🚀 go-pugleaf

In article <2851@eos.UUCP> eugene@eos.UUCP (Eugene Miya) writes:
>In article <7287@spool.cs.wisc.edu> shekita@provolone.cs.wisc.edu (Eugene Shekita) writes:
>>I was wondering how much you architects out there worry about our
>>microprocessor market falling prey to Japan.
>They have several micro-processor based projects projects, but they are
>moving cautiously.

The problem is that memories are the traditional vechicle for process
development.  Several reasons for this.  Memories are denser than uP's,
and therefore are more susceptible to defects.  To project target yields,
IC companies generally develop a curve for a process based on die area.
Different curves (or fudge factors) are used for memories and uP's.
Second, you really tend to push the layout rules and process parameters
in a memory to make that memory cell as compact as possible.  A small
gain in one cell yields big gains on the die when you step and repeat it
256K...or 1M...or 4M times.  Therefore, memories tend to force you to
develop and support aggressive design rules.  The importance of this is
cost.  A better process gives you better yielding dice and lets you shrink
to get more dice on that wafer.

So, you can discuss the issue of uP's from a viewpoint of design and
architectural dominance, which is probably the issue of greatest interest
to readers here.  However, process dominance can lead to ownership of the
merchant uP and controller market, and these "less interesting" processors
are the bread and butter which fuel R&D on advanced processors.

This is what makes the recent Intel/NEC copyright decision so interesting.

>Remember Pearl Harbor.  Remember the Alamo.  I am serious.  I now think
>we live in a racist society contrary to what I was taught to believe.

Japan bashing is in vogue.  And unfortunate.  It hides the real issues.
-- 
Chip Rosenthal     chip@vector.UUCP    |      Choke me in the shallow water
Dallas Semiconductor   214-450-5337    |         before I get too deep.