Archived from groups: microsoft.public.windowsxp.basics (
More info?)
Constant operation causes heat sensitive components to wear
or oxidize when powered. This destructive wear from too many
hours of operation makes hardware tend to fail on powerup.
Others then blame powerup rather than hours of operation.
Yes power cycling is destructive. And then we apply
numbers. 15 and 39 years. These numbers today are also
higher than numbers from the first small disk drives on 1980.
Numbers demonstrate that power cycling 'worries' are classic
urban myth. Too many observations not tempered by the
technical details - fundamental principles and the numbers -
create urban myths.
In the meantime, using personal experience of something
under 100 computers - almost all problems were with computers
left powered 24/7. Like the university experiment - it tells
us nothing useful, unless information such as what failed and
why is included. Autopsies are performed at the IC level to
learn why failures happen. Summary observations are not
sufficient and can create myths.
The Challenger exploded. That proves that god does not want
man in space? Without details and underlying theory, that too
could be proposed as a valid conclusion. Those who suggest
power cycling is destructive also cannot provide those details
- and the numbers.
Again a most damning example: If power cycling is so
destructive to a computer, then it is also equally destructive
to all expensive radios and TVs. Why power down those other
appliances? Either leave all radios, TVs, and computers on,
or power them all off when done. Consistency. One cannot
have it both ways. Disk drives wear most and therefore fail
due to hours of operation. A spec that most every component
manufacturer provides because hours of operation is the most
relevant number for failure. A disk drive with too many hours
of operation will wear and therefore experience failure most
often during power up. Those without the underlying knowledge
wildly speculate that power up did the damage when, in
reality, damage was due to too many hours of operation. Since
they never learn details, the naive just wildly assumed
powerup did the damage. This is how classic urban myths are
invented.
The numbers say something completely different. Damage from
power cycling becomes totally irrelevant once numbers put a
problem into perspective. Power it down or put it to sleep to
maximize value from that computer. After too many hours of
operation, a computer is most likely to fail on power up.
Powerup did not cause the failure. Too many hours did the
wear and damage.
If power cycling damages semiconductors, then power off
semiconductors when not in use. Digital semiconductors power
cycle constantly. Early Pentiums even went from less than 1
amp to more than 10 amps in microseconds. Far more
destructive than an AC power on. Even more nonsense is
massive expansion and contraction for thermal cycling. Please
show me one IC that failed because the substrate cracked.
Damage occurs during switching - during normal operation. One
example is electro-migration. AC power cycling does not cause
electro-migration.
Thermal cycling is many hundreds of degrees cycled many
times. And yet semiconductors are not damaged by this thermal
cycling. Now we are told than tens of degrees causes damage
that hundreds of degrees does not? Bull. Again, apply
numbers. More wild speculation that power cycling causes
damage - eliminated as soon as we apply a new perspective -
the numbers. If expansion and contraction caused transistor
failure, then it occurs when expansion and contraction
actually occurs - during manufacturing. That failure from ten
times more degrees just does not happen.
If thermal cycling is so destructive, then it occurs during
normal operation when most temperature changes occurs fastest
- during the so many less than 1 amp to more than 10 amp
demands for current. To avoid such damage, then don't leave
the computer on 24/7.
Those who claim powerup causes failures don't provide the
supporting numbers. No numbers means junk science reasoning.
Turn it off or put it to sleep when done - to maximize
computer value. Too much posted about 24/7 advantages is
provided without numbers - also called wild speculation or
myth.
The quick sound byte conclusion is that power cycling does
damage. Reality means the post must be long. Must provide
underlying principles and numbers. Turn it off or put it to
sleep when done - once we replace myth with long posts based
on science principles and experience.
Ron Martell wrote:
> The last definitive numbers that I saw on this were from the mid-1980s
> and were based on a study of computers at a University. These were
> IBM AT (80286 CPUs) models. One group of the computers were installed
> in a computer lab where they were turned on at the beginning of each 1
> hour class and turned off at the end of that class. The other group
> were installed in administration and faculty offices where they were
> switched on at the beginning of each work day and off at the end of
> the day.
>
> The computer lab machines began to encounter high rates of hardware
> failures (hard drives, RAM chips, motherboards, etc) after 18 months
> of use while the admin and faculty office machines were pretty well
> failure free after 3 years of use.
>
> Of course hardware reliability has improved by at least one order of
> magnitude since the mid 1980s but I believe that the factors
> identified with regard to the above noted study are still relevant.
> These specifics include:
>
> 1. Hard drives contain electric motors, and like all electric motors
> are under the greatest load and therefore the most stress when they
> are first powered up. The vast majority of electric motor failures of
> all kinds, including refrigerator compressors, washing machine pumps,
> etc. occur when the machine is powered up not while it is actually
> running.
>
> 2. Electronic components are comprised of different layers of
> materials. When power is applied to these components they become
> heated and with this heating there is expansion. However the
> different materials have different rates of expansion and therefore
> when they expand there will be stressed placed on the joins between
> these materials. And when the power is turned off the materials
> contract and the stresses are relieved. Repeated stressing and
> unstressing an object at the same point gives rise to a condition
> known as "metal fatigue" and the stressed item is likely to break or
> crack at some point because of this. Such a breakage or cracking
> within an electrical components would, of course, most likely result
> in the total failure of that component.
>
> But as "Bill" pointed out in his response the improvements in hardware
> reliability means that computers will be disposed of due to
> obsolescence long before these hardware effects reach any sort of
> critical level.