Threadripper even has two dies...Oh, no no, all CPU's die son. Right from their birth they die in fact. See, you got yer 14nm die, yer 12nm die, and in future CPUs will die in 7nm!
-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
Why do CPUs never seem to break?
- Thread starter 413x
- Start date
Soldato
I had a 2600K die on me...replaced under warranty by intel no questions asked....had it benching at nearly 5.5GHz and it did a sub 6.7 sec super pi...pretty sure that never killed it....its replacement had an easier life..still running 24/7 at a mere 4.5 to be safe 
Previously i had a opteron 1.8Ghz cpu die...had ran it at 2.8Ghz for 2 years then it just died...either that or MB died and took it with it as they both went at the same time....no warranty on that one
Previously i had a opteron 1.8Ghz cpu die...had ran it at 2.8Ghz for 2 years then it just died...either that or MB died and took it with it as they both went at the same time....no warranty on that one
A single processing mistake on a CPU is bad. Think about the ridiculous stress testing people do when overclocking for CPUs.
A random artifact with a GPU now and again that you probably didn't even notice? Not so bad.
I do remember that Intel 45nm originally was a bit dodgy when overvolted.
A random artifact with a GPU now and again that you probably didn't even notice? Not so bad.
I do remember that Intel 45nm originally was a bit dodgy when overvolted.
Soldato
The only one I've killed is an E8400 core2duo. That was over voltage. Intel replaced it though
Associate
I've had motherboards, power supplys, graphics cards and hard drives fail but never a CPU.
Good point someone else made about the comparison between a GPU and a CPU, the same comparison kind of applies to a motherboard as well!
Good point someone else made about the comparison between a GPU and a CPU, the same comparison kind of applies to a motherboard as well!
Associate
- Joined
- 2 Sep 2013
- Posts
- 1,902
I've had 1 (suspected) chip die on me so far in a fair list of CPU's since 96. It was the i7 920.
I never diagnosed the issue back then, I just replaced it with a 3570 Z77 system and transferred most of the insides over instead. From what I remember, it was either the processor or the motherboard that suffered failure; so I suppose it could have been the motherboard instead.
Since it was the 920, it required the x58 motherboards and when it died back in late 2012/early 2013, that had already been superceded by the X79 setups, so the cost was more to obtain a replacement motherboard to test. So I was forced to just abandon the whole setup instead of doing a proper test to find the real culprit.
I never diagnosed the issue back then, I just replaced it with a 3570 Z77 system and transferred most of the insides over instead. From what I remember, it was either the processor or the motherboard that suffered failure; so I suppose it could have been the motherboard instead.
Since it was the 920, it required the x58 motherboards and when it died back in late 2012/early 2013, that had already been superceded by the X79 setups, so the cost was more to obtain a replacement motherboard to test. So I was forced to just abandon the whole setup instead of doing a proper test to find the real culprit.
Soldato
I did IT Support for 15+ years. Warranty repairs in-store for a couple of years, corporate support since. Must have had 15000 PCs through the organisation in that time. Number of dead CPUs I've seen .... 2! One Cyrix 6x86 (they ran hot and close to the sun!) in the store. one Intel P3 when I first came to the corporate job. Number of faulty motherboards ... too many to recall.
Intel CPUs are generally extraordinarily reliable.
I've seen a couple of server CPUs need replacing but they are usually collateral damage. Things like VRMs blowing due to a Air Con failure ... taking the CPU with it...
Intel CPUs are generally extraordinarily reliable.
I've seen a couple of server CPUs need replacing but they are usually collateral damage. Things like VRMs blowing due to a Air Con failure ... taking the CPU with it...
One bit of hardware I've never really had much experience of - encountered in passing the odd setup I think running OS/2 but I might be wrong about that back in the day doing a bit of work in IT.
Soldato
Quite possible, they were fabbed by IBM. They did sell a fair few machines with 5x86 and some 6x86 CPUs in the late 90s.
I had a 6x86 PR200 for a while. They were 150MHz and early ones were 75MHz FSB which caused a few compatibility issues with hardware. Oh and 3.5V. It managed to melt the CPU socket, but never technically failed! The replacement was the later 6x86L and 2.8V. That got upgraded swiftly when I bought a 3Dfx card and the FPU couldn't keep up.
I had a 6x86 PR200 for a while. They were 150MHz and early ones were 75MHz FSB which caused a few compatibility issues with hardware. Oh and 3.5V. It managed to melt the CPU socket, but never technically failed! The replacement was the later 6x86L and 2.8V. That got upgraded swiftly when I bought a 3Dfx card and the FPU couldn't keep up.
OcUK Staff
I killed afew when doing XOC ofcourse with high voltages etc. Not as many as people think though.
On ambient very very few die unless a motherboard vrm failed or similar.
On ambient very very few die unless a motherboard vrm failed or similar.
Soldato
Mmmmmm. Not quite as i remember our conversation a few years ago in Cambridge Ian. Dag asked me to "bury the hatchet" between us, so i did. Anyway after it was "buried" we were talking about CPU "deaths" lol. As i remember we killed at least the same amount of Opteron 170's and 180's as each other ? Am i remembering a different conversation ?
I get the feeling that CPU'S are not really launched particularly close to their design limits, this seems to be changing with more recent CPU'S though. Whereas GPU'S are much closer to their limits than CPU'S are. This is probably a factor in their long livety.
.
.
Soldato
It's basicly a bit of rock glued to a pcb right? Not surprised they are so reliable
I've seen a single core fail on a Q6600, intel replaced with zero issues (general experience with intel is they replace pretty much anything remotely in warranty), a few socket A Athlon's blew (high clocks/volts, water + peltier), other than that i've had few issues.
In 22 years of using and building PCs I've had two CPUs that became faulty. The first was a Cyrix MII 300 that died a complete death, just a sudden black screen and it was dead to the world. The second was an Athlon 1700XP that worked ok but kept having cache failures, both were replaced under warranty with no hassle. AMD advised that the 1700XP had a small defect in the die that may have happened when mounting the heatsink or may have been a manufacturing defect, but as they couldn't tell they replaced it anyway.
Edit: For reference that's across 50+ CPUs from both my own and the Mrs machines, plus those we've built for family etc.
Edit: For reference that's across 50+ CPUs from both my own and the Mrs machines, plus those we've built for family etc.
Soldato
Mainly because they are built so well, and made from materials which will take the heat without issue. CPU’s right now are still one of the hardest things that we can make, there isn’t any room for error on the manufacturing side.
Every single piece in it will take its whole lifespan at 1-2c under thermal throttle.
Like others have said it’s one item essentially. A cpu doesn’t have to worry about power as it’s fed from the mobo which is much more likely to fail.
It also doesn’t have to worry about solder as it’s in a socket. GPU and other such dies if overheated or ran hot for a period of time, especially older ones can crack their solder connections between pcb and die rendering them useless.
Every single piece in it will take its whole lifespan at 1-2c under thermal throttle.
Like others have said it’s one item essentially. A cpu doesn’t have to worry about power as it’s fed from the mobo which is much more likely to fail.
It also doesn’t have to worry about solder as it’s in a socket. GPU and other such dies if overheated or ran hot for a period of time, especially older ones can crack their solder connections between pcb and die rendering them useless.
Vast amounts of money have been spent on R&D on this aspect with things like BGA, etc.
