fab

[jfox@xxxxxxxxxx (Jeff Fox)]

Dear MISC readers,

Just an update, we didn't make the June submission date for F21
at Mosis.  After Chuck corrected some things that didn't work
on P8, and presumably would not have worked on F21 we started
testing with the simulator in OKAD and found some problems.
In fact as the date approached I was finding serious problems
faster than they could be fixed.  So the new target is July... 

It really brought home the notion that all specs are preliminary
until we get to production.  Watching the chip go through several
versions in a single day makes it pretty obvious.

If we had a large budget we would submit the design even though
OKAD still shows serious design probems.  If we did we could get
parameters back to fine tune OKAD.  There is so much new stuff
to be tested on F21 that we wanted it to be mostly working in
OKAD before spending money on fabrication.  

The real problems may be those that OKAD doesn't predict because
it may not yet correctly model the actually device produced by
the fab process at Mosis.  We expect that it will go smoother
than MuP21 which needed about 8 fab runs, but so far P8 has
had various tuning problems not unlike some early fab runs on
P21.  It is encouraging to see OKAD simulating a working CPU
and coprocessors, now if we can just get it in silicon.

Jeff Fox
Ultra Technology