In article <MPG.1beacf1f1d7b12f8989807@News.Individual.NET>,
Peter Smithson <pgsmithson_deletethisbit@yahoo.co.uk> wrote:

What was wrong with the FAQ entry - I know it's simplified but what's 
inaccurate?

It implies that overheating is a threshold effect, when any level of heat,
especially with power applied, will affect the rate of some premature
failures mechanisms.  The machine that was running 25 degrees hotter
will fail through electromigration at least twice as fast as when it
is relatively idle.  It's likely that it will fail even faster than
that, because the heat sink temperature rise will be less than the core
temperature rise.

Failures are also statistical, so that some machines will fail very
early even though, statistically, the failure may still be very unlikely
in the 5 or so years service life of the machine, and the reduction in
failures due to expansion and contraction might compensate for it in
the overall picture.