Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!crdgw1!uunet!convex!swarren
From: swarren@convex.com (Steve Warren)
Newsgroups: comp.sys.amiga.misc
Subject: Re: 486SX
Message-ID: <1991May03.160136.6267@convex.com>
Date: 3 May 91 16:01:36 GMT
References: <1991May2.212247.12525@dartvax.dartmouth.edu> <3645@borg.cs.unc.edu>
Sender: usenet@convex.com (news access account)
Organization: CONVEX Computer Corporation, Richardson, Tx., USA
Lines: 26
Nntp-Posting-Host: neptune.convex.com

In article <3645@borg.cs.unc.edu> cullip@sargent.cs.unc.edu (Timothy Cullip) writes:
>I don't want to start an "I hate Intel thread" here, but I just can't
>resist giving my impressions of the 486SX.  Intel sells the standard
>486DX chip for about $500.  The SX is in fact a DX, but with the floating
>point unit turned off (i.e. it's on the chip, but disabled so it's just 
>as expensive to produce) but sells it for about $250. Kind of gives you 
>an idea of the profit margin on the DX.

I know this is hard for you to believe, but a major portion of the expense of
a hyper-integrated chip (>1million gates) like the 486DX is the testing and
the scrap costs.  Any 486DX chips that passed the integer diagnostics but had
a flaw in the FP unit are going in the trash unless they can be sold as a
486SX.

There is a lot of savings when you reduce your test time by not requiring a
test of a major portion of the chip like the FP unit.  Also, your scrap rate
goes down, and it may go down significantly if the FP unit is the most
sensitive part of the chip (which seems likely, since the FP unit is always
getting pushed the hardest for performance).  Not testing the FP unit could
conceivably cut their scrap rate in half, which would mean that it really
does cost half as much to produce.
            _.
--Steve   ._||__      DISCLAIMER: All opinions are my own.
  Warren   v\ *|     ----------------------------------------------
             V       {uunet,sun}!convex!swarren; swarren@convex.com
--