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 184.108.40.206 The LED Display 220.127.116.11 Configuration Switches 4.1.5 Internal Cabling and Board Settings 18.104.22.168 CPU Side Board Locations 22.214.171.124 IM1 RAM Switch Settings 126.96.36.199 Graphics Side Board Locations and Cabling 188.8.131.52 BP3 Bitplane Switch Settings 4.2 Disk Stuff 4.2.1 Controllers 184.108.40.206 Qualogy ST-506 220.127.116.11 Interphase ESDI 18.104.22.168 Interphase SMD 4.2.2 Drives 4.2.3 Maintenance 4.3 Video Stuff 4.3.1 Video Modes 4.3.2 Monitors 22.214.171.124 Models 126.96.36.199 Docs, Parts, and Repairs 4.4 Hardware Miscellany 4.4.1 Third-Party Multibus Boards 188.8.131.52 Central Data Serial I/O 184.108.40.206 Ciprico 9-Track Controller 220.127.116.11 Excelan Ethernet 18.104.22.168 Hyperchannel 22.214.171.124 Ikon DR11-W 126.96.36.199 National Instruments IEEE-488 188.8.131.52 Sky FPU 184.108.40.206 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 Mountainview, 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@example.com 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 -------------------------- 220.127.116.11 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 18.104.22.168 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 (firstname.lastname@example.org), 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. 22.214.171.124 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) 126.96.36.199 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 188.8.131.52 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. 184.108.40.206 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 ----------- 220.127.116.11 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 18.104.22.168 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] 22.214.171.124 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 126.96.36.199 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. 188.8.131.52 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. 184.108.40.206 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 --------------------------------------------- 220.127.116.11 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 18.104.22.168 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 22.214.171.124 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@example.com)) 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 126.96.36.199 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. 188.8.131.52 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 184.108.40.206 National Instruments GPIB-796 IEEE-488 -------------------------------------- This jumper information provided by Tom Haberlandt (firstname.lastname@example.org): 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 220.127.116.11 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 18.104.22.168 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!