14

What are the most common causes of CPU failure?

Are there intermediate states between a perfectly functioning CPU and a dead one?

1
  • 2
    Common causes include heat, and incorrect voltages. So make sure your computer is properly cooled, and you have a good power supply, and a good Power protection. Cooling will be be poor if your computers is overly dusty.
    – Zoredache
    Jul 24, 2011 at 9:35

5 Answers 5

10

It may only require one transistor to fail before a CPU stops functioning -- and since there are millions of transistors in a modern CPU, you might ask why it doesn't happen more often.

And, depending where the transistor is located in the CPU, the effect can be different, but I don't think we can expect a graded decline in performance: a failure in the ALU may not be noticed until a particular instruction is executed, and some instructions would be executed less frequently.

So CPUS die suddenly when a transistor fails. This might be caused by defects in the computer chip which are stressed too much, so time may be a factor.

Excessive heat can cause the minute impurities in the silicon which form transistors to diffuse and change operating parameters. Heat is an unavoidable copnsequence of simply operating the transistors, so a lack of cooling may eventually cause failures.

Other reasons might include failure of interconnections within the package of the CPU chip, but manufacturers are always looking for improved packaging methods with more reliable interconnections and better heat dissipation.

7
  • 3
    Heat is by far the most common way. And you'd never know it was failing til you start getting random crashes and bugs. The only way to really tell is to debug the kernel. If you application is crashing over simple instructions and memory copy, then it is a dead giveaway. Either that or you are overclocking too hard :)
    – surfasb
    Jul 24, 2011 at 9:38
  • 1
    My point is, if you aren't overclocking in the first place, then the alternative is your CPU is failing. And I don't know about you, but if my computer is crashing while the CPU is switching a one to a zero, I'd call that a failure. . .
    – surfasb
    Jul 24, 2011 at 9:54
  • 1
    Yes, I was being pedantic. In common usage the computer fails if it can't reliably perform operations. I should remember too, that when people speak of a CPU, they may not mean the chip inside the big square package. I would, but that's a professional perspective.
    – pavium
    Jul 24, 2011 at 10:07
  • 1
    Have a look at my answer on superuser.com/a/456511/122961 about CPU aging considerations
    – PPC
    Aug 1, 2012 at 23:44
  • 1
    Bad power or power surges can kill a CPU too.
    – LawrenceC
    Apr 30, 2013 at 18:28
5

Honestly, there are no common causes of CPU failure... at least relative to other parts of your computers. The CPU is generally the most reliable part of a computer. They just don't fail that often.

Instead, things you should watch for to fail are those things with moving parts: traditional hard drives, optical drives, and fans. More recently, we need to add SSDs to this list as well, even though they don't have moving parts. Capacitors also have a limited lifespan, and so power supplies and motherboards, which both use capacitors, can be suspect. Sometimes you'll have a bad stick of RAM, too, but I'm never really sure whey they go bad.

And now, at last, only after looking at most everything else in a computer, we come to the CPU. Even when a failure does occur, it's usually because the cooling fan (moving parts again) went bad first, and the CPU overheated as a result.

3

Among the other causes stated here, there can also be a broken internal connection. Several different techniques are used to tie the internal "chip" leads to the external package leads, and all of these are subject to possible failure.

This sort of failure could possibly be the result of overheating, and the likelihood of the failure increases with "thermal cycles", even in the absence of overheating. The failure may start out intermittent (though usually resulting in a hard crash when it happens) but get more and more persistent as the system is cycled.

This sort of failure mimics the failures seen from poor package/socket connections, etc.

[Added:] And I notice that "whiskers" have not been mentioned. A big problem with ICs and very tiny printed circuits is "whiskers" of metal that grow out of the plated wiring and short between adjacent "wires". This is especially a problem when you take out all the lead (see "RoHS"), as lead is commonly added to the wire alloys to prevent whiskering. This problem grows worse with increased temperature, of course.

3

In my experience, heat. How/why? Too much thermal paste! Many (most?) people know they need some thermal paste, but they may not realize how little they should use.

The rule is use as much as the size of an uncooked grain of rice, believe it or not.

Although the paste is about 10x better than air at conducting heat, the copper of the heatsink is 10x better than the paste, so you want it as close to the CPU as possible. The paste is truly only to fill in VERY TINY cracks so air isn't in there.

2

An interesting article regarding the subject of »Transistor Aging« appeared in IEEE's Spectrum magazine (http://spectrum.ieee.org/semiconductors/processors/transistor-aging). It lists several basic mechanisms that can lead to failure of an indiviual transistor, which may in effect reduce the full chip's computing power to that of a potato (or brick).

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .