Please consider a donation to the Higher Intellect project. See https://preterhuman.net/donate.php or the Donate to Higher Intellect page for more info.

Silicon Graphics IRIS FAQ

From Higher Intellect Vintage Wiki
Revision as of 05:38, 3 September 2020 by Netfreak (talk | contribs)

https://cdn.preterhuman.net/texts/computing/sgi/IRIS_FAQ.txt

                Silicon Graphics IRIS 2000/3000 FAQ
                -----------------------------------

0                       TABLE OF CONTENTS
                        -----------------

0       Table of Contents

1       Introduction
1.1             Revision History

2       The Machines

2.1             IRIS 1000 Series
2.2             IRIS 2000 Series
2.3             IRIS 3000 Series

3       Documentation

3.1             System and Software
3.2             Hardware

4       Hardware

4.1             Getting All the Wires and Switches Right
4.1.1                   External Cabling:  Ports, Keyboards, and Mice
4.1.2                   Starting It Up
4.1.3                   Booting UNIX
4.1.4                   The Other Back Panel Stuff
4.1.4.1                         The LED Display
4.1.4.2                         Configuration Switches
4.1.5                   Internal Cabling and Board Settings
4.1.5.1                         CPU Side Board Locations
4.1.5.2                         IM1 RAM Switch Settings
4.1.5.3                         Graphics Side Board Locations and Cabling
4.1.5.4                         BP3 Bitplane Switch Settings

4.2             Disk Stuff
4.2.1                   Controllers
4.2.1.1                         Qualogy ST-506
4.2.1.2                         Interphase ESDI
4.2.1.3                         Interphase SMD
4.2.2                   Drives
4.2.3                   Maintenance

4.3             Video Stuff
4.3.1                   Video Modes
4.3.2                   Monitors
4.3.2.1                         Models
4.3.2.2                         Docs, Parts, and Repairs

4.4             Hardware Miscellany
4.4.1                   Third-Party Multibus Boards
4.4.1.1                         Central Data Serial I/O
4.4.1.2                         Ciprico 9-Track Controller
4.4.1.3                         Excelan Ethernet
4.4.1.4                         Hyperchannel
4.4.1.5                         Ikon DR11-W
4.4.1.6                         National Instruments IEEE-488
4.4.1.7                         Sky FPU
4.4.1.8                         Xylogics Disk Controllers
4.4.2                   Power Supplies
4.4.3                   Batteries
4.4.4                   Extender Boards
4.4.5                   Loose Ends

5       Software

5.1             OS Versions
5.2             X
5.3             GNU
5.4             Applications
5.5             Software Miscellany

1                       INTRODUCTION
                        ------------

The Silicon Graphics IRIS 1000, 2000, and 3000 series workstations were 
the "first generation" 680x0/Multibus-based machines built and sold by 
Silicon Graphics Inc. of Mountain View, California.  The systems were 
obsoleted with the advent of the company's "4D" MIPS-based machines.

SGI formally announced end-of-production on the 3000s in November 1989.
Although at that time they committed to continued support through
November 1994 (subject to parts availability), no further software
was to be issued, and it is unlikely that anyone would avail themselves
of SGI's service.  I doubt that you could get a board swap out of SGI
for less than $1000, and the (informal) surplus market rarely sees a
machine sell for more than a few hundred dollars.

Silicon Graphics is a company apparently full of friendly and helpful
folks, but rapid advances in the workstation market have meant that 
inquiries about these old systems are generally met with blank stares.  
Hell, questions on early 4D machines are sometimes met with blank stares!

SGI claims to have sold more than 3500 of the 2000 and 3000 series
systems (no numbers are known on the 1000s), and as these systems have
trickled into the hand-me-down market, users have sought basic information
on bringing (and keeping) up the machines.

Hence this FAQ.  I hope it arouses enough interest to lead to the
formation of an informal, mailing-list-based users' group dedicated to
"obsolete" Silicon Graphics systems.

This is Rev 1.0, written by a hardware engineer with no prior UNIX
experience, working on the unconfirmed assumption that nobody else has
already done it.  It's also my first FAQ.  When I asked these questions, 
no one directed me to one, therefore...

Although I'm learning as much about UNIX and system administration as 
I can, I've learned about these machines from the "bottom-up".  As a
result this Rev contains, almost exclusively, hardware information.
Experienced users, more software-literate than I, are enthusiastically 
invited to make software/system/os submissions they feel are relevant.

If you wish to make a submission, please send only the diffs/deltas,
rather than adding to the existing body.

Oh, yeah... In case you were losing sleep over this, 
IRIS stands for "Integrated Raster Imaging System".

Please forward questions/answers/suggestions/errors/omissions to:

Jonathan Levine     [email protected]

Thanks, and I hope this helps!
May 24th, 1993

1.1     REVISION HISTORY
        ----------------

May 24th, 1993  First release
May 31st        More 3000 rear panel switches & mouse pinning
                (suggested by William Bardwell)
June 2nd        Completed IP2 S2 description
June 4th        Corrected Status LED information
June 7th        Added second Ethernet board addressing information
                (thanks to Randy Schrickel)
June 8th        Completed video mode section
June 15th       Added IEEE-488 controller information
                (thanks to Tom Haberlandt)
Aug 21st        Updated schematic availability
                Tweaked "Loose Ends"
                (A big thank you to Al Kossow)
Oct 1st         My mailer finally configured!
                (much thanks to Jeff Becker (ex-SGI))
Nov 29th        Added section numbering and EXOS prom cross-reference
Dec 8th         Added configuration switch mapping
Jan 5th, 1994   Added internal cabling and board settings
Feb 27th        Added "Power Supplies" section
Apr 7th         Added step-by-step fex instructions
Jun 6th         Added Video Mode CPU jumpers information (such as it is)

2                       THE MACHINES
                        ------------

2.1     IRIS 1000 SERIES
        ----------------

This information is provided in the interest of historical completeness,
as the remainder of this document will deal only with the 2000 and 3000
Series machines.

The 1000 and 1200 were 8MHz 68000s with 3/4Meg of RAM and no disk.  They
were built for use only as terminals.

CPU:            PM1 (variant of Stanford UNiversity "SUN" board) 
RAM:            Micro Memory Inc. Multibus
Ethernet:       Excelan EXOS/101
Graphics:       GF1 Frame Buffer
                UC3 Update Controller
                DC3 Display Controller
                BP2 Bitplane

The 1400 and 1500 were 10MHz 68010s with 1.5Meg of RAM.  The former had
a 72Meg ST-506 disk, the latter a 474Meg SMD.  My guess would be that
they used the same CPU (PM2) and RAM (PM2M1) as the non-Turbo 2000s.

The standard monitor shipped was 30Hz interlaced.

If anyone out there is still running one of these or just has more 
information they feel should be included, please let me know.

2.2     IRIS 2000 SERIES
        ----------------

The 2000 series consists of two groups of machines:  Turbo and Non-Turbo.
It is important to note that, owing to the different CPUs, these two
groups are NOT object-code compatible with each other.

                Common to ALL 2000s
                -------------------
Ethernet:       Excelan EXOS/201
Graphics:       GF2 Frame Buffer
                UC4 Update Controller
                DC4 Display Controller
                BP3 Bitplane
                        |
                ---------------------------------------------
                Non-Turbo                               Turbo 
                ---------                               -----
CPU:            PM2 (68010)                             IP2 (68020)
RAM:            PM2M1                                   IM1
FPU:            SKYFFP-M-03                             FP1 (optional)
                  |                                       |
        ------------------------------------    ---------------------
        2000    2200    2300    2400    2500    2300T   2400T   2500T
        ------------    ------------    ----    -------------   -----
Disk*:      none           ST-506       SMD         ST-506       SMD

*Refer to the section on disks and controllers (below) for further 
 explanation.

As you can see from the above table, the only hardware difference
between a non-Turbo system and its Turbo counterpart is the CPU and RAM.
For this reason, Turbo was also offered as an upgrade path from the
original 2400 (probably 2300 & 2500 too, but I don't have confirmation).
SGI simply swapped the CPU and RAM boards, reformatted and partitioned
the disk(s), and installed the new os.  NOTE:  Non-Turbo partitioning is
unusual!  Again, see the "Disks and Controllers" section.

The 2500 and 2500T are packaged differently as well.  They live in a
6' tall 19" EIA rack.  There is room for two SMD drives in the bottom 
of the rack (after all, they weigh about 150lb. each), and the electronics 
chassis resides in the top half.

I should note here that Turbo and non-Turbo RAM cannot be mixed.  
The IP2 introduced a new (ribbon-cable) local bus through which it
communicated with the IM1(s) and FP1.  The PM2 simply used the 
Multibus.

In short, the Turbos have much more in common with the 3000s than they
do with the non-Turbo 2000s.

The standard monitor shipped was 60Hz non-interlaced.

2.3     IRIS 3000 SERIES
        ----------------

CPU:            IP2 (68020)
RAM:            IM1
FPU:            FP1 (optional)
Ethernet:       Excelan EXOS/201
Graphics:       GF2 Frame Buffer
                UC4 Update Controller
                DC4 Display Controller
                BP3 Bitplane

Geometry Engines:      10x8MHz                    12x10MHz
                ---------------------   ------------------------------
                3010    3020    3030*   3110    3115    3120    3130**
                ------------    -----   ------------    --------------
Disk:              ST-506        SMD       ST-506           ESDI 

* As with the 2000s with SMDs, this is a rackmount system.

**The 3130 was also private-labelled by CDC as the Cyber 910.

In addition to the differences described above, there are also variations 
from model to model in the amount of RAM and number of bitplanes, but
since those are just optional upgrades anyway, I don't think it's worth
documenting them in detail.  You're better off opening the box and counting
up the boards.

All things considered, I don't view the differences between the models
in the 3000 series as very important.  After all, there's no reason why one
can't replace the ESDI drives in a 3130 with SMDs, resulting in a system
to which no equivalent factory model exists.

The standard monitor shipped was the 19" 60Hz non-interlaced, tilt/swivel.
Optional monitors were the 19" 30Hz interlaced and 15" 60Hz non-interlaced
tilt/swivel.

3                       DOCUMENTATION
                        -------------

3.1     SYSTEM AND SOFTWARE
        -------------------

If you got the manual set with the machine, good.  If you didn't, well,
not so good.  You'll have to do some serious schmoozing to get someone
to photocopy it, as it fills, more or less, ten (10) 1.5" binders.
I am including this roadmap to the docs so that you can quickly determine
your requirements.  Once you know what you need, it should be easier to 
arrange to have the specific portions copied rather than entire manuals.

PLEASE NOTE:  Although the following list was generated from a set of
2000-series manuals, I use these docs as reference for my 3000s.  I 
have yet to find any significant differences.  Your online man pages
should be regarded as the definitive reference for your release.

Within each binder I have listed each section delimited by a divider.
Here's the breakdown:

Vol.1   IRIS Series 2000

(This is the general reference and some assorted stuff.
The "IRIS Workstation Guide" contains information on optional video
hardware, making and using backups, some basic disk manipulation ie
fiddling with swap partition size and configuration of additional
disks, and beginner sysadmin things.  It's rather useful, and if
feedback indicates that a significant number of users are without
docs, selected portions may be included in later revs of this FAQ)

                Getting Started
                IRIS Workstation Guide
                GL2-W2.4 Release Notes
                GL2-W2.5 Release Notes
                Pipeline
                C dbx

Vol.2   IRIS Programmer's Manual Unix IA Commands A-L

(This binder and the next contain section 1 of the standard Unix documentation
set.  It's pretty much the same stuff as contained in the online man pages)
        
                Permuted Index
                A-C
                D-F
                G-L

Vol.3   IRIS Programmer's Manual Unix IA Commands M-Z

                M-P
                Q-S
                T-U
                V-Z

Vol.4   IRIS Programmer's Manual Unix IB System Calls & Subroutines

(This is the second volume in the standard Unix set, Sections 2 through 8)

                Permuted Index
                2. System Calls
                3. Subroutines
                4. File Formats
                5. Misc.
                6. Games
                7. Special Files
                8. Administration

Vol.5   IRIS Unix IIA Languages and Tools

(Environment and tools.  This binder written by the gods themselves)

                The Unix Timesharing System
                Unix for Beginners
                (Bourne) Shell
                Ed
                Ex
                vi
                Troff Tutorial
                MM
                Nroff/Troff Reference
                Eqn
                Tbl
                Refer
                Style & Diction
                Unix Programming

Vol.6   IRIS Unix IIB Languages and Tools

(Some of the Fortran, Assembler, and Floating Point stuff may be out
of place here, but that's where it wound up in the set I copied from.  
More key environment and tools stuff, too)

                Fortran Reference
                Fortran Debugger
                AS20 Assembler
                Floating Point
                C/Fortran Interface
                UUCP
                FSCK
                Lint
                Make
                Yacc
                Lex
                ADB
                M4
                SED
                AWK

Vol.7   IRIS Communications

                Iris Communications Guide
                Appendix A - Options
                Appendix B - Bibliography
                Appendix C - NFS Reference
                Appendix D - Mail Systems
                Appendix E - Sendmail Router
                Appendix F - Sendmail Guide
                Appendix G - Domain Convention
                TCP/IP User's Guide

Vol.8   IRIS Fortran

(I don't think we have a copy of this fortran; the manual probably came 
with my first system, the fortran tapes for which had been overwritten)

                Iris Fortran
                Fortran dbx

Vol.9   IRIS User's Guide Volume I Graphics Programming

(The part you've been waiting for)

                Graphics Programming
                Window Manager
                Sample Code
                Glossary
                Index

Vol.10  IRIS User's Guide Volume II Graphics Reference

(This binder contains the routine-by-routine GL reference)

                Graphics Reference
                Appendix A - Type Definitions
                Appendix B - Geometry Engine Computations
                Appendix C - Transformation Matrices
                Appendix D - Feedback Parser
                Appendix E - MEX Programs
                Appendix F - Tutorial man pages
                Appendix G - Fast Immediate Mode & User-defined Display Lists
                Appendix H - Image Libraries

3.2     HARDWARE
        --------

There's not much to say about hardware docs other than "Lie, Cheat, Steal".
I can confirm the existence of a "3000-Series Workstation Maintenance Study 
Guide", which is a training manual for field-service personnel.  I have also 
seen these schematics: IP2 (and block diagram), IM1, BP3, UC4.  If you know 
of (or have) any other drawings, let's talk.  I'm especially interested in
seeing the FP1, GF2, and DC4 sheets.

4                       HARDWARE
                        --------

4.1     GETTING ALL THE WIRES AND SWITCHES RIGHT
        ----------------------------------------

4.1.1   External Cabling (Or: Of Ports and Keyboards; Of Mice and Men)
        --------------------------------------------------------------

The rear panel has four (4) DB-25S connectors.  The first (top) is labelled
"Control Panel / Port 1".  This is the console connector, for use with the
IRIS keyboard.  The remaining ports, labelled "Port 2", "Port 3", and 
"Port 4", correspond to ttyd1, ttyd2, and ttyd3 respectively.  Personally,
I would have labelled them Port 0 - Port 3, but you know how it is.

There are two styles of keyboard, differing in their connectors:

Older 2000 keyboards have a 5-pin 180-degree DIN connector which plugs into 
the keyboard junction box.  The box may be free-standing or integrated into
the monitor case.  The box is in turn cabled to the "Control Panel / Port 1"
connector on the rear panel with a male-to-female (DB-25P to DB-25S) cable,
pinned 1-1,2-2,...,25-25.  The junction box also has a DB-9S connector
into which is plugged the mouse.

The 3000s (and maybe some Turbos) use the newer style keyboard.  It has a
longer cable with a DB-25P connector that plugs directly into the console
port.  The mouse plugs into the DB9-S on either side of the keyboard.

Mouse Pinning
-------------

If you hate your mouse, don't worry, it's an easy hack. The average
bus mouse is readily adapted for use with the IRIS.  The mouse is 
supplied with a regulated +5V from the keyboard or junction box, and 
outputs a quadrature X pair and a quadrature Y pair:

        DB9-P           function
        -----           --------
          1             +5V
          2             X (left/right) Quad out
          3             X (left/right) Quad out
          4             Y (up/down)    Quad out
          5             Y (up/down)    Quad out
          6             Left   button (active low)
          7             Middle button (active low)
          8             Right  button (active low)
          9             Ground

-Sorry, but I never bothered to write down which signal of each quadrature
 pair leads in which direction.

-I just took my meter to the optical mouse, and it appears that each button
 is a 1K pullup to +5 and a switch closure to ground.

I shouldn't have to explain monitor cabling beyond men-tioning that
all video signals should be terminated at the monitor.
 
4.1.2   Starting It Up
        --------------

On power-up, the CPU first attempts to establish communications with the
keyboard.  If all goes well, the LEDs above the keyboard's numeric row
will sequence left to right, the keyboard will beep, and the monitor prompt
will appear on the video monitor.

If the CPU fails to secure keyboard I/O, it will assume that either the
keyboard is defective, or that the system is a server (ie no graphics).
This having been determined, it outputs the monitor prompt to the "Port 2"
connector, to which I suggest you keep a terminal connected at all times.

One cause of this failure is the loss of power to the keyboard.  The
keyboard LEDs are not a valid indication, as they are powered from the
+12V supply, and the keyboard also requires -12V to output RS-232 levels.

If you suspect a problem here, look for the +12V on pins 9,21,22 and
-12V on pin 18 of the "Port 1 / Control Panel" DB-25 connector (pins
1,4,5,6,7 are ground).  If a supply is missing, remove the cover over the
power supply compartment, and check the two fuses along the top edge of
the I/O panel circuit board.  Visual examination is NOT adequate with
these fuses.  I had one in which the fuse link was clearly intact, but
was intermittent down in the epoxy base.  Believe me, that one was fun.

4.1.3   Booting UNIX
        ------------

If everything is ok to this point, you'll have the IRIS PROM Monitor
prompt.  It'll have read the switches and checked to see how much memory
is installed.

The monitor is your friend.  It has its own help, so I shan't go into
much detail here, except to touch upon booting the os.

When, at the monitor prompt, you type b, the monitor loads and
runs the file "defaultboot" from the default boot device.  This will
most likely be one of the following disks:  md0 (ST-506), si0 (ESDI),
or ip0 (SMD).  If all is in order, defaultboot will load and run, and
UNIX will start up with its device autoconfiguration.  Note:  When 
powering on the system, wait 30 seconds or so for the disk to spin up
before attempting to boot.  If the disk isn't up to speed when you try
to boot it, you'll get the failure described below.  No damage will result;
just wait a bit and try again.

If things aren't going allright, though, you will probably get a message
such as "md.0:defaultboot No such file or directory".  This likely means
someone wiped it off, and you're going to have to go in search of a mkboot
tape (see below).  In some cases, the boot routine will report finding
defaultboot, load it, and then bomb out on execution.  This may be indicative
of a bad copy on disk, and will also send you out for a tape.

The monitor can be of some diagnostic help here.  It has an "ls" that
functions just as UNIX's.  Use it to look at the disk, and confirm that
the usual root directory stuff is there ie bin, etc, tmp, lib, and usr.
You should see an entry for defaultboot right along with them.  Incidentally,
the monitor ls will examine subdirectories, too.  Keep in mind that you
can only check the contents of the root partition, as no other filesystems,
such as usr, are "mounted" until UNIX boots.

[how about something on serial/server operation here?]

4.1.4   The Other Back Panel Stuff
        --------------------------

4.1.4.1 Status Display
        -------------- 

I know of no key explaining the single hex digit display.
This, however, is what it does (at least, on a running 3130):

        State                      Status LED
        -----                      ----------
Idle at monitor prompt                 5
Booting or running os           Alternates 2 & 3
Shutting down (reboot)            F -> 6 -> 5

4.1.4.2 Configuration Switches
        ----------------------

Not so for the rear panel DIP switch.  It has different meanings, 
depending on whether you have a 2000 or a 3000 series system:

*Asterisks indicate factory defaults

Non-Turbo 2000 Switches:
------------------------

1 2                     Serial line speed
- -                     -----------------
C C                     300 baud
O C                     1200 baud
O O*                    9600 baud

    3                   Verbose 
    -                   ----------
    C                   No status reports in power-up testing
    O*                  Status reports in power-up testing

      4                 Autoboot
      -                 --------
      C                 Autoboot using specified environment 
      O*                Manual (PROM monitor) boot

      4                 Monitor display type
      -                 --------------------
      C                 Primary video driver
      O                 Secondary video driver

        5 6 7 8         Boot environment
        -------         ----------------
        C C C C         Floppy disk
        C C C O*        2400 ST-506 hard disk
        C O C C         Network
        O C C C         PROM monitor (Note effect on Switch 4 function)
        O O C C         Serial line
        O C C O         2500 SMD hard disk

                9       Reserved
                -       --------
                C*      Leave closed

No, the two sets of entries for Switch 4 isn't a mistake. 
To quote the Series 2000 User's Guide:

        The meaning of Switch 4 depends on the settings of
        Switches 5 through 8, which determine the boot
        environment.

        If the boot environment is set to the PROM Monitor,
        Switch 4 is interpreted to select a monitor display
        type.  Setting Switch 4 to CLOSED causes the workstation
        to initialize the primary video driver.  Setting
        Switch 4 to OPEN causes the workstation to initialize
        the secondary video driver.

3000 & Turbo 2000 Switches:
---------------------------

1 2 3 4         ($)     Boot environment
-------                 ----------------
C C C C*         0      Hard disk               \
O C C C          1      Cartridge tape          |          3000
C O C C          2      Floppy disk             |- Source: User's
O O C C          3      Network (XNS)           |          Manual
O C O C          5      PROM                    /

C C O C          4      Not used                \
C O O C          6      Diagnostic PROM board   |
O O O C**        7      Not used                |
C C C O          8      Not used                |
O C C O          9      ip - interphase (SMD)   |          3000
C O C O          A      st - storager 2 tape    |- Source: Maintenance
O O C O          B      sf - storager 2 floppy  |          Manual
C C O O          C      sd - storager 2 disk    |
O C O O          D      mt - DSD tape           |
C O O O          E      mf - DSD floppy         |
O O O O          F      md - DSD disk           /

        5               Autoboot
        -               --------
        C*              PROM monitor
        O               Automatic

          6             Quiet mode
          -             ----------
          C*            Display system information
          O             Don't display system information

            7           Monitor select
            -           --------------
            C*          Display on primary monitor
            O           Display on secondary monitor

              8 9       Reserved
              ---       --------
              C C*      Leave closed

** According to William Bardwell ([email protected]),
   this is actually a TCP/IP boot setting.

When the processor resets and enters the monitor, it reports the settings
of these DIP switches.  This is how the numbers displayed map to the
actual switches (1=OPEN, 0=CLOSED):

"Configuration Switch: 0xUBLB"  (As reported in hex by the monitor) 

                Upper Byte                    Lower Byte
bit position: 7 6 5 4 3 2 1 0               7 6 5 4 3 2 1 0
              ---------------               --------------- 
      IP2-S2: 1 2 3 4 5 6 7 8       IP2-S1: 1 2 3 4 5 6 7 8
                                Rear Panel: 8 7 6 5 4 3 2 1

What this means is that IP2-S1 and the Rear Panel switches are parallelled, 
so a CLOSED switch will override an OPEN one.  I strongly recommend ensuring 
that IP2-S1 is set to ALL OPEN so that the Rear Panel switches can be used.

4.1.5   Internal Cabling and Board Settings
        -----------------------------------

The IRIS Multibus backplane is divided into two sections:  The CPU
side (slots 1-9) and the Graphics side (slots 10-20).  This is not
arbitrary.  The P2 traces on the back of the motherboard are cut between
the P2 connectors 9 and 10, isolating the CPU half from the Graphics half.

4.1.5.1 CPU Side Board Locations:
        -------------------------

Brackets () indicate optional boards.

Without Floating Point:
-----------------------

  1       2       3      4        5       6       7       8       9
 ---     ---     ----   ----     ----    ----    ----    ----    ---
 IM1     IP2     ENET   DISK    (IEEE)                          (CG1)
(IM1)    IM1     IP2    ENET     DISK   (IEEE)                  (CG1)
(IM1)   (IM1)    IM1    IP2      ENET    DISK   (IEEE)          (CG1)
(IM1)   (IM1)   (IM1)   IM1      IP2     ENET    DISK   (IEEE)  (CG1)

With Floating Point:
--------------------

  1       2       3      4        5       6       7       8       9
 ---     ---     ----   ----     ----    ----    ----    ----    ---
 IM1     FP1     IP2    ENET     DISK   (IEEE)                  (CG1)
(IM1)    IM1     FP1    IP2      ENET    DISK   (IEEE)          (CG1)
(IM1)   (IM1)    IM1    FP1      IP2     ENET    DISK   (IEEE)  (CG1)
(IM1)   (IM1)   (IM1)   IM1      FP1     IP2     ENET    DISK   (CG1)

4.1.5.2 IM1 RAM Switch Settings:
        ------------------------

        Board:  1   2   3   4 
                -   -   -   - 
        SW1:    C   C   C   C 
        SW2:    C   C   C   C
        SW3:    C   C   O   O
        SW4:    C   O   C   O


4.1.5.3 Graphics Side Board Locations and Cabling:
        ------------------------------------------

Diagrams would make this much easier, but I think in this case bad
ASCII art may be worse than none at all.  Here's how the right side
of the card cage is cabled:

Board:   BP3     BP3     BP3    DC4     UC4     GF2      IP2
Slot:   10,11   12,13   14-17   18      19      20      varies
        -----   -----   -----   ---     ---     ---     ------

         no                     J7------J3
        J1/J3    J3-------------J6
        cables   J1-------------J5              J3--------J5
                         J3-----J4      
                         J1-----J1      J1------J1

Each BP3 board contains four (4) bitplanes, and each of those bitplanes
is 1024x1024x1.  The minimum number of BP3 boards needed to run graphics 
is two (2), that is, eight (8) bitplanes, or 8 bits/pixel.  These two boards 
are installed in slots 16 and 17, and bitplanes are added from right to left.  

As the above chart shows, the cables from the DC4's J1 and J4 daisy chain 
across the four (4) rightmost BP3s' J1 and J3 connectors, respectively.
The DC4's J5 and J6 daisy chain across J1 and J3 on the BP3s in slots 
12 and 13, and the BP3s installed in slots 10 and 11 have no cables 
connected to them.

4.1.5.4 BP3 Bitplane Switch Settings:
        -----------------------------

        Slot:   10  11  12  13  14  15  16  17
                --  --  --  --  --  --  --  --
        SW1:    O   O   O   O   O   O   O   O   (UC4)
        SW2:    O   O   O   O   C   C   C   C   (CD)
        SW3:    O   O   C   C   O   O   C   C   (BD1)
        SW4:    O   C   O   C   O   C   O   C   (BD0)


4.2     DISK STUFF
        ----------

The Multibus IRI used three drive interface types:
ST-506 (also referred to as MFM), ESDI, and SMD.  SGI never shipped
any SCSIs in this era, though I have little doubt that there's someone 
out there who's hacked one.  If you're him/her, please let us know.  
SCSIs sure are getting big and cheap these days.

4.2.1   Controllers
        -----------

4.2.1.1 Qualogy DSD-5217 ST-506/QIC-02  
        ------------------------------

"DSD" stands for Data Systems Design, which apparently was Qualogy's 
old name.  This board is sometimes referred to as the "Midas".  In the
ST-506 systems, it is both the disk controller and cartridge tape
controller.  It also has a connector for a 360K floppy, but I don't know 
whether anyone's ever used it in an IRIS.

(/usr/include/multibus/dsdreg.h)

Qualogy
2241 Lundy Avenue
San Jose, CA 95131-9884
voice: 408-434-5200

4.2.1.2 Interphase 3030 Storager ESDI/ST-506/QIC-02
        -------------------------------------------

The Storager will support both ESDI and ST-506 drives, but in these systems
is used only for ESDI and cartridge tape.  It too supports floppies
(3, 5, and 8 inch), but, again, I have no knowledge of their being used.

There have been reports of difficulty in using large ESDI drives with
the Interphase Storager, the suggestion being that there is a size 
limit on the drive that may be used.  In fact, the problem is not in the
size of the drive, but rather the higher data transfer rate that generally
accompanies larger drives.  Different versions of the Storager exist, 
capable of sustaining different data rates.  According to Interphase:

Storager        CC00047-****,Rev**      10MHz
Storager II     CC00058-****,Rev**      12.5MHz
Storager III    CC00105-****,Rev**      24MHz
Storager IIID   CC00119-****,Rev**      24MHz

For example, if you're running a Storager II with a Hitachi DK512-17 disk
(134 Meg, 10Mbit/sec xfer) and want to replace it with, or add, a larger 
drive, you may run into trouble with a Micropolis 1516 (678 Meg, 20Mbit/sec).
You'd likely have better results with a CDC 94186-383 (383 Meg, 10Mbit/sec).
This is largely anecdotal, and not based on personal experience.  
If you know otherwise, please supply details.

NOTE:  All references to the Storager in SGI documentation explicitly 
       state "Storager II", so I believe that to be the only version
       they shipped.  Either that, or they were sloppy in their field
       service docs and actually shipped the Storager III with the
       Hitachi DK514-38 (330meg, 15Mbit/sec).  Perhaps someone who has
       one of these drives will confirm the identity of the controller.

[/usr/include/multibus/si*.h]

4.2.1.3 Interphase 2190 SMD
        -------------------

This is the controller used with big (these days more a comment on 
physical size than capacity) drives such as the Fujitsu Eagles.  Unlike
the two previously described controllers, it does not have a QIC-02
tape interface.  Therefore, this board is used in conjunction with the
Qualogy 5217, which provides the catridge tape I/O.

(/usr/include/multibus/iphreg.h)

Interphase Corporation
13800 Senlac
Dallas, TX 75234
voice: 214-919-9000
fax:   214-919-9200

4.2.2   Drives
        ------

"Supported" Drives
------------------

The following is a list of drives known to the various versions of fex,
(fex is explained below).  Entries missing type or size information will, 
in practical terms, be just as useless as the 10 and 20 meg drives listed, 
but I again defer to completeness.  The first column contains a drive-type 
abbreviation used by fex.  (This list was created from the three versions 
of fex on my 3130's v3.6 os)

[drive sizes are formatted...calculate unformatted]

        Aim 130                         ?
        AMS 513                         ?
96202   AST 96202                       ?
96203   AST 96203                       ?
3046    Atasi 3046                      ST-506  39meg
WREN    CDC Wren II(86)                 ST-506  72meg
94156   CDC                             ESDI    40-72meg
WREN3   CDC Wren III                    ESDI    86-155meg
        CDC 9766                        SMD
        CMI 3426                        ST-506  20meg
D570    Cynthia D570                    ?
2243    Fujitsu 2243                    ST-506  67meg
2246    Fujitsu 2246                    ESDI    172meg
2249    Fujitsu 2249                    ESDI    389meg
        Fujitsu 2312                    SMD
        Fujitsu Eagle 2351A             SMD
5118    Hitachi DK511-8                 ST-506  67meg
5128    Hitachi DK512-8                 ESDI    67meg
51212   Hitachi DK512-12                ESDI    94meg
51217   Hitachi DK512-17                ESDI    134meg
51438   Hitachi DK514-38                ESDI    330meg
1085    Maxtor 1085                     ST-506  71meg
1140    Maxtor 1140                     ST-506  119meg
        Maxtor 2085                     ST-506  74meg
4175    Maxtor 4175                     ESDI    150meg
M514    Memorex 514                     ST-506  58meg
        Micropolis 1325                 ST-506  71meg
3212    Miniscribe 3212                 ST-506  11meg
        Mitsubishi 3426                 ?
        NEC 1055                        ?
QUME    Qume 592?                       ?
        Siemens 1100                    ?
        Siemens 1200                    ESDI    174meg
1300    Siemens 1300                    ESDI    261meg
T101    Tandon TM-101-4                 ?
TM252   Tandon TM-252                   ST-506  10meg
        Tandon TM-262                   ST-506  21meg
MK56FB  Toshiba MK56FB                  ST-506  130meg
156FA   Toshiba MK156FA                 ESDI    148meg
V130    Vertex/Priam/Bull V-130*        ST-506  26meg
V170    Vertex/Priam/Bull V-170*        ST-506  60meg
V185    Vertex/Priam/Bull V-185         ST-506  71meg
155015  1550-15?

*You can also interpolate the V150 here, even though fex 
 doesn't explicitly mention it.

"Actually Used" Drives
----------------------

These are the ones, as far as I know, that SGI shipped:

[check all of these]

2243    Fujitsu 2243                    ST-506  67meg
2246    Fujitsu 2246                    ESDI    172meg
2249    Fujitsu 2249                    ESDI    389meg
        Fujitsu 2312                    SMD
        Fujitsu 2351A (Eagle)           SMD
5118    Hitachi DK511-8                 ST-506  67meg
5128    Hitachi DK512-8                 ESDI    67meg
51212   Hitachi DK512-12                ESDI    94meg
51217   Hitachi DK512-17                ESDI    134meg
51438   Hitachi DK514-38                ESDI    330meg
1085    Maxtor 1085                     ST-506  71meg
MK56FB  Toshiba MK56FB                  ST-506  130meg
156FA   Toshiba MK156FA                 ESDI    148meg
V170    Vertex/Priam/Bull V-170         ST-506  60meg

Drives That Don't Work
----------------------

I haven't been able to get Seagate ST-4096s (that worked just fine
on a PC) to talk to the Qualogy controller.  mdfex just reports a
hard error when it comes up.  Anyone else out there have experience
with these (or other) drives?

4.2.3   Disk Maintenance
        ----------------

SGI provided three programs for disk maintenance: mdfex, sifex, and ipfex.
They live in your /stand directory, and are the first files copied to tape
when you do a "mkboot" in the course of doing backups.  They are identical
in functionality; which one you use depends on which disk controller (and
disks) you have:

Program      Controller         Drive type
-------  -------------------    ----------
mdfex    Qualogy/DSD 5217       ST-506 
sifex    Interphase Storager    ESDI 
ipfex    Interphase 2190        SMD 

Under normal circumstances (starting from the monitor prompt)
you will boot fex from the "mkboot" tape using, as appropriate:

b md0:mdfex     (if you have ST-506)
b si0:sifex     (if you have ESDI)
b ip0:ipfex     (if you have SMD)

fex will bring itself up, attempt to read the disk label, and present you
with a limited set of commands, which are largely self-explanatory.

To access the full command set, concealed from you for your own good, 
type a capital Z.  You will be prompted for a password:

Password:       carter
pAssword:       ludwig
paSsword:       chase
pasSword:       darrah
passWord:       donl
passwOrd:       bradley
passwoRd:       ellis
passworD:       luttner

You can get yet more commands by typing Shift 3 (#).

The new user is encouraged to fool around with fex early on, especially
if the disk was scrubbed when you got it.  If the system was bootable
and running when received, use your discretion.  You don't want to be 
learning about this stuff for the first time after your system has crashed 
and you have to reformat and restore from backups.

fex is a little idiosyncratic (particularly when you're trying to change
drive types), so familiarity is an asset.

In short, here's what you have to do:

        1. Start the appropriate variant of fex.

        2. Select the extended command set using password as described above.

        3. If fex failed to read the drive type from the disk label,
           select the drive type.

        4. Format and (if desired) surface test the disk.

        5. Enter bad block data if appropriate.

        6. Modify disk label partition table if desired.
           This is MANDATORY for non-Turbo 2000 systems - see below.

        7. Write the label to the disk.

        8. Copy the os from the mkboot tape to the disk.

           If system is a Turbo or 3000, use the "tape copy" defaults 
           (copy tape file 2 to disk partition a).

           If system is a non-Turbo 2000, you cannot use the defaults.
           First, copy tape file 2 to disk partition g (the boot record),
           then copy tape file 3 to disk partition a (the os).

        9. Assuming the copy operation(s) completed normally, exit fex
           and boot the machine.

Disk Partitioning
-----------------

If you have one of the above drives, you'll be able to use the default
partition tables contained within fex, providing your system is a Turbo 
or a 3000.  It is also possible to use a drive not in the list above
by employing the partition table from a drive with parameters (head and
cylinder count) close to yours.  There's some room for tweaking and
kludging here.

However, the fex default partitions WILL NOT WORK with non-Turbo 2000
systems.  These machines require a boot partition (g) for which the
defaults don't have the correct settings.  The couple of cylinders 
needed for the boot partition are pinched from the swap partition (b).
I have the boot partition parameters for the following drives only:

Drive                   Defaults                        Change to
-----                   --------                        ---------

Toshiba         b: 18870(111),   16830(99)      b: 18870(111),   16320(96)
MK56FB-1        g:     0  (0),       0 (0)      g: 35190(207),     510 (3)
                Root:(a) Swap:(b) Boot:(a)      Root:(a) Swap:(b) Boot:(g)

Vertex/Priam/   b: 17969(151),  17731(149)      b: 17969(111),  17255(145)
Bull V170       g:   119  (1), 115311(969)      g: 35224(296),    476  (4)
                Root:(a) Swap:(b) Boot:(a)      Root:(a) Swap:(b) Boot:(g)

REMEMBER!  THIS IS FOR NON-TURBO 2000 USERS ONLY!
           DO THIS TO A TURBO OR A 3000 AND IT WILL NOT BOOT!

This isn't to say that the Toshiba and Vertex are the only drives that
will work with a non-Turbo 2000.  The head/cylinder close-enough-match
rule mentioned above certainly applies here too.  We ran a 2400 with
an 80Meg CDC drive with a little partition-wiggling.
        
If you're running a non-Turbo 2000 system, PLEASE send us your
partition tables for addition to this list!  sgilabel will report
the partition table for a specified drive.

If you're looking for a mkboot (system) tape, pay attention!
------------------------------------------------------------

A common problem among those who have just obtained a machine is the
need to obtain an os tape after the anally-retentive previous owner
scrubbed the disk.  All in all, it isn't too hard to get someone to
provide a tape, but this must be kept in mind:

Be sure to get a tape made on a machine with the same type of disk
controller as yours (ie ST-506, ESDI, or SMD).  If you don't, you'll
wind up with a kernel with the wrong disk drivers.  I suggest opening
the box and checking the controller board itself to determine which
type it is, rather than relying on the system's model number.

4.3     VIDEO STUFF
        -----------

4.3.1   Video Modes
        -----------

The section on the Rear Panel Configuration Switches (above) suggests
each machine shipped is capable of two (2) video output modes.  Which
two of the possible seven (7) was an option determined at time of order.
SGI installed the appropriate DC4 board, and informed the CPU of the 
desired modes** with DIP Switch S2 on the IP2.

As you can see from this table, this switch also does some other stuff
unrelated to video, but I might as well keep it together here.  It would
seem that the hardware may permit a multiprocessor arrangement, but I
have no knowledge of it ever having been done.  Nor do I know of what
the ramifications of this serial port speed switch might be, having not
experimented with it.

For a clarification of how this switch is interpreted/reported by the
monitor, see section 4.1.4.2 above.

*Asterisks indicate factory defaults

IP2 S2 Settings
---------------

1                       Master/Slave
-                       ------------
C*                      Master processor
O                       Slave processor

  2 3                   Unknown
  - -                   -------
  C C*                  Leave closed

      4 5               RS232 Speed for ports 1-3
      - -               -------------------------
      C C*              9600 baud
      C O               300 baud
      O C               1200 baud
      O O               19.2K baud

          6 7 8         Display types**
          -----         -------------
          C C C         Mode 0
          C C O*        Mode 1 (seems to be the most common, anyway)
          C O C         Mode 2
          C O O         Mode 3
          O C C         Mode 4
          O C O         Mode 5
          O O C         Mode 6
          O O O         Mode 7

Unfortunately, I've inspected a number DC4s, and haven't been able to find
any of the -0x suffixes on the boards themselves, but it may just be that
all of my boards are -03s (probably the most common type).  Further 
experimentation and data gathering will be required.  European users in 
particular are requested to look at their DC4s and report whether any such 
numbers can be found (The full part number should be: 5000-090-0x). 

Mode**  Primary                 Secondary                       DC4
------  -------                 ---------                       ---

  0     60Hz Non-interlaced     60Hz Non-interlaced             -03
  1     60Hz Non-interlaced     33Hz Interlaced                 -03
  2     60Hz Non-interlaced     RS-170A                         -02
  3     60Hz Non-interlaced     European Video Standard         -05
  4     33Hz Interlaced         60Hz Non-interlaced             -03
  5     33Hz Interlaced         33Hz Interlaced                 -03
  6     33Hz Interlaced         RS-170A                         -04
  7     33Hz Interlaced         European Video Standard         -06

**Mode numbers are mine, not SGI's.

Observant readers of the above table will note five (5) unique video
output configurations in which the primary and secondary modes differ.  
The RS-170A and European Standard outputs were available in both 
genlockable and non-genlockable versions, for the total of seven.

Further, the video mode may be set from UNIX with the set60, set30, and
setntsc commands.  As the man page for set60 explains, this will only let
you select between the two modes installed as described above.  Selecting
a mode not installed in your machine will have no effect on the display.
European Standard, though, is not mentioned.  Again, perhaps European users
will enlighten us.

News Flash: If you've fiddled with the various combinations of CPU and
            back panel switches, and your monitor still won't sync (or
            you get the oversize characters/every second character
            missing syndrome), take a look at IP2 header H1, located 
            between U3 and U4 down by the P1 Multibus connector.  You
            may have to install or remove the bottom two jumpers, marked
            BPRN and BPRO.

4.3.2   Monitors
        --------

To the best of my knowledge, SGI never shipped anything but Hitachi
monitors with these systems.  There is, however, a variety of Hitachi
monitors.

4.3.2.1 Models
        ------

HM-4619-C-AC-2
--------------

This is the "bare" Hitachi monitor with top and sides that slope
toward the back.  I believe these to have been shipped with the 2000s.

Base model      Options
----------      -------

HM-4619                 19" 1024x768, Vsync 55-65Hz, Video 15Hz-100MHz
                        non-interlaced, separate sync or sync on green
                -C      Hsync 48-52KHz
                  -A    Housing 
                    C   P22 phosphor
                     -2 Digital Dynamic Convergence with Memory

HM-3619AC-11-2
--------------

This monitor is housed in a square, boxy case.  The keyboard junction 
box is housed within, in the lower left corner (as you face the screen).
The keyboard and mouse plug into the front, and the cable to the IRIS
connects to the back.  May have been standard with the 2400Turbo.

Base model      Options
----------      -------

HM-3619A                19" 1280x1024, Vsync 40-70Hz, Hsync 48-52KHz
                        Video 15Hz-100MHz, interlaced/non-interlaced
                        separate sync or sync on green
                C       Cabinet
                 -1     Long persistence phosphor
                   1    BNC connectors
                    -2  Digital Dynamic Convergence with Memory

CM2073ASG-517
-------------

This is the one in the nicely sculpted case with the tilt & swivel base.
Shipped with the 3000s.

Base model      Options
----------      -------

CM2073A           -517  19" 1024x768, Vsync 60Hz, Hsync 49.7KHz
                        Video 70MHz, separate sync, non-interlaced 
                SG      Silicon Graphics

I believe SGI also shipped these versions:

                  -511  19" 1024x768, Vsync 60Hz, Hsync 49.7KHz
                        Video 70MHz, separate sync, non-interlaced 
                        with Digital Dynamic Convergence

                  -515  19" 1280x1024, Vsync 60Hz, Hsync 63.9KHz
                        Video 107.35MHz,sync on green, non-interlaced 

Hitachi 15" tilt/swivel
-----------------------

I've got one of these (it's still dead) at another location.  When I
collect the numbers I'll add them to this list.

4.3.2.2 Monitor Docs, Parts, and Repairs
        --------------------------------

The following comments are based on experience in Canada. 
I'd like to hear from Americans (and others) on the subject:

Information on these monitors can be a little hard to come by.  Hitachi
stocks no replacement parts whatsoever for them, since they are OEM models
(as opposed to consumer product), yet they offer to sell the repair manuals 
for $200-$300 a shot.  Phooey.  The maintenance shop in your neighborhood 
University's Computer Science Department is often a repository for these 
docs, and they'll probably let you copy them for a case of beer.

This isn't the place for an in-depth discussion on repairing color
monitors, but here are a couple of general tips:

The switching power supplies in these Hitachis seem to be pretty good
about shutting down in overcurrent situations, so if the monitor is
totally dead, look for a short across one of the supply's outputs.
Odds are you'll find it in the horizontal output section.  Most faults
of this type are pretty easy to track down.

The 2073 in particular has a reputation for blowing its flyback
(primary to secondary short) *and* the power transistor mounted on 
the ceramic thick-film hybrid on the PC board down below the flyback.  
Whoever packaged this assembly was having a bad day and wanted to take 
it out on the world, because it's a couple of hours of work to get in 
there and replace them.  Before tackling this one, be sure to get the
schematic so you can continuity-check the transformer, which is pretty
complex, since it generates a bunch of lower voltages (-5, +12, +200, etc.)
for other boards.

One source of replacement parts is:

Video Consultant Service
100 Tyngsborough Business Park Dr.
Tyngsborough, MA 01879
voice: 508-649-4849
fax:   508-649-4843

They supplied me with the 2073 flyback and hybrid for $155 US, which is 
a little steep, considering the parts were RFE (Removed From Equipment).
I was in a rush, so I didn't haggle, but you might try.

4.4     HARDWARE MISCELLANY
        -------------------

4.4.1   Other third-party Multibus boards used by SGI
        ---------------------------------------------

4.4.1.1 Central Data CD3100 8-Port Serial
        ---------------------------------

I still don't have all the details here.  It seems that there have
been at least a couple of revs of this board.  I have a CD3025 that
came from an old NCR Tower system; the one SGI used may have been a
CD3044.  Other than using different I/O headers (3025 has two 34-pin
headers, 3044 has a 50-pin header (?)), I don't know whether there
are any functional differences.  I do know that the 3025 used a 2650
processor, and the 3100 has an 8088 and will emulate the 3044.  I'll
try to get enough information to piece this all together.  If your
IRIS has one of these boards, perhaps you can fill the rest of us in
on the details.

(/usr/include/multibus/cdsioreg.h)

Central Data Corp.
1602 Newton Drive,
Champaign, IL 61821
voice: 800-482-0315
fax:   217-359-6904

4.4.1.2 Ciprico Tapemaster 1000
        -----------------------

The Tapemaster 1000 (or TM-1000) is a Pertec-compatible controller
for use with a 9-track (1/2") tape drive.  The only thing to keep in
mind is that you can't boot from this tape.

Note:   This board is not compatible with the earlier 
        Tapemaster or Tapemaster A (TM-A).

(/usr/include/multibus/tmtreg.h)

Ciprico Inc.
2955 Xenium Lane
Plymouth, MN 55441
voice: 612-559-2034

4.4.1.3 Excelan EXOS/201 Ethernet Controller
        ------------------------------------

This is the standard Ethernet board used in all of the machines
being discussed.  

Adding a second Ethernet board
------------------------------

The Excelan board occupies two port bytes, addressed as follows:

bit      A15 A14 A13 A12 A11 A10 A9 A8 A7 A6 A5 A4 A3 A2 A1 A0
         --- --- --- --- --- --- -- -- -- -- -- -- -- -- -- --
block   |<------------J53------------>|<--------J52------->| 
pins    | 8   7   6   5   4   3   2  1| 7  6  5  4  3  2  1|

ex0(def)  0   1   1   1   1   1   1  1  1  1  1  1  1  1  0  x  0x7ffc
ex1       0   1   1   1   1   1   1  1  1  1  1  1  1  1  1  x  0x7ffe

Unlike most jumpers, 1=installed, 0=missing.  So, in order to change
a board from the default ex0 address to ex1, simply add a shorting 
jumper to J52-1.

(Based on information extracted from SGI by 
Randy Schrickel ([email protected]))

Changing the Ethernet address
-----------------------------

Changing the Ethernet address is a subject worth touching upon, because 
Exos was a real pain in the ass over this and wouldn't tell me how to do it.

Strictly speaking, changing an address is against the rules, since 
Xerox issues the addresses so that no two machines will have the same 
address.  This, however, is how many software vendors "copy protect" 
their code.  The "authorized" machine's Ethernet address is supplied
with the software in encrypted form, and when the code comes up it checks 
the encrypted number against the machine's actual address.  No match, no
run.  So, if you happen to find a tape of something interesting in a
Dumpster (tm), be sure to get the Ethernet address of the IRIS sitting
next to the Dumpster (tm).

IRIS addresses take the form 0800.14uv.wxyz (which the monitor will
report as 8:0:14:uv:wx:yz with leading zeros suppressed).

This address is burned into a 74S287 256x4 16pin DIP tristate bipolar 
PROM* (U20 on the EXOS/201).  Be forewarned that this PROM is soldered
into the board, and that a certain level of skill is required to remove
the device and replace it with a socket without damaging the board.
Consult a good hardware hacker.

The address PROM is coded in the following manner:

        PROM addr       PROM data
        ---------       ---------
          00-0F             0
          10-1F             v
          20-2F             u
          30-3F             x
          40-4F             w
          50-5F             z
          60-6F             y
          70-7F          checksum
          80-8F             1**
          90-9F             0
          A0-AF             0
          B0-BF             0
          C0-CF             0
          D0-DF             0
          E0-EF             0
          F0-FF            0-F 

* Equivalent parts:  AMD 27S21, Fujitsu MB7114, 
                     Signetics 82S129, TI TBP24S10

**At least this was a 1 in all of the boards I examined.  It's possible 
  that it's a variable part of the checksum and might have to be changed.

The address coding in the PROM is obvious, but what Exos wouldn't tell
me was how they generated the checksum, which must be correct, or the
board will fail its power-on self test.  I couldn't be bothered trying
to figure it out, and brute force costs six PROMs (worst case), so
that's what I did.  If anyone else would like to figure out how the
checksum is derived, I'll be happy to supply the raw data.

The way to tell whether you have it right is by the diagnostic LED on
the front edge of the board.  If everything's A-OK, it will blink
regularly (50% duty cycle).  If you've got it wrong, the LED will blink
a coded B9 (_.__  _.._) meaning "Physical address PROM checksum failed".

For those not so hardware inclined, Exos did say that one could change
the address with a memory patch.  This is all they told me, and I haven't
tried it, so let me know:

        EFD00   some version number
        EFD02   low  nybble byte 2
        EFD04   high nybble byte 2
        EFD06   low  nybble byte 1
        EFD08   high nybble byte 1
        EFD0A   low  nybble byte 0
        EFD0C   high nybble byte 0
        EFD0E   checksum nybble 1-6

Since the days when they made this board, Excelan was bought by Novell,
which then sold the Exos division to Microdyne, which (I think) used to
be part of Federal Technology.

Exos
207 South Peyton Street
Alexandria, VA 22314
voice: 800-255-3967
       703-739-0500
fax:   703-739-0558

4.4.1.4 Hyperchannel
        ------------

If you're looking to connect to one of those old Crays that keep
popping up in the surplus shops, this is it.  I assume that the
board was manufactured by Network Systems, but I've never seen one.

4.4.1.5 Ikon 10077 DR11-W Emulator
        --------------------------

The DR11-W is sort of an old de facto standard interface for
interprocessor communication and connection to peripherals such
as Tektronix and Versatec plotters.  These boards are referred to
as "emulators" because they emulate the Digital Equipment Corp.
DR11-W board.  DR11-W emulators can be bought for many buses,
and Ikon makes most of them.  So if you want to couple your IRIS
to the PDP-11 or Perkin-Elmer in the basement, this may be the
way to do it.

(/usr/include/multibus/ikcreg.h)

Ikon Corporation
2617 Western Avenue
Seattle, Washington 98121
voice: 206-728-6465
fax:   206-728-1633

4.4.1.6 National Instruments GPIB-796 IEEE-488
        --------------------------------------

This jumper information provided by Tom Haberlandt
([email protected]):

DIP Switch:     All ON except pos.11    (0xffd0)
Jumpers:        E1-E2                   (488 cable GND connected to logic GND)
                E4-E5                   (Serial DMA priority resolution)
                E8-E9                   (GPIB-796 doesn't drive Multibus LOCK*)
                E11-E12                 (16-bit I/O port addressing)
                E14-E15                 (GPIB-796 drives CRBQ* as required)
                INT 5

Incidentally, the board is all discrete logic (except for NEC 488 controller),
so it could be wire wrapped from the schematics.  No PROMs or PALs.

National Instruments
6504 Bridge Point Parkway
Austin, Texas 78730-5039
vox: 512-794-0100
     800-433-3488
fax: 512-794-5678

4.4.1.7 Sky SKYFFP-M-03 FPU
        -------------------

The non-Turbo 2000 systems used this Weitek 8440A-based FPU. 

Sky Computers Inc.
27 Industrial Avenue
Chelmsford, MA 01824
vox: 508-250-1920
fax: 508-250-0036

4.4.1.8 Xylogics
        --------

The standard distribution contains include files for the 
Xylogics 421 (ST-506 & QIC-02) and the Xylogics 450 (SMD) controllers.
Why, I don't know.  I don't believe SGI ever shipped them.

(/usr/include/multibus/xyfreg.h, xyreg.h)

Xylogics
Burlington, MA 01803
617-272-8140

4.4.2   Power Supplies
        --------------

All I have on the power supplies is an old phone number for
LH Research (714-546-5279).  It's a Southern California number, and
when I tried it was no longer in service.  If they're still in business,
perhaps SGI's purchasing department has a new number/address.  I haven't
checked.

4.4.3   CPU Battery
        -----------

The battery found on the 68020-based IP2 processor board (used 
in the 2400Turbo, 2500Turbo, and all 3000 series machines) is a:

Keeper II Lithium Battery
Battery No. LTC-7P
Eagle-Picher Industries Inc.
Box 130, Bethel Road, Seneca, Missouri 64865
phone: 417-776-2256
fax:   417-776-2257

Call them for the name of your nearest distributor.

4.4.4   SGI Multibus Extender
        ---------------------

If more than ten people out there think they might have use for an extender
board for these things (the card cage is too deep for a regular Multibus
extender), I'd be willing to lay it out and have it fabbed.  A minimum of
10 would probably set us back $60 per board (plus connectors).  The price 
would drop fast beyond 10 pieces.  Unless, of course, someone has a stash
of extenders that will work...

4.4.5   Loose Ends
        ----------

-Some information on the CG1 genlock board would be nice.
-I'd be interested in hearing from anybody who has a light pen.

5                       SOFTWARE
                        --------

5.1     OS VERSIONS
        -----------

The latest versions of which the author is aware are:

        2000            GL2-W2.5
        3000 (& Turbo)  GL2-W3.6

Rumors occasionally surface that suggest SGI had a 3.7 that was never
released.  It is unknown whether it was ever sufficiently completed for
even an alpha release.  Contact me for further details.

5.2     X
        -

The same rumor mills offer than an X was under construction along with the
3.7 os.  Not knowing anything about X, I can only relay the information I
have received from trusted sources:  That the graphics architecture makes 
it seriously nontrivial, if not impossible, to port X to these systems.
Addendum:  In late 1993 a discussion of this matter occurred in one of 
the SGI newsgroups.  I can provide a copy of the discussion on request.

5.3     GNU
        ---

>From time to time I've scouted around for ports of GNU stuff.  I'm sure
they exist, but I haven't found them.

5.4     APPLICATIONS
        ------------

As for commercial applications, I refer to the two editions of the SGI
"Geometry Partners Directory" in my possession:  1986 and Fall 1989.

The 1986 edition indicates whether each package is available for the
2400, 2500, Turbo, 3020, and 3030.

The 1989 edition lists only the "3000"; the remainder of the entries
are for the PI, 4D, etc.

I have had occasion to approach a number of these vendors to solicit
donations of software (for use on 3000s) for a worthy cause I'm
associated with.  Consistently, this is what I've been told:

        "Sure, we'd be happy to give you that old version,
         but we didn't keep any tapes of it."

Now, I don't want to get into a discussion about whether authors and
vendors of commercial software have (or should have) a professional
obligation to archive their work.  I enjoy working with a lot of
"obsolete" machines, so it's apparent which side of that argument I'd
be on.

That debate notwithstanding, the inference I draw from their replies is:
Since they don't care about the old versions of their code (even enough
to dedicate a shoebox in the basement to them), that software must have no
commercial value.

I therefore propose that a central repository be established for these
tarbabies (pun intended), from which users can obtain copies of the
software and related documentation on an exchange or cost-of-media basis.  

I appreciate that there may be readers of this document who have copies
of the software in question, sympathize, and would like to contribute,
but consider themselves bound by their original licensing agreements.
If that is the case, please contact me.  Judging from my prior
conversations with a number of vendors, I'm confident that I can arrange 
a letter of waiver, based on the promise not to distribute the software
for profit.

5.5     SOFTWARE MISCELLANY
        -------------------

Reading Sun tar Tapes
---------------------

The pre-SPARC Sun machines (like these SGIs) were 68k-based.  Despite 
this, tar tapes written on one machine cannot be directly read on the 
other.  For weird historical reasons once explained to me (I think it 
had to do with someone trying to maintain VAX-compatibility), they are 
byte-swapped relative to each other.  Which is which I can't remember.  
Anyway, here's what I do to read tapes written on a Sun:

        dd ibs=32k if=/dev/rmt1 conv=swab | tar xvf -

Of course, do make sure the tape you're trying to read is low-density (60meg).
 
Diagnostic Tape
---------------

If someone out there has one, please contact me!

Eebida eebida eebida that's all, folks!