A question about CPUs, why they never die?

[Derek12]

Why are they so resistant and *almost* never fails on it's own, being one of the most complex pieces on a computer and what does the most job in almost all situations? I am searching polls all around Google and the CPU is the least component that die (besides my personal experience in which none CPUs died).

GPUs are more complex but they are almost idle outside gaming, video, and 3D modeling.

Many thanks

[Outback Bronze]

Yep true. Personally ive never had a cpu die on me and neither have my mates. What does it take to kill one?

[BarbaricSoul]

Originally Posted by Outback Bronze

Yep true. Personally ive never had a cpu die on me and neither have my mates. What does it take to kill one?

running it outside of it's manufacturers specs(over-voltage) or pairing it with low quality hardware like a cheap PSU

[specks]

My Pentium D 820 is now 5 years old and it is still going strong even if I use it everyday.

[Fourstaff]

QC for CPU is very tight, there are only 2 people making CPUs, Intel and AMD. In other departments, there are plenty of people competing, and when that happens price wars will cause people to take shortcuts in QC, and which is why we have inferior products dominating the market because of strategic compromises. I think the first component to go out usually is the PSU or graphics card (for me at least).

[Jizzler]

Don't think I've ever had one die from age. Laying around here I have a couple P4D, a Sempron, Prestonia Xeons, and even some ancient P3 and Celerons. They all ran up until the day I decommissioned the system (4-8 years of use). Some of them have seen 3 different motherboards (cheap ones usually blew caps after 6-12 months of 24/7 FAH).

[IlluminAce]

CPUs are self-contained units protected by a large heatspreader, which are installed into a well-defined socket design and typically left well alone. All that's left to worry about is cooling one area (the HS) and supplying stable Vcore. As for the design, it's not too difficult to write extensive tests for them - this can be done virtually. Plus, there are very few CPUs, and each sells an extraordinary number of units, unlike other components such as graphics cards and motherboards. The manufacturers have to make sure they work as stated.

Motherboards typically have so many different component chips, all which need to be cooled to differing degrees, depending on ambient case temp and usage (which can be hard to predict). There are plenty of places to cut corners too. Finally, during installation, it's easy to over-tighten MB screws, damaging the multi-layered design.

Hard drives, opticals and fans have moving parts. SSDs have a limited number of IOPs.

Anyone know why RAM failures aren't too uncommon?

[Frick]

Originally Posted by IlluminAce

CPUs are self-contained units protected by a large heatspreader,

Not in the olden days. A friend of mine crushed a corner of an Athlon XP Barton with the heatsink. The P4's of the day had them, but there are some that were naked.

[specks]

My father even has a p2 in his office and it still works!

[Frick]

Originally Posted by specks

My father even has a p2 in his office and it still works!

I have some of those as well, still functional.. Not very useful though.

[_JP_]

Originally Posted by Frick

The P4's of the day had them, but there are some that were naked.

Those would be the socket 423 Willamette P4s.

Originally Posted by specks

My father even has a p2 in his office and it still works!

As does my Pentium!!! (11 years of service).

Originally Posted by Frick

I have some of those as well, still functional.. Not very useful though.

As a router they're useful.

[lilhasselhoffer]

Originally Posted by Derek12

Why are they so resistant and *almost* never fails on it's own, being one of the most complex pieces on a computer and what does the most job in almost all situations? I am searching polls all around Google and the CPU is the least component that die (besides my personal experience in which none CPUs died).

GPUs are more complex but they are almost idle outside gaming, video, and 3D modeling.

Many thanks

Simple answer: CPU components are created at the microscopics level, and rely on material properties rather than physical properties to work. This makes them, functionally, the most durable component of a system.



Long answer: A CPU is a chunk of silicon, or other semi-conductor. It is etched and doped with other materials to make it into either a conductor or insulator. This process is either successful, or not. If it is successful, the material has new properties, which can only be destroyed if the material is chemically altered (through additional doping, burning, or other chemically changing process).

Everything else in a computer relies on both chemical and physical properties to work. Assuming that you do not have a basic understanding of electronics, I'll cover the basic components.
Capacitors - Conductor plates separated by a dielectric medium. This dielectric can be anything from oil to air, to glass. The dielectric constant of the dielectric influences the capacitors capacitance, and very small changes in the dielectric constant produce very large capacitance differences.
Diodes - Doped semi-conductors that largely only allow current to flow in only one direction. They have a breakdown voltage where they will not stop current from flowing, and always have a small amount of current that flows in the opposite direction, which is functionally nothing in simple calculations.
Coil/Choke/Inductor - A wire loop. These things are composed of wire wrapped around some sort of core. The core material matters a lot, but the biggest difficulty is finding out what the resistance associated with the wire is, and how this will influence the system.
Transistors - Doped semi-conductor that allows current to pass through only is a signal from another pin is present. These allow smaller voltage signals to control larger current devices.
Other IC - A varying combination of transistors, diodes, etc... combined into one package that performs a specific function.

The coil/choke/inductors generally never burn out. The ICs in a circuit also rarely burn out if not over-voltaged. Transistors and diodes fail, but very infrequently. They usually only fail when higher voltages are present, and the feedback in the system melts their shells.

What we are left with are capacitors and the boards themselves. Capacitors dielectric ages. Whether it be the heavy heating and cooling cycles, or the dielectric breaking down due to age, the dielectric constant changes. This negatively influence the capacitor, and will often partially vaporize the dielectric. If you've ever experienced this you'd either have the capacitor blow-up, or otherwise greatly swell. Then, we've got boom goes the capacitor, pop goes the circuit, and off goes the board.

The boards are even more fun. They can fail via heating/cooling cycles, pregnant traces, or simply poor layering. Starting with poor layering, the QC checks at the factory are basically to make sure everything initiallizes. The layers could well be intact and working there, but jostling during the long shipping process leads to layer separation. No contact=dead mobo. Pregnant traces appear over time. A minor deviation in the lead on the motherboard generates a point of increased resistance, which creates excess heat. The heat distorts the metal, and with continued heating the trace eventually pops. Pregnant traces are named because before they pop they generally create a bulge in the solder mask. Heating and cooling cycles are always an enemy. They create stresses in the board, which cause gradual failure over time. The only way to address them is to have an awesome cooling design to wick away all of the heat from your components. This is generally more expensive than just knocking a couple of years off your warranty...


So in summation; CPUs have maybe a couple points of failure. Motherboards, Graphics cards, and PSUs have thousands of points of failure. It doesn't take a genious to see why one fails more than the other.

[Derek12]

Many thanks to all, specially Lilhasselhoffer for the extensive and useful explanation

Before this, I'd never realized that semiconductors (specially VLSI IC's) were so reliable in comparison to less complex components like capacitor or PCB boards!

[Super XP]

My brother in law still running a 486 hard and strong.

[cadaveca]

Originally Posted by Derek12

Many thanks to all, specially Lilhasselhoffer for the extensive and useful explanation

Before this, I'd never realized that semiconductors (specially VLSI IC's) were so reliable in comparison to less complex components like capacitor or PCB boards!

CPUs (and all silicon for that matter) is judt glass and metal, which is inherently inert. Vacuum tubes don't die because the glass breaks...

Nevermind that there are both thermal and electrical protection built into the CPU, the board's VRM, the BIOS, and elsewehre to hlep prevent premature death. Of course, it's possible to bypass those protections, and some OEMs even design their boards so that this is possible, but most users aren't gonig to be changing/modding that sort of stuff anyway.

[Derek12]

Originally Posted by Super XP

My brother in law still running a 486 hard and strong.

I remember my old K6/2 used in my first computer months, in some day the computer started to went a bit mad: programs crashed sporadically with the usual "This program has performed an illegal operation" and some random reboots, but it was usable to browse internet and such. We though that the PC was infected so we scanned with some antivirus and it was clean, then a friend who had better computer knowledge than us reinstalled Windows 98 but the problem continues, he accessed the BIOS, checked the temperatures and the CPU was 110ºC, the fan was totally stopped (it was very silent anyways) and the heatsink totally clogged with dust (2 years without cleaning it sitting under the desk lol). After replacing the fan and cleaning the HS the computer returned to normal.

Short: the CPU was OVERHEATING in IDLE at 110 ºC during months[/U] and it lasted another 5 years until we bought a new computer. Incredible the CPU wasn't toast.

Originally Posted by cadaveca

CPUs (and all silicon for that matter) is judt glass and metal, which is inherently inert. Vacuum tubes don't die because the glass breaks...

Nevermind that there are both thermal and electrical protection built into the CPU, the board's VRM, the BIOS, and elsewehre to hlep prevent premature death. Of course, it's possible to bypass those protections, and some OEMs even design their boards so that this is possible, but most users aren't gonig to be changing/modding that sort of stuff anyway.

But it has millions of tiny transistors, capacitors, resistors and traces right? Don't they broke at a time in normal operation? This intrigues me

[cadaveca]

Originally Posted by Derek12

But it has millions of tiny transistors, capacitors, resistors and traces right? Don't they broke at a time in normal operation? This intrigues me

Sure, it can happen, and that's why overclcoking, in any form, isn't warrantied, and why I stress "stock" ram speeds and such. They're delicate, and the current pushed through them is usually a delicate balance of lifetime and speed.

The same sort of problem caused the recall fo the P67 B2 chipset, so it's definitely more of an issue as things get smaller, and this is why the industry is researching alternative methods of transistor design and construction.

[arnoo1]

i still have a p3 500mhz oc'ed to 667mhz running, windows xp on it runs just fine, also stock heatsink, the cpu is always hot but never dies

[Derek12]

Originally Posted by cadaveca

Sure, it can happen, and that's why overclcoking, in any form, isn't warrantied, and why I stress "stock" ram speeds and such. They're delicate, and the current pushed through them is usually a delicate balance of lifetime and speed.

The same sort of problem caused the recall fo the P67 B2 chipset, so it's definitely more of an issue as things get smaller, and this is why the industry is researching alternative methods of transistor design and construction.

Then if you don't OC nor have any PSU, temperature, issues, would a CPU (or any semiconductor) have a "infinite" lifespan theoretically? I read in Google that CPUs can "wear" and become "slower" over time with high usage, but I am not sure if that is true, seeing old CPUs working perfectly as some users commented here

[claylomax]

My E8400 half-died, well sort of ... After being at stock for the last year one day the computer BSODed and is been stuck at 2000mhz ever since. More here: http://www.techpowerup.com/forums/sh...d.php?t=148341

[Derek12]

Originally Posted by claylomax

My E8400 half-died, well sort of ... After being at stock for the last year one day the computer BSODed and is been stuck at 2000mhz ever since. More here: http://www.techpowerup.com/forums/sh...d.php?t=148341

hmm strange problem out there, seems at first the motherboard became mad

[cadaveca]

Originally Posted by Derek12

Then if you don't OC nor have any PSU, temperature, issues, would a CPU (or any semiconductor) have a "infinite" lifespan theoretically? I read in Google that CPUs can "wear" and become "slower" over time with high usage, but I am not sure if that is true, seeing old CPUs working perfectly as some users commented here

I suppose, that yes, it's possible a PCU could last forever, as there are no moving parts. However, there is this thing, called electromigration...over time, electrons can "tunnel" through imperfections in the circuit, causing failure. Electromigration can be increased by changing the amount of current, but on a "perfect" chip, with just enough current, it really could last forever.

Sadly, I do not know if current chips are designed that good. Many chips have numerous "redundant" transistors to deal with problems in the fabrication process, so I am not 100% ocnfident current chip would last forever, as the margins are so tight...

[Derek12]

Originally Posted by cadaveca

I suppose, that yes, it's possible a PCU could last forever, as there are no moving parts. However, there is this thing, called electromigration...over time, electrons can "tunnel" through imperfections in the circuit, causing failure. Electromigration can be increased by changing the amount of current, but on a "perfect" chip, with just enough current, it really could last forever.

Sadly, I do not know if current chips are designed that good. Many chips have numerous "redundant" transistors to deal with problems in the fabrication process, so I am not 100% ocnfident current chip would last forever, as the margins are so tight...

Many thanks again I wasn't aware of the electromigration issue

Lets hope the new CPUs and other semiconductors with reduced semiconductor process would be reliable to the long term