Experience with MicroVAX II

[Avadis Tevanian]

As everyone now knows, DEC has officially announced the MicroVAX II.  There
have already been a couple of articles about this new machine, but far fewer
than I had expected to see.  So, here we go:

I have been fortunate enough to field test a MicroVAX II (serial # PROTO
112!).  I have been using my machine now for 2 1/2 months and have been
running Ultrix-32m 1.1.  My configuration:

	MicroVAX II CPU with Floating Point Chip.
	5 Megabytes main memory.
	DZ (4 serial lines)
	2 Floppies.
	2 RD52 hard disks (~30 Meg each) (not the newer RD53's).
	DEQNA (10MB ethernet).
	TK-50 tape streamer.
	QVSS (bitmap display) (just installed yesterday!)
	(let's see, what else could I want?)

2 1/2 months ago a couple of us were told that a MicroVAX II was set up in
our machine room.  We couldn't find anyone who knew exactly were it was at
the time, so went in search of it.  We walked through three of our machine
rooms, and finally found it hiding under a table.  At this time, the machine
only had a Decwriter attached to it, with about 5 sheets of paper in it.
What had scrolled out was the results of diagnostics, which claimed that the
machine had 5.0 Meg.  Well, we didn't believe it, figuring that DEC
diagnostics were screwed.  So, we then proceeded to look at the box of
floppies we had.  We found one labelled "Boot 1 of 3" and another labelled
"Boot 2 of 3" and inserted them into the floppy drives, typed "B" and
watched Ultrix boot (we were then convinced the machine really had 5 Meg).
It started asking us to insert more disks, we did.  About 1/2 hour later, we
were running multiuser Ultrix.  Enough of this anecdote, time for some hard
facts:

Speed:  The WORST CASE I have seen is 75% of a 780.  Usually runs about
80-90% of a 780.  Programs that make good register usage get 100% of 780.
More importantly, it "feels" like a 780 (with only 1 user!).

Floating point:  OK, lots of machines can get close to 780 perfomance, but
what about floating point?  I thought for sure that this would be the place
were the 780 would outperform the MicrooVAX... WRONG!  The floating point
chip is just as fast as a 780 with a FPA (actually, add/subtract are a
little slower on MicroVAX, but not much).  My instruction timings showed
differences of 20% or less in floating point speeds (the MicroVAX actually
seems to do floating divide faster than the 780).

Compatibility:  I copied emacs and scribe off one of our 4.2 780's and ran
them without even recompiling them.  There are a few problems though (like
with CLU), so it is recommended that you recompile on the MicroVAX, but it
is generally unnecessary.

Bitmap display:  I've only had a day to work with it, and I have only
software than I have written.  The display isn't overly impressive, but it
does work.  The main problem is that the display hangs off the Q-Bus... this
seems to slow it down.  For example, I can do graphics on the II about twice
as fast as on a MicroVAX I, but for most other types of computation, the II
is 3 to 5 time the speed of the I.

Summary:  The MicroVAX II is (in MY OPINION) clearly superior to other
workstations I seen (SUN is going to take a BIG hit).  Of course, I still
haven't seen any 68020 based machines yet, but I doubt they will perform as
well as the MicroVAX II.  With the MicroVAX II you get everything on two
chips (memory management is on the processor chip also), you get
785/780/750... compatibility (how many people couldn't run GNU emacs on the
68K box when it was first released, and how often to you run into problems
with byte order because you want your program to run on both VAX and 68K?).

Again, these are MY OPINIONS after using the machine for 2 1/2 months.  I'm
sure some people will agree, others will not.

			HIGHLY RECOMMENDED!

	Avie Tevanian
	Carnegie-Mellon University
	Department of Computer Science

PS:  The only gripes I have with DEC about the MicroVAX pertain to the
announcement film:  my advisor is "Rick" on Rich (pronounced incorrectly on
film), and they cut all the film they took of me.