Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: notesfiles
Path: utzoo!watmath!clyde!burl!ulysses!bellcore!decvax!decwrl!pyramid!
hplabs!hp-pcd!orstcs!nathan
From: nat...@orstcs.UUCP (nathan)
Newsgroups: net.micro
Subject: IBM RT: first impressions
Message-ID: <1400012@orstcs.UUCP>
Date: Sat, 1-Feb-86 18:15:00 EST
Article-I.D.: orstcs.1400012
Posted: Sat Feb 1 18:15:00 1986
Date-Received: Fri, 7-Feb-86 08:36:51 EST
Organization: Oregon State University - Corvallis, OR
Lines: 89
Nf-ID: #N:orstcs:1400012:000:4909
Nf-From: orstcs!nathan Feb 1 15:15:00 1986
IBM's RT-PC: first impressions
I went to a demo of IBM's new supermicro, the RT (RISC Technology,
i guess). They seem to be trying to change the name from PC-RT to
RT-PC for typically mealy-mouthed reasons.
They have two operating systems for it: one System V and the other
BSD4.2. The V port was done by ISC, who did PC-IX and the mainframe
ports: it's described as V.1 with Release 2 and 4.2 enhancements
(whatever that means). It does seem to have paged virtual memory.
The BSD port was apparently done by Berkeley, MIT, and/or Carnegie-Mellon.
It seems fairly vanilla. One or both OS's has loadable device drivers.
There is no OS window support in evidence, and the system V port doesn't
seem to have jobs control. No Version-8'isms are in evidence. Only
one compiler (C) comes standard; they sell an F77 for ~$1000, and Pascal
and BASIC (guaranteed identical to PC versions!). There's a sleazeball
command called "dos" that makes the shell look like PCDOS (del for rm,
dir for ls, etc.) They also have a "menu shell" in the works, which
should be amusing enough to try before discarding. They promised to
publish protocols for writing one's own installable device drivers:
expect to see somebody's window shell on it shortly.
Hardware: 32-bit semi-RISC pipelined/cached processor (50K devices on-chip)
with a separate paged-MMU (60K devices) providing 40-bit virtual
address space. No indication of how the extra 8 bits of address are
handled. Runs at about 68020 speed. Memory: minimum, 1 Meg; max, 4 Meg
but you have to figure out what to do with the 1 Meg card it comes with, as
memory additions come in 2Meg chunks and there's only room for two
memory cards. It uses a PC-AT peripheral bus (it's not clear whether
memory rides there too) where disk controllers, graphics boards,
and network interfaces go.
Standard display is a monochrome 512x768 pixel (they call them PELs!?);
optional is a color board, same resolution but displays 16 of 64 colors.
They have a bunch of other options, the higher-end ones actually
attached graphics processors. The high-end CAD stuff requires a
separate workstation processor to do 3D rotates, etc.
You can also use a PC-AT EGA, which is necessary to run graphic programs
on the...
Optional i286 coprocessor. Compatible enough with the AT to run MS Flight
Simulator, according to an IBM honcho. It has a socket for '287, and
up to 1 Meg of RAM (apparently independent of RT memory). A DOS/program
runs on the 286 as a subtask under Unix on the main processor. No windows,
but you can flip back and forth from DOS screen to Unix (er, AIX) screen.
Floating point performance with no FP accelerator is about as fast as
an AT with a '287. With the FP board ($850), they say it's faster...
I suppose I can believe that without proof.
Disks offered are 4OM and 70M. The cheap box only holds one drive, the
spendy one can have three for up to 210M of disk. If you add a
serial expander you can connect *up to* 7 other users. (Apparently
a Bell licensing restriction). There are some built-in serial ports
for such as the mouse (2-button, rollerball), printers, etc. Ethernet
ports can be had, and they say they're doing MAP drivers.
Documentation: This is where the big surprise is. One can understand
IBM rewriting Bell's manuals. The surprise is that they made the manuals
readable, and understandable -- with little diagrams explaining
inodes and links, and what fsck does. A readable Unix manual is
inexplicable enough, but a readable IBM manual is shocking.
Pricing: high. It's priced to compete with the Microvax II,
i.e. $11K to $30K. It's strange how it's bundled: the OS and
mouse are optional, the 1Meg ram card isn't. If you want 2 or 4 Meg, you
have to throw away the 1M.
I saw a fairly impressive presentation-level editor on it: WYSIWYG, with
"MacDraw" and "MS Chart" (or equivalent) built in. They have the Interleaf
stuff (too? that may be what I saw...).
Judgments:
This product will definitely give DEC (and everybody else) a run for
their money. Although it's a bit pricey, DEC is too and high prices
never stopped IBM before. Apple or S. Jobs had better hurry or IBM
will sew up the academic market. It's hard to say whether support will be
up to mainframe standards or down to PC standards -- apparently, almost
none of the software was written within IBM, and most wasn't even
commissioned by them. Much as I hate to admit it, though, IBM has produced
a desirable product. I asked the rep what kind of political machinations
within the company made this possible; he seemed to acknowledge the
premise, pointing out that the group that put it together wasn't answerable
to any of the traditional business units: their charter directs them
to address the scientific/engineering market. They are being very
quiet about offering accounting software on the machine....
Nathan C. Myers
orstcs!nathan
Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!henry
From: he...@utzoo.UUCP (Henry Spencer)
Newsgroups: net.micro
Subject: Re: IBM RT: first impressions
Message-ID: <6382@utzoo.UUCP>
Date: Sat, 8-Feb-86 20:56:06 EST
Article-I.D.: utzoo.6382
Posted: Sat Feb 8 20:56:06 1986
Date-Received: Sat, 8-Feb-86 20:56:06 EST
References: <1400012@orstcs.UUCP>
Organization: U of Toronto Zoology
Lines: 25
> ...Runs at about 68020 speed...
>
> Floating point performance with no FP accelerator is about as fast as
> an AT with a '287. With the FP board ($850), they say it's faster...
>
> ...It's priced to compete with the Microvax II, i.e. $11K to $30K...
Then why bother? Buy a Sun 3 or another 68020 machine. If it isn't
distinctly superior on either price or performance (and the 68881 is a
good deal faster than AT-with-287, I believe), then why take a chance
on an oddball product that IBM may drop next year? Especially with only
a 16-bit bus for peripherals.
> This product will definitely give DEC (and everybody else) a run for
> their money. Although it's a bit pricey, DEC is too and high prices
> never stopped IBM before. Apple or S. Jobs had better hurry or IBM
> will sew up the academic market...
Don't forget Sun and Apollo, who are much more price-competitive and
function-competitive than DEC. Beating DEC prices and performance has
never been terribly hard, so comparing the RT against DEC hardware is
a joke.
--
Henry Spencer @ U of Toronto Zoology
{allegra,ihnp4,linus,decvax}!utzoo!henry