The collected works of SAQ - Page 21

hamei wrote:

chicaneuk wrote: Sadly multimeters and oscilloscopes are not something I've ever owned or been familiar with - not an electronics guy at all

If you're going to get one or the other, don't be misled by the apparent "complexity" of an oscilloscope. In general they are really easy to use and tell you much more about what's going on than a meter. And the current craze for digital meters can be wrong-headed. Inexpensive digitals can't even see fast-moving changes.

Simple example : you check the voltage of your power supply. Digital meter tells you 4.93 volts. Cool, within spec ! Oscilloscope shows a base of 4.93 volts but more noise than the dining hall of a Chinese university. You need new caps !

An inexpensive 20 mhz one would probably be fine for anything you're likely to do. Just don't pick up one of those thirty-ton antique Tektronix thingies for $5 at the surplus store ... they look cool but not worth the hassle. You want small and light.

I'd go in the following order: Decent DMM (Fluke 70 series is a good general purpose unit), scope, analog meter (VTVM if you can). DSOs are sweet - once you get one you'll wonder how you managed with an analog. Downside: my LeCroy (pre-Windows) doesn't do X-Y mode . Don't know about them new fangled LCD thingies, guess they might. Actually can't remember about my Tek analog storage unit either. Troll surplus lists and used postings, scopes have always been expensive new.

Yep, mine has the cover on the top bracket.

They wouldn't leave it open from the factory because someone's just gonna stuff a floppy in there (you know it) - or something worse (especially if kids are involved).

Nope, SGI didn't do that. Sun did on the Sun4m lunchboxes, but never SGI. Being on the side it was a bit more hidden, and the Indy was the only one with an exposed possible floppy port anyway.

The ethylene glycol isn't needed if it's not going to be below 1C or above 90C, but the anticorrosion ingredients in auto antifreeze are a good thing, especially if you have mixed metal engines.

It would need some work...

GXEmul has some of the parts in place for IP32 emulation, but it doesn't work enough to run IRIX
MIEP is working on IP22 emulation, which may wind up being easier as SGI released the documentation for Indy and XL graphics.

Gotta love the backwards "y" - probably the worst thing in the whole SGI font. What were they thinking? Yes, it "matches" the G better, but there's a big market difference between SGI and Toys R' Us.

I'm getting less and less "gotta be the real thing" as I get older. Emulators are great - a wide variety of people can get the experience with much less frustration. It's possible to do even the formerly"special" stuff such as SGIs now that computers have crazy amounts of processing power in even low-end models, and more people can be exposed to alternative ways of doing things on the software side, anyway. I decided that while messing around on my VAX - sure, the DIGITAL storage interface systems are technically sweet, but it's frustrating to have to find all the little, very proprietary bits to get it working especially if you don't have experience with it. SIMH gives someone a dose of VMS much quicker, and without the "cross fingers and pray I don't get any smoke now that I've finally found this at a price I can afford" worry.

What would be really nice would be to have a full PowerPC Mac emulator, now that 10.6 is getting very old and the PPC hardware is getting old... I still have PPC/SysX and PPC/classic apps that I use.

I'm also deciding that there are many days when I don't want to juggle boxes to run stuff. Emulation and virtual machines require much less lifting, bending, and cord fiddling.

Vladio wrote: I got uucp installed and followed these parameters "Using Another Workstation as the System Console" in http://techpubs.sgi.com/library/tpl/cgi-bin/getdoc.cgi?coll=hdwr&db=bks&srch=&fname=/SGI_Admin/ChallS_OG/sgi_html/ch03.html .

When I type cu Direct at the prompt I get "connected" then it hangs, does nothing. any suggestions?

Make sure it's going out the correct port in the config file. If it's the wrong port (or if there isn't anything coming over the serial) it will appear to hang. Cu (and the similar BSD "tip") aren't very good at the whole "tell you what's going on" thing.

If you do much of the headless box thing than a network console server/ terminal server becomes very nice. Plug your machines' TTYS0/OPA0/console whatever in, plug a network feed in, then you connect to the box over the network (via telnet, ssh, whatever it supports) and via that connect to the machine. No cable swapping needed.

hamei wrote: It's been a long time since I did the overclocking thing, but as I remember the cpu boards from the 195 through the 300 were the same. 175 was different and 360 was different but in between, I think the boards were all the same. I believe you can put the 300 mhz chips into a 195 mhz module and just change a few resistors.*

Visual inspection time ....

* Unfortunately all those photos are gone from Nekochan because they were hosted off-site Shoulda caught that error in the first place

Excluding the cache chips (different speeds/sizes) and the SEPROM with the setup code. Also you need to keep track of different voltage requirements - 195 was first-gen, 300 used lower voltage core.

But they do use the same socket.

For a while I had a spare MSC, but I think I may have given it away. Will message if it turns up.

computron wrote: Hello

i'l looking for an IRIS 4D/20 complete Skin

i'm in europe

cheers

Eve

Typical. 4 months after I finally throw mine away because "nobody wants them"...

kjaer wrote:

SAQ wrote: Gotta love the backwards "y" - probably the worst thing in the whole SGI font. What were they thinking? Yes, it "matches" the G better, but there's a big market difference between SGI and Toys R' Us.

I'm not sure what makes you think the "y" is backwards...?

Don't have a SGI with me right now, but I'm pretty sure that in the "new new" System Manager with the "SGI font" (6.5.14+) the y in System has the tail pointing to the right.

computron wrote: hello SAQ

you have one ???

Eve

Nope, tossed them 4 months ago after saving them for 9 years.

Looking around I noticed that Netatalk is currently at 3.1.x, but all of the Linux distros package 2.2.x. My suspicion is that this is your common, garden-variety "you're not my friend anymore" Linux hissyfit, (especially as BSD packages 3.1), but thought I'd check in to see if anyone knew anything more in depth. 3.1 has some advances over 2.2, and unfortunately some things require Linux and I'd like to minimize admin overhead.

Have you tried setting "sgimode" on in the app-defaults?

I am off and on trying to build a IRIX 5.3+XFS for All Indigo2 IMPACT including 2MB SC, and I've gotten it to install on High IMPACT but not work right on Solid IMPACT.

I looked into some things and noticed that High and Max IMPACT were the first cards released, followed by Solid. I can't recall the exact timing, but I think that 5.3 was early enough that I suspected that it may not have had Solid support. The difference would be meaningful as Solid lacks the texture hardware that High/Maximum have. Look at the brochure on Ian's site: http://www.futuretech.blinkenlights.nl/ ... actpg.html - it lists 250MHz R4k, but only High and Max IMPACT graphics.

Anyone have a definite date for Solid IMPACT release?

I have Deb8/nVidia.

Indigo is one of the earlier ARCS boxes, pre-ARCS are really picky so it wouldn't surprise me if Indigo might be a bit pickier than later ARCS boxen such as your O2.

See if you can get your hands on a Plextor or Sun SL-CD. Those work nicely with SGIs. Toshibas can work also, but they seem to be more often broken for me (except for the 4101/SL-CD). I think I used a Plextor to install on my IP20.

Supplementary Applications can create a system that you don't expect - it removes Impressario and installs a demo version of ESP PrintPro (which was the $$ version of CUPS). There's also CDE on there, and an older version of OpenOffice 1.x (probably the most useful part).

My advice - ignore unless you know you want it.

I'd install with the overlay version of apps, then add what you know you want from the 1998 disk (demos). Don't do a blanket install, b/c you can get a whole bunch of outdated servers.

One thing that SGI with MIPSpro (or some other commercial UNIXes with mfr's compiler, say XLc or Sun Studio) will do is make sure you learn how to do things right (for the time period of the compilers, though). Linux/GCC lets you get away with much that IRIX/MIPSpro will flag as a warning or error out on.

That said, things are old enough to where you'll be a bit of a curio if that's your main platform. The latest OpenGL on IRIX is 1.4, for instance (and depends on your graphics option, so Express and IMPACT will be even older).

SGI may not care, but that does not mean that SGI's licensors will not care. Best to tread carefully and slowly, and not "tickle the tiger's tail" or some similar phrase that came up at one point.

I do a "dd" image, then a tarball of the contents. The latter is in case you reach a point where it is hard to extract the files from the image.

I started doing this with tapes. Originally I had dd images, but with IRIX tapes you'd need to write them out and then find a working IRIX box with a distcp that wouldn't mess up the header and require a new version of inst to install (e.g. if you distcp a 4D1-3.3 tape on IRIX 5.3 it will not install because it will now need a 5.3 inst to install, but 4D1-3.3 of course only has 3.3 inst). Now I have the tarballs and dd files, and it makes network install easier.

The 3-phase is the killer (some VAXes and IBM boxes, big SGIs, probably big other boxes).

Dexter - do you know anything about Graphicd Rollup patch 1332? I can't get it to install on any of my boxes, it has a depends failure.

ivelegacy wrote:

jan-jaap wrote: Tapes don't support directories

this is a problem for me, I need my folders
but, I can put a tar (Tape ARchive) on tapes

Not sure completely what you're trying to do then - distcp will create a install tape set. If you're trying to create a bootable backup set your backup tape will generally be different from the tape you have as a bootable install. If you have the CD-ROM and a drive then you can boot off of that and then restore off of any tape (...that is supported by the IRIX miniroot...) you want.
It's been a while, but I think SGI's "official" archive program was cpio. Dump and restore (or xfsdump/xfsrestore) are good to look into as well.

I think there was one model of DEC tape drive that "looked" like a block device, the rest never had directories- at least ones that IRIX will be able to communicate with.

Found this - seems to be pretty straightforward: http://www.backupcentral.com/wiki/index ... l_Recovery

Of course techpubs also has a good and official doc, but it ain't up anymore.

foetz wrote: simply because sgi was more popular for desktop use hence in comparison there's more available

+ the cool factor. SGIs were (and are) "cool", so it's easier to maintain a hobbyist community to go to the work of porting, packaging and maintaining a software repository.

OpenVMS has some, although update frequency isn't as often as Nekoware. Solaris used to have a lot, until Oracle went Oracle on Solaris (but OpenCSW is still around). DEC Unixes likely suffer from BSDitis - NetBSD has an excellent selection of software, and (especially for VAX) BSD is historically considered to be "the UNIX" for DEC. Note that that is pretty much "science by Aristotle" (go into a closet, ponder something for a bit, come out and make a big pronouncement without any verification first), so don't give it too much credence.

AIX is IBM, and IBM is, well, IBM, so don't expect too much there. It's better than OS/400/i or the mainframe OSes, so be happy, I suppose... Groupe Bull still has their freeware collection at http://www.bullfreeware.com/ , which is targeted more towards servers (naturally), but exists and is kept updated.

I know of nothing for SunOS 4 or Tru64.

At this point I would replace any accessible services with ones from xBSD or Gnu/Linux. Probably even consider a SSH or VPN host to go through to get to the SGI. Not sure about client apps on the SGI - I do occasionally use Firefox, but generally for "trusted" sites (for whatever degree of validity "trusted" is ... not much).

Krokodil wrote: HP-sUX(lol)

That dates from the Unix Haters Handbook era (along with Slowlaris, ScumOS [but it boots fast!] and the like). HP-UX 10.20+ isn't a showoff OS (like, say, IRIX), but it is solid and stable.

HP did go through a period where they thought they'd challenge SGI for graphics, but the software end of it just wasn't there.

Another thing to look at is the number of connectors. Indigo had HP1/HP2-> backplane and backplane ->graphics, and both of those had to be up. Indigo2 and Indy were both SBCs with graphics on them - though I2 did have the EISA riser board connection. In a similar vein SGI simplified the Octane/Origin architecture to a SBC with I/O expansion riser for Fuel/Tezro/O3x (though cost reductions in other places took their toll there).

Keep in mind that there's speed and there's latency and you have to look at them both together. ATA100 ain't super fast, but it doesn't have to go through quite so many layers.

Kind of a buzzkill, but do keep in mind what you will actually be doing with the machine. Does it need to have a better disk interface? Will you really be transferring that much data and/or making use of applications where command queuing will make a significant difference? Do you need more than the max disk size for ATA100? If so but it isn't that often than NFS may be a good solution that's a bit less trouble/expense than getting a compatible SATA/SAS/SCSI card. If not than as-is ATA100 might work just fine.

Don't panic about the compression connectors. The rumors are worse than the reality.

Just don't be egregious. Don't drag them through centimeter-thick dust, don't grab them with your hands, don't clean them with a paper towel and Windex. I've inadvertently bumped mine gently and they still work.

Wouldn't surprise me - Linux is the buzzwords, so it is by definition buzzword compliant. Add on that, while somewhat mediocre and less committed to standards and long-term interoperability than Apple is, it's cheap and you have a "winner" as far as MBAs are concerned.

Yeah, Oracle should have priced it more in line with an equivalent Windows installation. I remember being dumbfounded when my university "replaced" their almost-new Sun Enterprise setup with a Winfarm in the NT4 days (evidently someone from MS gave lunch and licenses to someone high up in admin with little actual tech experience). It was supposed to be "more reliable", but it seemed to break more often and they had to keep up the Suns anyway for the CS department. Since then Windows, while not my favorite, has become functional, and at least at first glance competitive or cheaper than Solaris (though the whole CAL muddle still is a mess, but profitable to MS).

I've never used Sol11, because the licenses are too expensive, but it still has some technical merits.

Be nice if at least ZFS made it out in a easily-used (say BSD) license status, but I'm not counting on it.

I've never really trusted FreeBSD - there's something wrong about a BSD that doesn't support any DEC architectures, but that's just on a gut level as technically it seems pretty good.
That said, even I realize that ZFS on a VAX would be ... painful. Even NetBSD can get slow if you aren't careful what you load your box up with.
Anybody know if Oracle really took on Sun in good faith or was this a HP-style "we promised to continue development for X years and by golly we had the day we could drop it circled on our calendar before we even signed the deal"

pentium wrote: I swear I made a post in here several days ago....


The video is great. The only complaint is he drags his heels and talks about the keyboard being rubber dome and the quality of the mouse. Hey man, neither were issues in the 90's. Mechanical crap only became a fad in the last half a decade.

To be honest that is about the worst SGI keyboard they shipped, though I haven't used the later post-granite ones. NMB was used by a bunch of
PC makers as well, whereas the big slab that preceded it does feel nicer.

I am, and have been, a keyboard snob though. On my computers, typewriters and pianos.

The slightly galling bit is the implication that the lack of an emulator is just because of lack of interest, instead of the collection of proprietary chips SGI used, the documentation for which has not been released and is likely lost.

If I understand properly the PSU is shutting down, correct? If so then that means that something is either going wrong in the PSU (voltages off), there's something drawing too much power in the computer and tripping the protect circuit, or the PSU can't put out the needed power. What exactly was "todays trial?" Yes, SMPSUs need a load on the main rail, but if the PSU is failing you might as well fix it now to support the load you need rather than trying to halfway hack something that kind of works. Try pulling all boards but the backplane and seeing how it runs with several disk drives for load. Indigo does not have soft power, so that won't be an issue.

No rails. There are brackets that bolt it in to the rack and heavy angle support brackets to take the weight on the bottom.
Officially it has the plenum to direct the air to the front/back, but I pulled mine on my O2k so I can drop the lower module and put some more equipment in the rack. It blows down so I'm not worried about dust.

The other issue is that documentation for the silicon starts drying up at about the R12k. You can find lots on the R10k and earlier, but after 12k it was pretty much SGI-only. I don't have much that new, but on the earlier chips multiplier was set on startup from a serial EPROM, along with cache size and timing. Indigo2-era used jumpers, Origin used the flash utility, can't recall about Octane (though I do have a busted R12k IP31 PIMM @300MHz that I'm considering transplanting to my 2x250 R10k).

N.B. that voltages HAVE been known to change within the same processor number - R4400 has 5V and 3.3V parts, and R10k >200MHz is 3.3v, <200MHz is 2.5V. I have never seen docs on R14k, so can't speak to that one.
I'm not sure if the 550 -> 600 jump would be really worth it, especially given the risks.