[rescue] This Just In: HP to buy Compaq

[Greg A. Woods]

[ On Sunday, September 16, 2001 at 02:57:14 (-0500), Scott Newell wrote: ]
> Subject: Re: [rescue] This Just In: HP to buy Compaq
>
> >The Intel specs require the PPro case temperature be at 85 C for ideal
> >operating conditions (I can't find any spec of the junction-to-case
> 
> What?!  I see that the spec for operating case temp is 85C max (range of
> 0-85C), but I certainly don't read that as the 'ideal' temp.  (Table 11-5,
> Power Specifications, Pentium Pro Family Developer's Manual, vol 1)

Hmmm....  OK, I missed that table  (man do I ever hate electronic
versions of books, especially as PDFs!)

I guess they've a firm handle on what the junction-to-case thermal
resistance is, but I wish they'd state it anyways as it makes
understanding the cooling parameters much easier.

(I am somewhat surprised that they give a minimum case temp of 0C.
That's got to mean the min junction temp is not much higher (maybe 10C)
and that's lower than I would expect.  I finally see where they mention
the THERMTRIP# pin and the internal sensor being set to trip at ~135C,
(which is a bit lower than older ICs would need).  The storage temp
range though is almost MIL-spec IIRC at -65C through +150C!  Quite an
achievement for a package that sophisticated!)

It still surprises me that tables 14-2 and 14-3 would be worked out for
achieving an 85C case temperature.  Obviously this means that once you
know your heat-sink and its fan airflow speed you can use the values in
those tables to set the cut-off temperature for ambient air.

Unfortunately few system designers measure ambient air temp in the
heat-sink area, nor do the heat-sink fan makers supply air-flow speeds
at given rotational speeds (even though many now have sensors for
detecting the fan RPM), and even when air-flow speeds are given they're
in CFM not LFM so presumably you've got to work that backwards given the
fan diameter, etc.  No consumer would ever do this math.  Indeed no
clone maker I know would ever do that math either.  Maybe Compaq, Dell,
HP, IBM, and the former DEC engineers would, but that's about it.

Indeed most socket-8 and socket-370 motherboards with temperature
sensors measure the ambient temperature under the CPU chip (i.e. in the
void in the centre of the socket) (or worse off on the die of some LM78
chip mounted usually down between two card edge connectors).  Despite
the lack of air flow the temp in that void is unlikely to ever exactly
match the case temp.

Now if Intel would smarten up and put a set of termocouples on the die,
or maybe mounted properly on the inside of the die cover, then there'd
never be any question of when the chip is getting close to overheating.

-- 
							Greg A. Woods

+1 416 218-0098      VE3TCP      <gwoods at acm.org>     <woods at robohack.ca>
Planix, Inc. <woods at planix.com>;   Secrets of the Weird <woods at weird.com>