Mysterious crashes - solution?

Ian Molton spyro at armlinux.org
Thu May 9 12:01:56 PDT 2002


On Wed, 8 May 2002 09:18:09 +0100
Declan Moriarty <declan.moriarty at ntlworld.ie> wrote:

> Was it Ian Molton who wrote on Wednesday 08 May 2002 10:25:

Yes it was :-)

> > Declan Moriarty <declan.moriarty at ntlworld.ie> wrote:
> > > Heat is only an issue as junction temprature is approached (150
> > > degrees C) by some component. In practise, if you can put your
> > > hand on things and hold it there, heat is fine.
> >
> > Hmm. not /quite/ true.
> >
> > CMOS chips doping degrades over time, and this decay is accelerated
> > non-linearly by increasing temperature.
> > what that means is that a few degrees rise in temperature can mean a
> > few/TIMES/ reduction in lifespan for the chip.
> 
> We all know heat cycling ages a chip.

Actually I wasnt thinking cycling,  I was thinking junction doping
degradation. Obviously X86 class high-temperature CPUs have more
resilient dopings (and the flip-chip packaging is MUCH better from this
POV)

> If there is real evidence that
> heat degrades cmos chips in a fashion we have to worry about (pcs are
> obsolete in 5 years anyhow) I expect the market would have a different
> attitude to the cmos technology.

I dont think all CMOS uses the same doping strategies. StrongARM CPUs
run cold, and so probably dont have expensive doping.

> > of course, not all chips are born equal. modern CPUs are designed to
> > run hot, I assume this is achieved with newer doping tech, among
> > other things.
> > the StrongARM in my RiscPC, OTOH is not designed to 'run hot', and
> > wont survive that sort of treatment.
> 
> The issue is not hot on the outside, but on the inside, at the
> junctions.

Of course.

> There is a 'thermal runaway' phenomenon, and where it
> applies, great care has to be taken to avoid it.

and again, the doping tech. is important here, although less so than
some other design considerations...

> We can only measure heat on the
> outside, which isn't much of a guide. So the good rule is if it's too
> hot to touch (~50 degrees C) it's too hot.

That doesnt work for TQFP plastic packages - the plastic is a poor
thermal conductor, and so the chip develops hot-spots internally.

> Because by the time it
> works back in through all the thermal resistances, your junction
> temprature could be up at 130 degrees - close to danger.

way beyond danger for cold-doped chips... SA is rated to 50 deg C MAX,
IIRC.

> Theoretical arguments aside (this isn't really the place), you would
> agree he needs a bigger fan on his cpu, or more cooling, wouldn't you?
> And that this alone will probably solve the problem?

Well, yes - it'd help, but as his room is VERY hot (according to him,
passive cooling might just not be adequate given the size constraints.

I think it may well be worth investing in a forced system, using (say)
peltier effect heat pumps.
-- 
Unsubscribe: send email to listar at linuxfromscratch.org
and put 'unsubscribe blfs-support' in the subject header of the message



More information about the blfs-support mailing list