On 29/09/2020 09:31, druck wrote:
> On 24/09/2020 05:14, paul lee wrote:
>> If you use a heatsink and a small fan you could get
>> down to the mid 40s. I suggest an iceCooler - I run in the mid 30s on
>> normal use...
>
> Is there any real point to getting it that low? The cooler is probably
> using twice as much power as the Pi.
>
> One of mine Pis in shed got down to 19.7C CPU temp during the winter, it
> didn't run any better than when it was 72.8C in the summer :)
>
> ---druck

Serious answer from an ex electronics engineer.
Semiconductors and indeed all components age. possibly the worst
offenders are capacitors with liquids inside - the old wax paper
condensers and modern electrolytics are two culprits that change values
with age and may fail completely.

This form of ageing is very much a function of heat.
As far as semiconductors go, there are two completely different
thermally related failure modes.

One is failure of the hermetic seals to the chips caused by temperature
*cycling*. Hot and cold cycles can also fatigue internal wires.

The second is electromigration. Over time the actual dopant ions
physically migrate in the junction area, leading to loss of performance
and eventually out of spec. behaviour. Temperature accelerates this.

Operation at junction temperatures near the limit of usually around
150°C will destroy a transistor in minutes,  not hours or years.
I have seen cases of CPU fan failure on SPARC and INTEL CPUs that have
destroyed the CPU altogether, although usually the users switch them off
when performance becomes erratic.

So in the limit, yes, high temperature and temperature cycling destroys
chips. Just how much an elevated temperature of say 80°C matters, is a
very hard thing to quantify.

I used to run power transistors in power audio amplifiers at over 100°C
case temperatures, they sizzled with a wet finger. I have had a Schottky
diode accidentally shorted across a battery physically unsolder
itself...without apparent damage.

So there is a real point, but how important it is? is very hard to quantify.

Its like SSD failure, everyone knows that writes destroy SSDs over time,
but how fast?

And then examination of SMART data suggest 'a lot slower than bearings
wear out in, or oxide flakes off, a conventional drive'

In reality I suspect your Pi will be obsolete before it dies of heat
exhaustion. :-).


--
If you tell a lie big enough and keep repeating it, people will
eventually come to believe it. The lie can be maintained only for such
time as the State can shield the people from the political, economic
and/or military consequences of the lie. It thus becomes vitally
important for the State to use all of its powers to repress dissent, for
the truth is the mortal enemy of the lie, and thus by extension, the
truth is the greatest enemy of the State.

Joseph Goebbels

--- SoupGate-Win32 v1.05