[comp.sys.mac] 20 Megaflops in Nubus, and a question/Nubus cards

jeffb@eleazar.dartmouth.edu (Jeffrey C. Buchsbaum) (06/06/89)

Two things to say, a question and a bit of FYI stuff for the scientific 
users of MacII family boxes.

I have a National Semi. ram board that is great except for one thing, it 
will not "boot" unless I start from a floppy or insert a floppy during my 
hd's startup/init loading period.  Oddly, the same setup fails to boot the 
ram disk board (Nubus board, 8 meg, 32bit) without the floppy, unless all 
inits are removed, even the init that supposedly runs the board....(?)
Paul Mercer wrote the board's init, so if he gets this, I hope he can email 
me a patch of some sort. 
System
MacII, loaded with inits, 5meg, 120 meg in two hd's, LaCie hard disk software
(silver lining)....  

The FYI stuff is about a Nubus board made by Mercury Computer of Lowell, MA
(508-458-3100, ask for marketing, Paul Martino for academics, I think).  It 
comes with C or Fortran dev. systems, assembly, lots of utilities (debuggers....),
support for the MPW 3.0 environment, support for Think C 3.0x, and a very wide
array of scientific and mathematical algorithms that are in the board's 
microcode.  The board uses WEITEK r i s c chips and has three main processors,fpu,
int unit, and process controller.  The code gets parallelized as you compile it, 
flags galore, and you can optimize via vectorization.  I just confirmed a bench 
we developed at Dartmouth(plasma physics) that took the following times on
various machines:
(by memory of last fall, so +/-.5 hrs...)

VAX 11/785  =   14 hrs BSD unix 4.3
VAX 11/785  =   13.5 hrs VMS 4.?
Convex C1-XMP 2.42 hrs   
Mercury MC3200nu (today) = 2.30 hrs
and things like Alliants with one and two CE's that were about 3-6 hrs(I forgot)

The above code was not optimized, but I estimate it reflects about 12 Mflops. Most
impressively was that it ran completely on the board (the board can be a slove or
a master in their terms).  Since I was in Multifinder, I chugged out of MPW and
did other things.  My cpu saw NO slow down since ALL the processing is taking place
on the board.  You can run up to five boards (100+Mflops) in a parallel envronment,
and my guess is that each could be autonomous.  One is plenty for me.  

The company supports academics, and has a program in place for such people. I
visited the company and got a tour of the entire operation, from shipping, to 
marketing, to the burn-in area (a big box that was quite warm +130 degrees f),
to the engineering and support areas.  It was very clean, professional.  What I
had alway guessed these places looked like.  Their support is superb, and the 
whole experience I have had with them has been fantastic.  Installation of all
software is automated....  I highly recommend getting info on this board for the
macII family.

They make sun, ibm, and vme boards too.