[net.works] Unix Workstation query - results.

MCB%MIT-MC@sri-unix.UUCP (09/15/84)

From:  Michael A. Bloom <MCB @ MIT-MC>


I can't resist.  You requested info on PC's running Unix, and I've just got
to throw in my two cents worth.

Last July, I asked people on works and info-unix what computers they
would recommend (or not recommend) as a personal computer.  I
indicated that I was particularly interested in machines that run
Berkeley UNIX (either 4.1, or 4.2).

The replies I received are at the end of this article. Missing are
those replies that only asked for copies of what others sent.  I
have also done some slight editing to reduce the size of this summary.
This editing consists of header removal, and removal of text extraneous
to the subject material ("Please forward your other replies").

Some commentary first:

A couple of machines that I received no mail about are the AT&T 3b2
and the DEC MicroVax I. The microvax runs one of the two versions of
Ultrix. This version is based upon 4.1BSD, unlike the Ultrix for
larger Vaxes, which is based upon 4.2. This machine is of interest to
me especially as a DEC rep at the UNIX Systems Expo (This past week
in L.A.) informed me that the machine is available under their
educational discount plan, which (if I choose the machine) will mean
a substantial discount when ordered thru the computer dept. of my
school's bookstore.  The machine is, however, reputed to be slow, and
is not a true VAX; it lacks compatibility mode.  This isn't much
of a problem, as playing zork is not one of my intended uses.  Tho it
would be nice to be able to run pdp-11 binaries under compat.

The 3b2 seems like a nice machine.  I'm waiting for some technical
data from ATT, but I'm not likely to buy it as they are unlikely to
supply Berkeley Unix in the near future.  Now if someone were to sell
a port at a reasonable price for it... One thing that bugged me about
the AT&T rep's speil is that he kept throwing words at me about how
"Berkeley enhancements" are continually being "added" to the system.
The same is true for a lot of vendors.  MANY of them have literature
that touts their UNIX xx system with "{4.1BSD,4.2BSD,Berkeley}
enhancements".  I was surprised at myself when scanning through the
literature from a major mfr. (I forget which it was at the moment - it
may have been Honeywell), while watching their windowing demo. I
caught the words "4.2 BSD" and until I discovered otherwise shortly
afterward, I was believing that this major mfr had a system that might
meet my criteria.  The wording itself was not misleading, but I
believe that it was designed so that one making a quick scan would see
something that wasnt there. Several vendor reps have tried to tell me
that they've turned their system V port into Berkeley UNIX by "adding
the Berkeley paging code". Some asked "what else do you want?". One
told me that he couldn't believe that AT&T wouldn't provide a reliable
signal mechanism. "I'm sure we have that" was his response.  Come to
think of it, that may have been the AT&T rep. 

A friend of mine who works for an oem software house suggested strongly
that I wait until the end of the year before buying.  He couldn't tell
me why (due to non-disclosure agreements), but I trust his judgement.
Unless something spectacular appears on the market before then, I'll
hold off on making a decision.

Here are the replies:


>From tektronix!teklds!upvax!bobbyd

TRCC demonstrated at SIGGRAPH, in *July, 1979* the prototype of the Perq 1A
graphics workstation.  In August of that year, I believe, they began
production.  Keep in mind that these facts I now present are in 1979-80
terms:

	o	a 16-bit, bit-sliced microcoded cpu with 170-nsec
		microcycle, with 4K writeable control store

	o	~1-1.2 MIP Pascal performance*

	o	20-bit physical addresses (1MB dynamic RAM, 680-nsec
		average cycle), 32-bit virtual addresses

	o	768 x 1024, 60Hz, non-interlaced b & w portrait
		display (100 pixels/inch)

	o	Hardware RasterOp, 32Mbits/sec transfer rate

	o	10Mbit/sec Ethernet (to be fair, it wasn't fully up
		until arount '81, when my stepdad worked there...)

	o	24MB Winchester, 1MB DSDD floppy

	o	CVSD speech/audio output hardware

*Their Pascal is the greatest programming language I have EVER encountered.
It puts today's Modula-2 to tears - modules are BEAUTIFULLY implemented!  It
has dynamic strings fully supported, and even lets you do C-like bit
twiddling with shift() and rotate() and 'and' and 'or' etc.  Graphics through
rasterop().  The entire operating system accessible through the very modules
it was built from!

"Ah, so this marvelous machine from the old days...why haven't I heard of
it", I hear you cry.

Hate to admit this, but they did some things wrong.  TRCC has GREAT
engineers, but LOUSY management...

	o	They didn't bring up Unix as their primary OS.
		Look at how big a plus that is NOWadays...

	o	Color.  They STILL don't, I believe, have a color
		machine out.

	o	Software floating point.  Um, an engineering work-
		station that can't do FAST floating point...?

	o	Software development tools.  Their Pascal is, as I
		mentioned, the GREATEST.  But they didn't bring up
		a good debugger, or any kind of SCCS (?) or real
		development tools.

	o	Standards.  They are soon to announce a Multibus
		adapter.  Though I am tremendously proud of their
		CPU (it's one HOT machine, and is a hacker's dream),
		they didn't put it on a very accessible bus...so now
		folks go out and buy a Nu Machine or a Sun and slap
		just about any peripheral or uprocessor they choose
		on it...

	o	Advertising.  Like I said, they could have had better
		management (money troubles for a while, but they're
		pulling out of their slump)

Well, if you haven't seen their new ads, here's (finally) Perq 2.  I'll note
that I saw the prototype of the gate array version of Perq 1 in ~81-82,
BEFORE uVax and uPDP-11 and Tower-1632 and MegaFrame and all the skinny
lookalikes there are now...(no gate array version - money problems).  Anyway,
Perq 2 is essentially a souped-up version of P1, but actually lower priced.

	o	2MB RAM (from 20-bit addresses?  I assume they really
		mean 1Mwords which turns out to be 2Mbytes...dunno)

	o	1280 x 1024 landscape display optional (100 pixels/in,
		60Hz non-IL)

	o	16K writable control store

	o	Carnegie-Mellon's Accent OS

	o	FORTRAN, C and CommonLisp (Poplog from ICL in Britain?)

	o	up to 288MB Winchester disk internal (faster than the
		old units @ 30msec)

Now, you asked about Unix PC's.  Well, I've finally gotten to that.  Accent.
A damn good hack, if the old CMU folks are still around...(did I mention
that Three Rivers Computer (Perq Systems) is in Pittsburgh, PA?  Oh, well,
my stepfather went to CMU, and I used to hang out there even before I knew
him, and he says that they write far better code than the Bezerkeley
types...)

OOPS!  Verbosity!  Back to Accent:

	o	The 16K WCS and the great demand for Perq microcoders
		back when they were hacking Accent as part of CMU's
		Spice Project makes me want to believe that the Accent
		kernel is in microcode.  THAT means, and if you see
		the old issue of Byte that showed the Perq 7TH (YES,
		7th, AHEAD of ALL the DEC machines benchmarked!) fastest
		at the Sieve, that Accent SCREAMS!  (Like, what if they
		wrote a Pascal->microcode compiler for that Luscious
		Pascal?)  SCREAMS, as in >5 MIPS...

	o	Well, ATOP the Accent kernel can be Qnix, which is their
		validated SysV.  Or you can have the Lisp environment,
		or the default Accent shell.  Why they never tried
		SmallTalk on the Perq mystifies me...

	o	INSIDE the Accent kernel is their completely distributed,
		virtual memory, message passing networking scheme.  Now,
		I'm no wizard, and I've never used or seen or touched 
		Accent.  But they claim:  Up to 64 windows (viewports);
		completely transparent networking; icons for window and
		PROCESS control; completely distributed processing, with
		automatic distribution of network-wide loads; priority
		scheduling with pre-emption and aging; demand paging;
		etc.

I'll just end (abruptly) this rambling with

		Perq Systems Corporation
		2600 Liberty Avenue
		P.O. Box 2600
		Pittsburgh, PA  15230

		(800)-222-4489 (outside PA)

I'd estimate ~$15K for a low-end machine.  A bit steep?  Maybe if you find
out anything from PSC you could let me know...just your impressions should
you get a chance to see the machine or some literature...

I'm through.  Flame off.  All that jazz.  I notice that my friendly Vax has
put 1:52am in my mode line...about time to hit the sack...or better yet...
time to play rogue!

Hope my flaming wasn't too much of a good thing...good luck with your survey!

World B. Free
Perq Fanatic
...!ucbvax!tektronix!teklds!upvax
----------------------------------------------------------------
From:  Wilkinson@HI-MULTICS.ARPA

I am currently trying to get detailed info on NEC's recent Unix/Graphics
machine (Advanced PC - III or APC-III).  It is 8086 based but has two;
one for memory management.  Also resolution is great; 600 X 400 on a 14"
NEC monitor (colour or mono), with NEC7220 graphics chip.  Can partition
hardisk to run Unix and MS-DOS.  This I gleaned from a news release kit
and am looking for more DETAILS from NEC.  I am VERY interested in both
subjects (Unix and Graphics) on PC's
          Richard
 BTW NEC is out in Boxboro MA 264-8000.  I am trying to get the
details from Steve Abt 264-8128.  He is blaming mail for slowness but
that is over a month.  I think they are still putting documentation
together, maybe even the system - it was supposed to start shipping this
month.
----------------------------------------------------------------
From: nbires!mccallum@Berkeley (Doug McCallum)

I work for NBI and we make a 68010 based workstation which runs
4.2BSD UNIX.  The UNIX workstation was just announced this week
at NCC.

It is a full port of 4.2 with virtual memory, job control, etc.  We have
our own multiwindow system for the bitmapped display which makes job
control less necessary, although it is still there if you want it.  The 
windows are a similar to SUN's in that a process thinks of it as a terminal.

I don't remember all the details of the configuration and pricing, but
I think it is something like 1 Mbyte memory, 24 Mbyte disk, 4.2BSD, bitmapped
display, keyboard and mouse for slightly under $16,000.

If you need more information I can probably get it.

			Doug McCallum
			NBI, Inc.
----------------------------------------------------------------
From:     Rusty (Where's the fish?)Haddock <haddock%ti-csl.csnet@csnet-relay.arpa>
Michael,
	I'd be very interested in receiving anything you get
with regards to UNIX machines as I plan to buy one in about the same
timeframe.   I'd like to get something based on the National Semi's
Series 32000 (alias NS 16000) uP.  Unfortunately there remains several
small bugs in this processor family; the main one being that a handful
of instructions will not restart after a page fault.   I've been told
that this should problem should be fixed sometime this fall.

	May I suggest that you stay away from the Altos machines
at least the ones based on the 8086 (e.g. the 8600 [discontinued]
and the 586).  My experience with Microsoft's XENIX for this machine
is not very good and Altos/Microsoft do not deliver everything listed
in the UNIX documentation.   Their service could be called friendly
but they don't have the manpower to fix bugs in their software.
I don't know about the 68000 and 568 machines which are based on
Motorola's 68000 uP.  It may be possible that the software on this
machine is more reliable as it is System III and Microsoft is not
involved. (Sorry for the harsh words toward Microsoft but, from my
experiences with them and their software, I think they deserve it.)

	Another good possibility for a UNIX machine may be the
LMC MegaMicro which is advertised in BYTE.  I spoke with those folks
last year and was impressed with, what sounds like, an exceptional
piece of hardware.  Unfortunately they boosted their minimum system
configuration from $15K to $20K by supplying a 30-Meg disk instead of
the 10-Meg plus some other.   They are running HCR's version of
Berserkley 4.1 (could be 4.2 by now) and their disks are the
30-millisec average access time which ain't too bad.
----------------------------------------------------------------
From: masscomp!tjt at mit-vax
I work for Masscomp.  The MC-500 is a 68K based machine.  Our current
software is a mixture of System III and 4.1BSD, including virtual
memory and job control, although the terminal driver is basically
System III and multiplexed files are not supported.  We also have
4.1aBSD style Ethernet support, including ARP support plus support for
FTP and TELNET.

Soon to appear (i.e. November release) will be a 4BSD compatible
terminal driver, System V source code compatibility (shared regions,
messages and semaphores), remote file access (not just remote disk
access), and symbolic links.  By next January, we will support 4.2BSD
source code compatibility.

The machine is fast: at usenix benchmarks the machine has consistently
outperformed 750's and most other 68K based machines, including those
running at 12.5MHz (our processor runs at 10MHz).  Our performance
degraded somewhat with our virtual memory release as the system had not
been tuned.  Our November release will recover that performance.

	Tom Teixeira,  Massachusetts Computer Corporation.  Westford MA
	...!{ihnp4,harpo,decvax}!masscomp!tjt   (617) 692-6200 x275
----------------------------------------------------------------
From: Rich Zellich <ZELLICH@SRI-NIC.ARPA>

At the NCC last week, NBI was showing a nice system for $15,475, which
was a 68010 running an apparently-total 4.2bsd port with a bit-mapped screen,
1 meg memory (2 meg optional), 22 (or 24?) meg Winchester, 640K floppy, and
custom raster-ops processor.  The bit-mapped screen was used for Star/Lisa/
MacIntosh style icons and overlapping windows; a separate process could be
run in each window, and the system could also emulate a VT100 (and other
terminals?) in a window connected to a mainframe while running local programs
in other windows.  Besides UUCP and cu, TCP/IP are also supported (as said
above, apparently \all/ of 4.2 was ported).  vi, ed, and ex are of course
available, and I think emacs is, also.  Oh yeah, the "U!" system also has
a 3-button mouse.

The literature, as I remember it (I haven't unpacked all the NCC stuff yet)
compares the U! machine as slightly slower than a single-user VAX 11/750 and,
in some cases, slightly faster than a 4-user 11/750.

-Rich
----------------------------------------------------------------
From: Mats Wichmann <mats@Berkeley>

How long a line would you like? We build 68k machines, and have been
delivering for a long time (since January, 1982 - introduced at Comdex
November 1981). I don't want to give you the whole sales pitch, but our
machines perform very nicely - have benchmarked very well against
people like Plexus (blow them away in disk-intensive stuff, in fact),
and far superior to things of the class of Callan, Heurikon, Pixel, etc.
However, we ain't real cheap, and we run Sys V, with no job control. 
(Deal is: $21k gets you .5MB memory, 80MB disk, 68000 @ 10 MHz, 4 serial 
ports). For further information, please call our sales people - phone is
415-549-3854. 

The following is my personal opinion, having nothing to do with Dual:
If you must have job control and 4.2BSD, look at Integrated Solutions.
They really have their act together.

	    Mats Wichmann
	    Dual Systems Corp.
	    ...{ucbvax,amd,ihnp4,cbosgd,decwrl,fortune}!dual!mats