home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.arch
- Path: sparky!uunet!snorkelwacker.mit.edu!bloom-picayune.mit.edu!athena.mit.edu!solman
- From: solman@athena.mit.edu (Jason W Solinsky)
- Subject: Design Cycle Times
- Message-ID: <1992Sep12.005655.7652@athena.mit.edu>
- Sender: news@athena.mit.edu (News system)
- Nntp-Posting-Host: m37-318-1.mit.edu
- Organization: Massachusetts Institute of Technology
- References: <1992Sep9.061933.28304@news.eng.convex.com> <1992Sep9.213319.7894@fasttech.uucp> <1992Sep10.142410.3351@terminator.cc.umich.edu>
- Date: Sat, 12 Sep 1992 00:56:55 GMT
- Lines: 41
-
- In article <1992Sep10.142410.3351@terminator.cc.umich.edu>, sarr@sinshan.citi.umich.edu (Sarr J. Blumson) writes:
-
- |> Computer architects are SUPPOSED to design for process technology that
- |> will exist when the architecture is built, not what is available when
- |> they start.
-
- Moving away from little technical details, Have there ever been any
- microprocessor design efforts based primarilly on design cycle time. I don't
- mean efforts which thought about it, but efforts where the architecture was
- actually governed by design cycle time considerations. It seems to me that
- with speed doubling every year or so, that the ideal effort aught to be
- focused around reducing the design time.
-
- It seems to me that the ideal effort would begin by designing a microprocessor
- with a lot of room (logistically, not area-wise) for future additions. They
- wouldn't even bother laying it out, but would let the computer do that and
- immediatelly send it out for a quick fab run. Naturally the die size would
- be huge, but its only a first run and you can afford a really low yield. While
- that was gone, some people could go to work at hacking the circuits, while
- others could start laying out the circuits which appear to be finished.
-
- The first fab run would get back within a couple of weeks, since this design
- cycle time consciencous team would have made special arrangements with the
- process people for a quick turn around time. Despite this, the team may already
- have found a bug with their original design. Not to worry though, before the
- last run gets back, send out a new one. Its only $100K. The testing, redesign
- and layout would all go on simultaneously. Naturally a really large amount of
- effort would be wasted, but you are left with two advantages. 1) Your time to
- market is going to be greatly reduced (but they try to do this with your
- typical process anyway.) and 2) After you have a selable product, there is no
- need to take the design team apart. Why introduce a new chip every three years
- with twice the power when you can introduce a new chip every three months with
- a 20% increase. Don't even bother making them instruction set compatible. Just
- include a program which will instantly transform a program for one chip to
- another. This might make copyright protection a little difficult, but that's
- not your job. Let William Henry Gates the Nerd worry about that.
-
- Anyway, I'm interested in any architectural efforts where concerns related to
- these were the PRIMARY FACTOR governing architectural decisions.
-
- Jason W. Solinsky
-