Path: utzoo!utgpu!news-server.csri.toronto.edu!rutgers!mephisto!bloom-beacon!eru!luth!sunic!mcsun!ukc!inmos!brac!davidb
From: davidb@brac.inmos.co.uk (David Boreham)
Newsgroups: comp.arch
Subject: Re: Peltier cooler
Keywords: Velox ICECAP
Message-ID: <8856@ganymede.inmos.co.uk>
Date: 27 Jul 90 07:44:58 GMT
References: <10022@pt.cs.cmu.edu>
Sender: news@inmos.co.uk
Reply-To: davidb@inmos.co.uk (David Boreham)
Organization: none
Lines: 63

In article <10022@pt.cs.cmu.edu> lindsay@MATHOM.GANDALF.CS.CMU.EDU (Donald Lindsay) writes:
>
>I just got some literature from Velox (in Santa Clara) about a
>product they call an ICECAP.
>
>It turns out to be a box, some inches deep, which uses a solid state
>technology to cool a chip to zero degrees Celcius.  Peltier-effect
>refrigeration has been known since 1838, so the technology is
>believable.  It's intrinsically inefficient: one puts 3 to 6 watts
>into the box for each watt that one wants transferred from the chip
>to the heat sink.
>
>The literature claims that they've used it to run an 80486 at 50 MHz.
>Does anyone know if this is a reasonable claim?
>Don		D.C.Lindsay


Yes, sounds reasonable.

1) Peltier heat-pumps are used commonly to cool down
   various hot bits of electronics such as laser diodes
   in optical fiber systems. As long as you've got somewhere
   to dump the heat, then they work quite well. The datasheets
   I have go up to a unit offering 68.8W of pumping and a 
   max temp differential of 65 degrees C. That is 40 by 40
   milimeters, which is about the size of a 486 package.

2) CMOS devices like the 486 will definitely run faster
   if cooler. It's common to find vendors shipping samples
   of fast parts with ``relaxed temp spec'' for instance,
   because they really don't have a fast enough part.

   A couple of years ago someone was rumoured to have 
   an IMST414 running at 50MHz in liquid N :)

   The trouble with this kind of dodge is that nobody is
   going to trust the thing to work under all conditions.
   It is not the case that all the ``speed related'' 
   characteristics of the device track temperature in the
   same way. So, you are going to be requiring a VLSI tester
   and intel's test programs to test devices at zero C
   and speed bin. Then you're going to to have to do 
   some kind of qualification testing to ensure that 
   the asumptions made in the testing still hold when the
   maximum clock speed is used at low temperature. 
   You would also need to do a fair amount of thermal
   work on the packaged assembly to find out what the
   die temperature varies between given an ambient temp
   range and different device dissipation figures.

So, yes it is realistic but only really if you can
do a large amount of re-testing and qualification or
if you can persuade Intel to ship parts tested for
50MHz operation with a heat-pump.

BTW these pumps cost about $50, what are these guys
charging for their unit ?



David Boreham, INMOS Limited | mail(uk): davidb@inmos.co.uk or ukc!inmos!davidb
Bristol,  England            |     (us): uunet!inmos.com!davidb
+44 454 616616 ex 547        | Internet: davidb@inmos.com