad (if they're all good, then I suppose they'd kill one at random).
On top of that, they do the binning thing for dealing with larger defects.
>>52480091
>It's my understanding that the chip design itself has redundancy to it.
For extremely high-end chips, i.e the ones that are used for powerful servers, yes. There are redundant caches and whatnot that only kick in when one is seen to go bad via ECC In consumer grade chips, even enthusiast, this is not really considered cost effective so they don't do it.
>>52479908
CPUs degrade constantly. Faster with more heat or more voltage.
At some point a given voltage will start to leak at places, when it becomes to much, it will crash. There is no way to tell how much life a CPU has left.
>>52479921
There are plenty of cases where CPUs run under crazy voltages or extreme heat for extended time. Settings that cause crashes are of course not the problem.
Still, overclocking in general is only a small issue compared to 'regular' heat death, because of terrible cooling of a not overclocked chip. Just look at how how some laptops run.
>>52480142
You misunderstand me. I'm talking about at the factory.
They intentionally design redundant portions on the chip because they know that they'll never achieve 100% perfect production. It's more cost effective to scan and blast every chip than to toss a whole chip due to statistically inevitable flaws.
>>52480142
Even consumer chips have redundant transistors. It's practically impossible to make a die with billions of transistors and zero faults. Most of them don't matter. At worst they'll slow the CPU down by a small fraction of its speed.
>>52479872
>>52479811
they do it though, that why we have i5, i3 and pentiums, they're all defective i7 that were dumbed down.
>>52480196
Oh you're talking about how they just rebrand lower quality silicon into lower end hardware? of course, that makes sense on too many levels to do otherwise... I don't know if redundancy is the word here but they certainly know how to disable certain parts of the silicon without rendering the entire thing useless. I don't think it involves completely cutting it out from the die, though.
>>52480237
and i7 extremes are just defective high-end xeons.
>>52480196
The speaker in this talk actually mentioned that.
https://www.youtube.com/watch?v=eDmv0sDB1Ak
If there's a significant amount of "white space" off to the side they'll add gates that don't really go anywhere. The idea is that those gates can be connected together using one of the metal layers which is supposed to be cheaper as opposed to making an entirely new mask for the silicon layer which could cost thousands of dollars if not millions.
>>52479361
I had a CPU die on me once and it was RMA'd
I think it was a cheapo MSI board that was the culprit though
>>52480241
>Oh you're talking about how they just rebrand lower quality silicon into lower end hardware?
No. No, I'm not. In fact, I quite clearly contrasted it to binning.
I'm tired of repeating myself though, so whatever.
