🚀 go-pugleaf

biggfredd@NOSPAMcomputech-online.net wrote:
[snip]
> When factoring in design costs and the costs of building new facilities, etc
> (that is, unit cost accounting) those figures are certainly correct.  But as
> my original post was worded, the materials (only) in a chip probably aren't
> more than 50 cents.  Put another way, the incremental cost of the next
> additional chip isn't very much.

  The incremental cost of the next chip is certainly more than 50 cents.

  You are correct in that raw materials are cheap (silicon, oxygen, and
  aluminum are some of the most common elements in the earth's crust) but
  this is like an insurance company trying to settle with someone's
  benefactors for $10 because a dead body only has $10 worth of chemicals
  in it.

  You start with sand (special kind) and reduce it to metalurgical grade
  silicon.  You then zone refine it to an utterly incredible level of
  purity to obtain electronic quality silicon.  You then have to melt
  it and then pull a 8" diameter single crystal boule from the crucible
  for slicing into a wafer.  The wafer is polished and possibly gettered
  and at this stage is probably worth several tens of dollars.  You then
  run it through a fab where it goes through many dozens of repetitive
  steps of lithography, etching, implantation, polishing etc using very
  expensive machinery, very expensive chemicals, and tons of high purity
  deionized water, in an expensive, vibration isolated and highly filtered
  clean room facility.  The finished 8" wafer processed in say a 0.25 um,
  5 level metal CMOS process has probably cost you several thousand dollars
  so far.  You have to probe the wafer (i.e. step a probe card across the
  wafer stopping at each potential die site to run basic functional tests
  and ink the bad die while occupying a multi million dollar tester).  You
  have just added hundreds of dollars of cost to your now tested wafer.
  The probed wafer is now shipped off-shore for slicing and packaging
  (maybe several cents per pin or more depending on package type).  You
  now have to retest each packaged die but with a longer, more extensive
  final test program that occupies the many hundreds of dollar per hour
  tester for possibly several minutes per device.  For a PC class micro
  the variable cost is probably $40 - $80 per device and twice that for
  a workstation/server class device.  Not quite 50 cents!


All opinions strictly my own.

--
Paul W. DeMone                 The 801 experiment SPARCed an ARMs race
Kanata, Ontario                to put more PRECISION and POWER into
demone@mosaid.com              architectures with MIPSed results but
PaulDeMone@EasyInternet.net    ALPHA's well that ends well.