The collected works of SAQ - Page 7

You'll need to either use low resolution or get a Cosmo board, as the Indy 10MB/s SCSI interface can't handle full NTSC resolution video.

To an extent. IRIX 5.3 for Indy including R5000 works on all Indys (R4k, R5k), and may work on Indigo2 (probably will), but it will not work on any non-FastForward/IP22 (Indy IP24 is a small IP22) machine.

While it would have been very nice for SGI to qualify all releases for all supported systems, it would have taken way too much time to run the tests against everything, especially for the 5.3 releases.

Note that this doesn't work for everything, you want to look for the "including" (Indy including R5k) as there are some releases that are for a specific hardware version and won't work on others (the confusingly named "IRIX 5.3 for all Indigo2 IMPACT" only works on IP22 IMPACT machines and no others). The naming isn't even constant - IRIX 6.2 for Indigo2 R10k will work on Indigo2 R4k and Indy, but doesn't have the "including", so I guess you just have to know. Gerhard's pages are a good start, as are SGIs release notes and whatever is printed on the CD (sometimes a supported system list is printed)

Supposedly that version has issues with the 2MB cache processor cards, R5k, and IMPACT graphics, other than that it's pretty general.

There's a good story behind this one to make up for the lack of silicon graphics. Back when I was in grade school we played guts in gym class. all the time. also dodgeball. (are kids still allowed to chuck hard rubber balls at each other or was that banned along with peanut butter and sugar pop?) [/quote]

Dodeball is still very much played, but they have these great balls now called "Rhinoballs" - they're the Nerf foam-rubber material on the inside with a fairly durable rubber/neoprene coat on the outside and don't cause any damage to living creatures.

Bursting your bubble - Photoshop was hardware accelerated on the Mac AV models as well (and you could also get DSP add-in cards that would work, too).

You're running up against the reason that UNIX guys snort at the Jurassic Park line "this is a UNIX system - I know this". That line is hubris exemplified. Really, though, IRIX is much closer to other UNIXes than, say, Windows or OpenVMS (which is really a different beast). IRIX is great (along with some others, such as DEC's old OSes (Tru64/OpenVMS) and recent Solarises in that all the documentation is available online and is easily organized - and for IRIX it goes back to 5.3 (1994!). SGI was at the forefront of the Internet for support at that time, pity that they've been going through and removing patches over the past couple of years.

XBOW rev 1.2 was the subject of a recall by SGI, so my guess is that there were some issues there.

You have a license for IRIX, since it is tied to the machine. What you do need to do is buy or borrow media, as SGI doesn't make it available for download.

Machines can be CrayLinked/NUMAlinked as long as they are of compatible architecture and have the same speed NUMAlink ports/midplanes/etc. All O200s are the same, so all you need is a cable.

I wouldn't recommend changing PIMMs until you have much more experience. Many SGIs use processor modules, which incorporate the oscillator and setup PROMs, but Origins need to have clock speed, cache, and other data flashed into the nodeboard PROM whenever the setup changes, and if it is done wrong the node will not boot. You can correct this if you have a second node to CrayLink together with the first, but it's a bit difficult.

Linux mostly supports IP27 (Origin 200/2000), but you'll wind up with a machine that isn't much different than a dual-socket Opteron in use(except that it is noisier, takes up more space, and is slower). Stick with IRIX for now. OpenBSD also supports IP27 on O200, but I'm not sure about CrayLink.

PROMs usually get updated during IRIX installs, otherwise look up "flash" on TechPubs.sgi.com (the man section).

For classic Mac systems you can just copy everything over using the Finder.

I loathe and despise this trend towards three single word not-quite-sentences as a motto, a trend evidently devised by people who failed English repeatedly.

Used sparingly by authors who know what they're doing single-word sentences are OK. Their large-scale use by marketing types should be suppressed, violently if necessary.

They did. Unfortunately it was too late when the port of OpenOffice (SGI contributed to development) made it to IRIX. For a while there was WordPerfect/IRIX, but in common with most UNIX apps it was quite highly priced.

It depends on where it's stopping. Is the problem in the PROM or in IRIX (does it give you the PROM menu or not?)

What was the original processor/IRIX combination in the Octane that provided the mainboard/drive? There are slight changes needed for the R12k, but those were included from IRIX 6.5.4 I believe, so it is very likely that support for R12k has been installed (though not guaranteed).

I can see why administrative types would be quick to say "no" to including software - there are so many different licenses and some of them are transferrable without any paperwork, some require paperwork and fees, some are non-transferrable. It's work, and most admin-types aren't interested in the "computers as fun" part, so they see the machine as an asset or liability to be disposed of in the most convienient or required manner.

For O200 all IRIX licenses are transferrable. As of about 2004 SGI revised their licensing to a Sun-esque model where the license self distructs when the machine leaves the original purchaser's hands. Chances are Rackable/SGI isn't going to come gunning for you if you're just a hobbyist, but just so you know...


Going from 1 processor to 2 does require reflashing the setup PROM, otherwise the machine will only use a single processor. It isn't too hard, but you do need to keep track of a bunch of numbers for all the characteristics of the setup. I've done it, and it isn't impossible, but I would recommend setting up a time where you can call/PM/IRC a Neko Guru if you run into questions. As another possiblity you can replace the whole "logic carrier" assembly with a new IP29/PIMM that is already flashed to the proper speed. It's a bit more to ship, but it is the safe route. Three processors isn't bad, so I'd recommend getting a feel for the machine and IRIX before diving into modifying hardware.

Why sell all of them? If you still like the hardware and software keep at least some - it's still the same. You can probably prune, though - the lunchbox Suns are fun and small, as are the SS20s. The Blade1k is a good box, and I might keep the Ultra-2 as well (good design, fastest SBus desktop). If you get 300MHz + processors it will wipe the floor with your Netra, and with a shelf can be mounted in a rack. Haven't worked with the Blade chasses, so can't comment there.

To add another dimension to Japes' argument in favor of software RAID on commoditiesque hardware:

If you have a dedicated proprietary RAID controller subsystem (I would have said "hardware RAID" except that I've already been chewed out once by someone who pointed out that since it wasn't implemented using dedicated logic but rather a microprocessor running a custom program it was no more "hardware RAID" than any other software setup) you have the additional lurking time bomb of other controllers very likely not being able to read the array, so if your controller dies you need to acquire the exact same model or else restore from dumps. Software RAID is much more forgiving of being moved to different equipment.

HP PA-RISC boxes are interesting, but I have to then admit that I use my HP9000s about the least of any of my workstations. They're more the sort of machine you get because you have something that needs to be done and HP-UX is what it gets done on. The ones I have (G70, C180, B1000) are very well built, and they seem to be quite reliable. HP graphics are somewhat interesting - at some point they decided they'd try to challenge SGI on the desktop, and so they released the Visualize-FX line. Stats-wise they're not bad cards, and many run a variant of the PA-RISC processor for geometry. Unfortunately once HP did the hardware they decided they were done, and the Visualize-FX series never got the back-end and software support that SGI has.

HP-UX is pretty vanilla but sadly, unlike Solaris or IRIX, there really isn't much in the way of precompiled freeware selections. SAM is a reasonably-good hand-holding tool for administration, what it looses in hidden occult workings against AIX's SMIT is offset by the more-comprehensible nature of the underlying OS. Newer boxes sport USB, but in common with IRIX there's no block USB device support.

If you can lay your hands on HP-UX 11i it'd be worth playing around a bit, as HP-UX is one of the few proprietary UNIXes still kicking and likely to stay that way (along with AIX and probably Solaris).

Around the mid '90s wasn't Aldus Premiere the go-to software for presentation? If I recall right the only reason PowerPoint took over was because of bundling (and then MS started to make it better).

Of course, few people had the LCD overlays for the projectors, so most big presentations weren't computer assisted.

Unless something very unusual happens they won't. They could spend millions developing the R22k and millions developing V16 graphics or InfiniteReality5, and for the most part Intel, AMD and nVidia would match them speed-wise in 2 years. Another big SGI strength (probably the major remaining one) is NUMA, which is now commodity on the scale that would be used for the desktop. Their best graphics bet would be to come up with some sort of scaling out to many commodity GPUs that really works well and is transparent to the user and apps programmer (IOW do UltimateVision properly), and then bundle it with UltraViolet, Altix IA64, and possibly a really well made desktop multisocket NUMAbox (dual or quad socket). The only way they'd be able to get back in the proprietary business would be if the government sent them a big check, and that's highly unlikely - they'd send it to Oracle or IBM (and with the cancellation of Rock Oracle doesn't really have much in the way of selling points). As Skywriter has pointed out you need to move up the chain to survive now - come up with something to harness masses of commodity silicon easily, either with glue logic or software. That was the other big positive of the classic SGI - they made top-drawer graphics, yes, but they also made the software resources to easily use them (IRIS GL/OpenGL/OpenInventor/etc.)

Sadly IRIX is also unlikely to come back. You'd be a fool to get involved with the flaming remains of USL. Sure, SCO is write-off material now, but there will be someone else with a brilliant way to squeeze money out of dead IP who will pick up the pieces, and you don't have to loose lawsuits to have your company killed off by lawsuits.

MBus speed, used cache size (meaning that 1MB/2MB doesn't make a difference for SS10/SS20), chip family (SuperSPARC, SPARC, HyperSPARC) and MXCC version both have to be the same. MXCC 4.x introduced some significant changes over MXCC 3.x and you will run into problems.

Minor MXCC revs should be fine - give your config a go.

You're forgetting the marketing potential of having a different interface and different file format. Your organization might be doing just fine with Word 97, but the new people coming out of school where they have the new software (that MS almost gives away) look at your software and say "how do you do ____? I don't know how to do _____!" and MS points out the lost productivity because you didn't "upgrade". For the home users, well they start getting stuff in an "incompatible" file format. True, you can download the filters from MS (back to a point), but how many go out and buy a copy of Office 2007 just so they can open .docx files? Probably a not-insignificant number.



Again it's marketing, same deal with the "why do we have so many Linux desktops trying to do a mediocre-to-passable job of emulating Windows with all its UI gaffes". Management or end-users see it and recognize it and think "this looks like what it's supposed to (i.e. MS), it must be OK." That and the trend today emphasizing flash over substance (there must be more people out there who find transparent windows irritating, distracting and occasionally confusing).

Getting my cane and hobbling back to my rocker to growl at the passing kids...

Not really. I'm just pointing out that now generally a big focus of the change is on the look because the substance doesn't sell as well. The people who are interested in the substance often don't control the checkbooks, and many end-users don't really understand or care to understand the substance. I was pointing out how you can use targeted UI changes to encourage upgrades where there might not otherwise be a sale, although I am slightly irritated that MS felt the necessity to take a standard and craft a new proprietary format around it that they claim is "open" when there are plenty of other true standards that work.

Reading the other posts (especially R10ks) it seems like exactly the opposite is being argued - that there really isn't any substantive change going on at all, it's just the window dressing (and features of dubious usability such as the infamous "auto summarize"). The true, groundbreaking changes often seem to stall out.

Yep, I2/Indy both have the RS-232/RS-422 device-selectable interfaces

Not sure if it looks at the buffer chips, but my guess is that you'll be able to fool the system to run in -232 only mode easily enough. You might need to wire jumpers in place of the tristate, but it should work.

Instead of looking for a full IP28 you can just look around for mainboards from systems with similar serial port hardware. IP22/IP24 are obvious candidates, and IP20 probably has it as well. Chances are someone here will have a parts board that has something wrong with it that's probably not the serial drivers.

Might be upstream, then. Does IP28 use the IOC2 (integrated I/O (kbd/mouse/parallel/serial) used on Indy) or separate UARTS? The SGI docs indicated that they were considering moving to IOC later on in the Indigo2 lifecycle, but I can't recall if they went through with it.

It'd be a nice VMS box or Tru64 box. NetBSD will support Alpha for a while, but Linux is starting to move away from it.

OpenVMS Hobbyist comes with enough layered products to be useful - it even has the license for DECwrite (a old but still usable word processor). VMS makes a good Internet server also, as it can be locked down very tightly and has few bugs.

Not sure about ES25 - are you sure it isn't a misread ES45? ES45 is a SMP deskside/rack box.

I can't stand Jobs, but he's great at business.

Now that 10.4 is completely unsupported (read: no more security patches), what are the Neko types using to run classic Macintosh applications? As I see it, the options are

(a) keep around an old PPC or 68k Mac (or both, or more than one of each) to run the stuff on, but that gets bothersome with moving boxes around (no remote display).

(b) keep a 10.4 on PPC box around to run it (same as above, since 10.4 is getting less usable with the discontinuance of security patches <rant> LESS THAN 3 YEARS AFTER THE RELEASE OF 10.5 - how does Apple expect anyone in IT to take them seriously if they don't commit to 5 years - even MS does that! </rant>

(c) use one of the emulators to run System 9 on a newer system (Macintosh, Linux, xBSD, whatever), and if so what one (SheepShaver, Basilisk II, PearPC). Sadly most of them seem to be stalled out as far as developments. SheepShaver and Basilisk II seem to be the best (for PPC and 68k respectively).

(d) something else?

Industry-standard (as epitomized by IBM, HP, SGI, Sun, etc.) is support for 5 years after last ship.

Apple pretends to have servers (Xserve, OS X Server), but their support belies their marketing department evidently.

I think it was similar to the DSP accelerator products for other architectures in that certain plugins/ops were sent to ICE.

I guess the really irritating thing is that Apple builds a pretty good product. Really, they think about the whole system and plan for it - and then they go and do annoying things such as:

Stopping support early (not just software, the hardware/firmware gets shut down fast, too). I don't think the have any sort of support commitment.

Doing their best to eliminate customization (when Apple was headed to bankruptcy you could find plenty of add-in hardware and little things to make the system more custom, now just try to find a higher-end graphics board let alone more exotic hardware options - or for that matter a slot to put them in (Mac Pro is the only one with expansion vs. most Mac models in the '90s). Even the little software hack to replace the white Apple with "Welcome to Macintosh" was broken in the next release).

Not updating their lines or cutting prices. A year ago a Core2 Duo was OK - now try to find one from anyone except Apple, who happen to want the same price that they did before. This might have worked in 1990, but not now.

Whaddaya expect from a calculator created by computer science people?

Much better than one calculator I tried (think it was the Windows calc) that didn't reliably do order-of-operations correctly.

That part's easy - just boot from the miniroot on the Overlays Disk 1 tarball and make sure that you open the overlays as well as the base disks.

The only other thing -12V would be used for would be RS-232 levels. You could probably live OK with it disabled - try it before buying a new (probably expensive) board.

Another possibility would be to rig up a circuit to inject -12V where needed

For IRIX: you have one person who knows what they're doing read it and document the interesting bits (DGL, GL, graphics hardware, ARCS PROM, drivers, etc.) and then the documentation can be passed on to people to program support into xBSD, Linux, what-have-you. In short it would make reverse-engineering much easier.
The other, more legally questionable, thing you could do is integrate in bugfixes and security patches from other code.

For Windows, not sure what you'd do. Help the Eastern European/Chinese/Russian criminals break into more systems? Send MS suggestions that they'd ignore?

Has anyone here worked with the "Raven-040" products from MicroNet? They're some sort of RAID that runs over the Q900/950's integrated SCSI ports. Not sure if the control is software or hardware based.

Note that performace will still suck relative to modern drives b/c the Quadra's SCSI ports are slow (5MB/sec sync SCSI). Chances are that a fast modern drive will saturate the port more than the old MicroNets, even with the array.

You would need to find some way of laundering the process so your (xBSD/Linux/X.org) code wouldn't be contaminated, which adds another layer of difficulty, but not nearly as much as trying to reverse engineer from scratch would be.

Works great if you're on Solaris 8 or above. Not so good for early Solaris 2 users, and hopeless for SunOS 4.1 users. SunOS 4.1 and earlier are very different from Solaris 2, being BSD based rather than Vr4.