Xref: utzoo comp.sys.ibm.pc:46242 comp.sys.intel:1167
Path: utzoo!utgpu!jarvis.csri.toronto.edu!rutgers!mit-eddie!uw-beaver!zephyr.ens.tek.com!tektronix!psueea!parsely!bucket!loop!keithl
From: keithl@loop.UUCP (Keith Lofstrom)
Newsgroups: comp.sys.ibm.pc,comp.sys.intel
Subject: Re: Burnin' hot math co-processors - heat sinks
Message-ID: <39@loop.UUCP>
Date: 9 Mar 90 19:08:07 GMT
References: <1990Mar6.033339.2016@cs.dal.ca>
Reply-To: keithl@loop.UUCP (Keith Lofstrom)
Organization: Launch Loop, Portland, OR
Lines: 37

In article <1990Mar6.033339.2016@cs.dal.ca> lane@cs.dal.ca (John Wright/Dr. Pat Lane) writes:
>
> ...
>I asked him about putting a heat sink on the chip and he recommended 
>against it, especially gluing one on with epoxy, since this might disrupt
>the chips own heat dissapation ability and make things worse.
>
>Now this might be a conservative, "if-it-isn't-in-the-book-don't-do-it"
>response.  What do the engineering-types here on the net think?  Anyone
>have any experience in this area?
>...

Putting heat sinks on things might possibly break them if they weren't
designed for it.  Probably not, but possibly.  I suspect this is simply
C.Y.A.;  if the heat sink causes something to fail, Intel might be liable
if they told you to go ahead, and loses nothing by telling you not to.

Silicon, ceramic, plastic, and metal all have different coefficients of
thermal expansion.  When a packaged IC changes temperature, this creates
stress in the materials and the adhesives bonding them - I've seen large
die pop loose from their epoxy die attach because of thermal stress.
Still hanging on the bond wires, but liable to get VERY HOT with only
conduction through the bond wires to cool them, killing the chips.

There is a small possibility that the Intel packaging engineer designed
the part so that particular combination of metal and ceramic would bend
and flex with temperature in a special way - more likely they just tried
it and it worked, and changes in the formula didn't work.

Note: if you do try a heat sink, find a sheet of thermoset epoxy to mount it.
Trying to smear on liquid epoxy may create bubbles and voids which cause
stress and lead to failure - another reason most people shouldn't add their
own heat sinks.  And be sure to let the rest of us know how it turned out!

-- 
Keith Lofstrom    keithl@loop    tektronix!psueea!qiclab!loop!keithl
Launch Loop, P.O. Box 1538, Portland, Oregon 97207   (503)-628-3645