Can a CPU degrade with time?

Can a CPU degrade with time?

Other urls found in this thread:

rkirchhof.com/Electromigration.html
en.wikipedia.org/wiki/Electromigration
minuszerodegrees.net/vcf_motherboard_failure_history.htm
classic-computers.org.nz/blog/2010-09-04-fixing-an-atari-400-screen-issue.htm
vcfed.org/forum/archive/index.php/t-52813.html
twitter.com/NSFWRedditVideo

No

Yes

It can and it does. Passing electrons through transistors will wear them down over time. Fortunately as long as your processor does not get too hot then it should be able to outlive you. You're very likely to have a dead PSU before your processor dies.

Yes. Especially when you or the motherboard fucks with the voltage for whatever reason.

>degrade
No.
It dies, it doesn't degrade. Only hard drives degrade.

It does degrade...like said, electrical resistance will kill transistors very very slowly

Dam, that sounds very... degrading.

And like he said, not in our lifetime. Outright failures are much more likely to kill it.

>tfw will never need to upgrade my haswell for the foreseeable future

Yup! I was't refuting how slow it is... just emphasizing that it does happen

Yes ICs can fail through electromigration which is the result of overheating causing the connecting wires in the chip matrix to melt and produce an open circuit. This is more likely to happen the more power and heat the chip uses/generates.

rkirchhof.com/Electromigration.html

This guy claims you can repair it by baking the IC in an oven for a few hours. No idea if it actually works, but sounds cool.

>tfw will never need to upgrade my sandy bridge for the foreseeable future

Yes

en.wikipedia.org/wiki/Electromigration

Actually CPU failures are very rare because of the nature of the chip; it's essentially a giant collection of logic gates. The most likely ICs to go bad are RAM and GPUs.

minuszerodegrees.net/vcf_motherboard_failure_history.htm

From VCFED--this was a list of repairs on IBM PC, XT, and AT boards which were all 25+ years old and there is not a single CPU failure here. The vast majority of faults were caused by caps or RAM.

So in all likelihood, a CPU will probably outlive you while RAM or your NVIDIA might not.

He's asking if performance degrades.
Does it?

>or your NVIDIA
Linusshilltips did a video, paired an old day 1 880 or something with a boxed one. No performance difference.

No. ICs are digital devices meaning the chip either works or it doesn't.

Does that make hard drives analog?

That wasn't the argument. The argument was that GPUs are relatively unreliable as far as ICs go.

Of course not.

>disable a core on a dual core CPU
>only half works
Checkmate

Yes. Overclocked chips that require slight increases in vcore after a couple years use is such an example. It's normal

But their performance degrades terribly.

stupid question, but why are RAM and GPUs unreliable

Overclocking (dumb thing to do anyway) is likely to cause electromigration damage which causes the bonding wires in the IC to weaken. As this happens, the IC leaks current, will have to draw more to operate properly, and also get hotter. In short, if a chip is suffering electromigration, it will run at an abnormally high temperature prior to failing like a star swelling into a red giant before it goes supernova.

Still not analog

All things fail or become obsolete eventually. If you look at it from a cost/utility perspective it is most certainly not a dum thing to pverclock! Ask anyone still running a 2600k at 4.5 after 5 years

But it doesn't maintain performance, making invalid.

I never said was valid.

GPUs get hot enough to roast weenies on. This is true of modern ones, it was true of 80s ones like the VIC-II and ANTIC. They nearly always need a heat sink and/or fan.

As for RAM, this is mostly because of the high density of the chip matrix in a DRAM and its construction, which includes a capacitor in each cell. SRAM for comparison is much more reliable simply due to being made of flip-flops.

You're just being obtuse, then.

Not at all. You asked if HDDs were analog, I answered the question.

There was no reason to have this argument if you had indicated that.

Okay, kid

He's wrong. But you can "fix" tantalum capacitors by heating them. By "fix" I mean temporarily, they'll still shit out afterwards.
A computer that doesn't work might start magically working for a while if you heat a tantalum capacitor that has failed.

Heating the chip doesn't do shit.

will this actually work or is it one of those old wives tales about improving your digestion by eating some poisonous mushroom you found in the forest

I have no idea, but no harm can be done by trying it.

If you have a failed tantalum cap, it might work. If there is something wrong with the chip or solder, it won't do anything.

Au contraire, baking dead graphics cards in an oven is a well known trick to revive them.

It's worked on certain GPUs, but if you don't know the melting point of the solder used on the specific device then you're fucked.

Basically, the link is saying that if a connecting wire failed from electromigration, that heating it with no bias (ie. current) will melt the wire back into place. Again, no idea if this is true or not but no harm can be done by trying it.

Yes OP. The transistors get worn our and sloppy after being used too many times. This is because the electrons gain extra electromagnetic charges after passing through too many hyperloops. It's kind of like how your dick gets small and shriveled up after fucking too many tight pussies.

>Yes OP. The transistors get worn our and sloppy after being used too many times

Absolutely false. Transistors do not and cannot degrade from use. The primary method by which an IC fails is as I said, the connecting wires between transistor pairs melting and breaking like a light bulb filament.

You're definitely not melting any "wires" inside a chip. They are made of copper, which has melting point above 1000C.

Old ICs from the 90s and earlier used aluminum connecting wires though. Modern stuff has mostly switched to copper to cut down on electromigration.

What? IC means integrates circuit
Maybe has a digital controller but is analog at the root of it

Evidently if electromigration is true, it takes much lower temperatures than 1000C to melt the things.

Electrification has nothing to do with melting. Electrons just fuck off from where they are supposed to be.

It has everything to do with melting. That's just physics 101 - any time you heat something, you cause the molecules in that something to move faster.

This is basically what does happen in an IC. Enough heat causes the connecting wires to melt and eventually break.

Damn right, especially if you overclock it long enough

classic-computers.org.nz/blog/2010-09-04-fixing-an-atari-400-screen-issue.htm

This is a classic electromigration failure. His ANTIC spontaneously blew while running. If I were him, I'd have first tried the oven trick and only replace the chip if that didn't work.

Or you could just dust your computer out every 3-4 months...

no dumbass. long after all the stars have burned out and planets been sucked into black holes, the only thing that will survive in the entire universe is fucking CPUs

CPUs are set to a factory speed based on an ambient temp of 90 degrees F

If you dont have 90F ambient temperatures, you are retarded if you dont overclock

vcfed.org/forum/archive/index.php/t-52813.html

This as well.

my amd won't stick with 3.75GHz overclock anymore.
Need to come down to 3.6

Only if you forget to wash it under the faucet every month.

>Will running electricity through something, causing it to thermally expand and contract, cause it to degrade?

Obviously, yes.

Keep in mind that baking chips in an oven will absolutely not work if it was ESD or overvoltage from a bad PSU/caps.

Troll post. It degrades from heat if you want to be exact. Unless you assume the chip is in use, time will do nothing with proper storage.

Everything degrades with time. Even the entropy responsible for this phenomenon is not immune. That is how the universe dies.

I've seen quite a bit of evidence that Commodore ICs like the VIC-II and SID tend to blow from electromigration since they used 12V power and get burning hot. The damn things should have been heat sinked from day one.

Sometimes but don't forget a bad PSU or caps blowing out ICs, especially RAM.

But again, despite all that, it's unusual for the CPUs to go bad.

Would a cpu survive an emp wave?

Just like I said. TTLs quite rarely fail and a CPU is just a giant collection of TTLs. Graphics processors and DRAM are the most common ICs to fail.

I remember one VCFED thread where a guy said the Z80 in his TRS-80 Model 4 spontaneously failed. It's rare, but not unknown. The good thing being they still make new Z80s for the embedded market (in a modernized CMOS variant).

Yes, the only thing that might be affected in a huge EMP blast might be the hard drive being erased and RAM/CMOS being cleared if the computer is still running.

There's a documentary detailing the process but I forgot what its called.

TRS-80s luckily don't have many ICs in them that can't be easily replaced and you could implement the whole thing on an FPGA. It's not like Commodore and their golf bag of custom ICs.

After reading a couple of Commodore PET restoration threads on VCFED, I've heard of virtually every component in the whole thing going bad except the CPU; I don't remember any instance of that happening.

Have you tried it?

...

Not that one would witness it. I've been building PCs for 20 years now and never had a CPU fail under normal use. They outlive their usable life by far.

That one time I overclocked an AMD thunderbird and it cracked cleanly in two pieces with smoke, but I've never gotten a non-working CPU if it hasnt been broken with misuse. And I've been building 5-10 PCs a year all this time. Sometimes I buy older xeons for workstation builds and sell those to clients too, no one has ever complained.

not really
only when you overclock a LOT will you notice it needing more voltage for the same clocks.
but even then it takes at least a year, usually 3 or more.

I know that there are some series of old TTL logic (74xx series) that will fail after some time because some crystals start to form on the bondwires. I've had a function-generator that suffered from this issue. Some of the logic gates stop working.
Also transistors from the 50's and 60s, e.g. AC126,AF116 etc, suffer from similar issues.

To be fair, those old ICs made on a larger process are more robust against electromigration. The new ICs might not last as long.