>From: Kevin Schoedel <schoedel@kw.igs.net>

>>And speaking of Qbus cards, definitely need to make multi-function cards,
>>if possible.  Save space, save energy, save a bunch of chemicals need for
>>extra cards, etc.
>
>If you do that, why not do it *first*? If you can start by producing a
>nice Qbus multifunction card at a reasonable price, I'm sure plenty of
>current VAX & PDP owners would want one, and it might finiance the CPU
>development.

Don't be surprised if something like this does turn up before too long!

>From: "J.S. Havard" <enigma@sevensages.org>

>As I think of this, one thing that I think we should do is to not only
>make a vax clone that out performs real vaxen, but how to do it at as low
>a cost as possible, while still having a good quality machine.  I believe
>somebody's estimate was US$5000-$7000.  This won't do!  I can barely
>afford everthing I have, as it is, and be able to eat.
[...]
>So, let's just think here, what can we do to make them not cost much?  

The obvious answer:  do it as a software emulator.  Hardware development is
*expensive* (not as if SW dev is that cheap either).  Even once that's paid
off (by whom?), actually building boards and putting together boxes around the
CPU design costs lots of money, especially in the relatively tiny quantities
we're probably talking about.  IMHO there's no way it's going to happen for
only $7K a box without losing money, unless someone orders a million of them.

Re putting some fast CPU on a board and using a ROM-based emulator, that's
excellent if you need a plug-compatible CPU board, but otherwise it's the
worst of both worlds.  It's not as fast as a hardware CPU would be if it
used all the latest magic tricks, but you still have to redesign the board
every couple of years as each series of CPU chips is phased out and the new
faster chips won't fit the socket you've put on your board.  Using totally
off-the-shelf computers to run an emulator isn't as satisfying as having a
custom design, but it has practical advantages;  every speedup and economy
of scale that comes about from the mass market, filters down and gives you
immediate benefits, without YOU having to build e.g. a new 66 MHz PCI bus
controller to graft onto your old design.  So you write the code once and
all you have to do for future host CPU generations is some fine tuning,
and meanwhile it just magically gets faster every year thanks to Samsung or
AMD or whoever.  Of course, I'm biased...  :-)

John Wilson
D Bit