Re: [Phys-l] computers

[John Denker <jsd@av8n.com>]

On 08/07/2007 08:56 AM, Anthony Lapinski wrote:
> Not sure if this is really a "physics" question. My old laptop died (Mac),
> and I now have a new one. I am wondering if it should be shut down every
> night, or simply be left on "sleep mode." The manual states to "shut the
> computer down if it won't be used for a day or two." The tech guys at my
> school say to keep it on at all times because the heat up/cool down cycle
> causes the most damage.
>
> Any thoughts or personal experience with this? What should be done to
> extend the life of a computer?

Experience?  My experience includes being responsible for hundreds 
of computers deployed in all sorts of nasty environments, and also 
consulting for people who did research into semiconductor reliability.

My experience is that there is not much you can do to extend the life
of a computer.  The vast majority will become obsolete before they
break.  I just threw out a ten-year-old laptop that still worked the
way it did when it was new.

The part most likely to fail is the /fan/.  Obviously it is one of the
few moving parts.  Also it is just asking to get jammed by dust and
cat hair.  I have resurrected more than one laptop by blasting the
krud out of the fan with a shop-vac.

Desktop fans are less vulnerable than laptop fans ... and easier to
clean ... and easier to replace.

So:  At one extreme:  I don't shut down my desktop machines overnight.
I don't even shut them down for vacation.  I have desktops with multiple 
years of nonstop uptime, dating from the last time they were moved 
from one office to another.

Uptime is a convenience issue (not a longevity issue).  I don't want 
to wait for the things to boot up when I need them.  I certainly 
don't want them to be down when I'm trying to log into them remotely.

At the other extreme:
There is no way you can hurt electronics by cycling it on and off.
The electronic parts are made of things like silicon which is an
incredibly hard material.  Also the thermal gradients associated
with normal operation are huge compared to anything you could
do with the on/off switch.  (This does not count the filament
in a CRT display, which could be hurt by lots of cycling.)

My recommendations:

1) Back up your files.  Implement a program that backs them up 
automatically and frequently.  Do it now.

2) Don't worry about other hardware longevity issues.  Do
whatever you find convenient concerning on/standby/off.