The collected works of jan-jaap - Page 24

kramlq wrote: Did you retry any downloads again just to be sure?

If you go to the page for the CD: https://archive.org/details/sgi_O2_Demo ... ing_R10000 and hover over the "ISO image" download link, it says it's 212.2MB. That number is way short.

cris_adder wrote: It is worthwhile just trying to burn one of those shorter images to CD to see if it might work? I thought that some CD tools created larger images (with some sort of buffer bytes)

We're not talking about some buffer bytes here, but a ~ 600MB image truncated to ~ 200MB. Nu burner can make up for that

I skipped the physical media and tried loopback mounting the disc image in Linux. If you try to copy the files off, you'll get I/O errors at some point.

robespierre wrote: I can't see any reason you would install analog CCTV for that purpose today.

OP didn't mention what sort of VBOB he intends to use, but DMedia breakouts for DM2, DM3 etc. are all digital: SDI and HD-SDI. Cameras with (HD)SDI outputs are not all that uncommon.

Regardless, (HD)SDI is uncompressed video. A single HD-SDI stream can be > 200MB/s so this wouldn't be very practical.

SkyBound wrote: the numbers on the board (30-1235-001 REV A) does not return anything useful.

030-1235-001 is the PCI Riser for an Origin 200, not an Origin 2000 (O2K)

Did 195MHz CPUs even exist in the Origin 200? IIRC there's 180MHz and 225MHz (and faster).

I read somewhere today that IBM's Watson is capable of passing the CISSP exam

foetz wrote:

hardware this old is not reliable anymore. you better not bet the farm on this stuff to keep trucking along

quite the opposite in my experience. my sgis run; outlived all other stuff i had in between. the cheap x86 junk in particular.

The old hardware may have been constructed with a longer service life in mind than the current "cheap x86 junk", I doubt it was designed for an expected lifetime of 15 ~ 20years, the age of the average Indigo2/Octane today. Even the best elcos have a limited lifespan before they can expected to fail. The same for mechanical bits, fans, disks etc.

Also, while reliability of current hardware may be limited by the relentless pressure to keep prices down, the reliability of SGI hardware was often limited by the fact they were pushing the technological limits. Already back then the RMs of Onyxes would fail, and you were crazy to let the service contract expire if you depended on the machine. But people accepted it because the Onyx could do things no other system could. IIRC the entire first run of R10K CPUs was replaced, and the R8K wasn't all that reliable either. I think they even gave free upgrades to R10K to some customers.

The Phobos P1000 works on IP27 and IP30 as well, and I suspect it works quite well on those platforms.

The other way around, the QLogic QLA12160 or QLA2342 (dual 2Gb FC) work very well on Octane or Origin 200, but poorly on O2. And my O2 has a faster CPU than my Origin 200. The O2 PCI interface is simply under performing, probably due to the way it interfaces the UMA architecture of the O2.

jodys wrote: Last time I used the CDROM on my O2 I had some strange errors. These CDs are getting on in the years.

I think it's more likely the CDROM drive of the O2 is getting old. But backups never hurt.

pentium wrote: I've always dumped to raw disk images using dd.

Same here. If using an IRIX system, be sure to back up the entire disc (s10), not just the EFS partition (s1). But any Linux or OSX machine will do.

I have a little script which reads the .IM file from the CDROM root and renames the image accordingly, e.g. the CD with this .IM:

is renamed: 812-0707-001_MIPSPRO_C_COMPILER_7.2.efs

I somewhat order CD images by IRIX version or whatever else is common between them, but it's rather vague Many CDs work on many IRIX versions or were part of several distributions.

Rev.Bubba wrote: I had planned on using a switched PDU to allow for a system not in use to stay offline until needed and also for reboots. That would be rather simple as well.

robespierre wrote: You could automate the earliest stage of installation by using a terminal server connected to each machine's console port.

I have much of this in place: I have a pair of APC 7953 PDUs (one in my 19" rack, one behind the workstations). This allows me to remotely switch power to just about anything. Admittedly, my motivation was not having to bend over an Onyx ever again to reach for the breaker, but it works really well.

I also have an Cyclades ACS48 Console Server connected to just about anything with a serial console port. The amount of wiring is .... something.

Both of these I picked up for reasonable amounts of money and because they are enterprise hardware they come with neat goodies like scriptability, support for Radius etc in addition to a web interface.

As far as RoboInst is concerned: I believe it works best if the dist server and the target run the same IRIX version, and I also believe it requires a license. But I never messed with it myself.

eBay & co then. Whether that's legal depends largely on where you live.

Sacrifist wrote: Workstation eh? Might need decent graphics adapter support. Are there open source drivers for AMD/NVIDIA (or other notable) graphics on POWER/PPC?

Open source drivers: yes. But if I understand correctly, the claim to fame of this platform is "open", "owner controllable" etc etc. So no binary firmware blobs. Good luck finding an AMD/NVIDIA graphics card without those.

First of all, try to reseat it and re-run the test (irsaudit).

Then, if it really turns out to be dead, pay attention which half (raster or texture board) failed. If the raster board failed: try to buy an RM9. An RM10 is essentially an RM9 with an upgraded texture board. RM9's are cheaper and easier to find. Use that as a raster board donor.

Do pay attention to ESD when performing this kind of transplantation, though.

pentium wrote: Has there ever been a push to update the adapter using cheaper and more modern components?

I'd like to have something that actually worked behind a KVM switch.

Also these days there's not just the matter of connecting a PS2 keyboard to an Indigo. I'd like to connect an Octane2 an an Onyx2 to a modern (USB + DVI) KVM switch. DVI KVM switches with analog switching capability are easy to find, but PS2 + USB support and DVI with resolutions > 1600x1200 seems impossible to find.

What would be truly awesome would be some sort of module design where you can connect a modern (USB) keyboard on one end, and it emulates an SGI "4D" or PS/2 interface on the other. It could probably be done using an Arduino or Teensy. Cost isn't even my motivation, I just want something that works, and works reliably, with a KVM so I can eliminate the zoo of keyboards, mice and glass tubes at my place.

I was sort of expecting to tackle this myself at some point, but it will be a while ...

pentium wrote: I couldn't care about USB support. PS/2 ain't gonna go out of style any time soon.

Consider this situation. I want to cluster these systems on a single keyboard + mouse and two screens:

* An Onyx2 (PS/2 + analogue)
* An Octane2 with V12 + DCD (PS/2 + 2x DVI)
* A Tezro (USB or PS/2 + 2x DVI)
* A Hackintosh PC (USB + 2x DVI)

As you can see, I've got analogue and digital screen outputs, and USB and PS/2 keyboards and mice. I don't want to dumb down the displays to analogue for the sake of that Onyx2.

This restricts me to a DVI based dual head KVM switch (DVI-I to be precise -- it must pass the analogue signal of the Onyx2).

Next issue is practical: I've never seen a KVM with DVI and PS/2 ports which supports resolutions > 1600x1200. I guess nobody bothered to make the chipset for that? And I'm pretty sure that was not a dual head either.

Dual head DVI + USB KVMs on the other hand are easy to find, I've been using them for years at work. And unlike the older VGA + PS/2 KVM's they actually *work*. Not work "most of the time, except ...", or " works except this system XYZ has issues with keyboard repeat rate", or "works except this system ABC (sometimes) needs to be rebooted after I switch inputs...", if you've ever used a PS/2 KVM you know the drill.

So that's why PS/2 is out too. I will happily sacrifice the capability to support more than a maximum of six simultaneously depressed keys plus the modifier keys (shift, alt etc..) in exchange for that. (that is apparently the N-key rollover limit for USB, even though apparently many keyboards have a problem with more than 3 keys due to the way their diode matrix is wired). In either case, this never bothered me. Unpredictable KVM behavior on the other hand can be immensely frustrating.

So this process of elimination leaves me with a DVI + USB based KVM, plus the need for two USB-PS/2 converters. These exist btw, although PS/2 emulator is the better term, but cost ~ $100.

The alternative is to buy monitors with 3 digital and 1 analogue input and switch video inputs at the monitors, and use an old PS/2 KVM to switch only the keyboard + mouse. This hack would be rather unpractical in day-to-day use.


On the other end of the room, I have a cluster of older systems, basically ranging from Professional IRIS to Octane1 MXE. For those I can settle for an analogue, PS/2 based KVM (and live with the occasional weirdness). Except, back on topic, the existing Indigo keyboard/mouse converter doesn't work with a KVM. I've spent hundreds of euros and tried many versions, including ones supposedly made for SGI & SUN, and it won't work. What works (mostly) is powering up e.g. an Indigo with a real keyboard and mouse, and then hotplugging the KVM into the loop. Sorry, but that's BS. Had this worked I would have bought a dozen of these things years ago and lived happily ever after.

Normally you'd expect "P0 M1 C", where:
P0 = partition (or processor?) 0
M1 = module 1
C = has the console

P255 is way off. I'd advise to hook up the serial console cable to the MSC port (bottom at the back of the chassis), and jumper the MSC to stop in POD mode. Then re-initialize the NUMA configuration (search the archives for the infamous "clearallogs ; initallogs" sequence). Reset the PROM environment for good measure once you're able to reach that. Then jumper the MSC for extra thorough diagnostics at boot to get some confidence in the system.

You're not the first one: viewtopic.php?f=3&t=3654

In that system apparently the XBow fan was stuck, so that could easily be the culprit. You'd need another system to mix and match parts to know for sure.

I'm servicing a Saeco Royal Professional super-automatic espresso machine:

My parents in law introduced me to these, I serviced that one a couple of times (it has > 56000 coffees and counting ...), so I bought them a new one, bought one for us, one for work, one for my father in law's yacht club and now this one which I got for €50 because it had a "short circuit" -- a burned out motor for the brew group transport (cost me €12 to replace). I'll keep this one as a spare if one of the other Saecos develops a problem. These things are (semi) professional units, quite reliable and parts are readily available and cheap. Unfortunately, the engineering is typical a Italian job:

I mean, get real. Why not bundle wires and keep them away from the hot boilers? Half of those wires carry 230VAC, btw.

This particular unit is from 2001, so I think I'll do a complete rebuild, replace the boiler gaskets all other O-rings. Costs only a couple of bucks and then it's good as new

zkzkz wrote: Hm. I will happen to be in Vantaa in a few weeks (and have a free checked bag allowance). I might be interested in some/all of these though I'm not sure what.

An Onyx2 rack is roughly 100x80x185cm . People have transported them as-is in vans or station wagons, but normally you transport it in a crate and then it's ~ 120x120x200 cm, weighing some 300-400kg. If you decide to take that to the airport and check it in, make sure to bring someone to film it because I'd like to see the look on the face of the airport personnel when you show up at the desk

The smallest unit of an Onyx2 rack one could still consider a computer is the compute module. That would give you a Origin 2000 deskside stripped of it's skins and wheelbase. Still circa 70x70x70cm and some 75kg (my estimate, non-trivial to lift off the floor by one person)

There are two SCSI cables inside the Fuel: one to the harddisks, one to the CD/DVD-ROM. The latter is a traditional 50-pin cable, so narrow SCSI. IIRC, narrow SCSI only allowed 8 devices on a bus, numbered 0 ... 7. So if you jumpered your device to ID 8, I can imagine there's a problem.

dexter1 wrote: You need a recent IRIX installation CD overlay containing the diagnostic tools.

If the system has IRIX installed, PROM will boot the diagnostics from the hard disk. It will try to load diagnostics from a CD first, but failing that it will start them from the system disc.

NB: since we're talking about an Octane2 here: the standard diagnostics do not test VPRO graphics boards.

Next up: Exchange on Linux

Playing Nickelback in reverse to verify this:

So far we've only confirmed the 2nd part of the claim

NB: Yeah, that's not the real Dave. Still funny.

astouffer wrote: Any difference in speed when using this compared to an "old fashioned" drive?

Performance

Any SCSI disk made in the last 20 years should be able to beat that, except maybe the 512byte block reads.

Moving the stereo equipment in my computer room. Unfortunately that means new (longer) speaker wires. I was happy with what I have in the living room so got some more of that. Then realized that pulling a ca 15mm cable through a 5/8" pipe in the wall is ... a challenge. Much swearing with the help of a can of silicone spay the cables are in:

My Dynaudio speakers are insensitive (86dB), 4Ohm. 200W amp recommended. My Rotel RB991 does that into 8Ohm, and twice that into 4Ohm. Evidently, the word "overkill" does not exist in my vocabulary

Ugh. Is that a Lyrec TR532? There's a service manual here: http://analogrules.com/manuals/strangerecorders1.html

Archive password is 'AnalogRules.com'

ivelegacy wrote: ups, I had forgotten to ask … tips about SCSI/Scanner (A4 size) compatible with SGI ?
and in case, which Irix App do you suggest ?

The 'app' is Impressario, and it comes with IRIX. The release notes or manual will tell you which scanners it supports, but it does support certain older HP SCSI scanners: the HP ScanJet IIc/IIcx/3c/4c/4p/5p. I have one of those. It's been a while, but it did work.

Not the one you're talking about I think, but I made a poster of this one:

That's the "single board" graphics card, right? I have one of those, but it's in the attic and there's a desk blocking the path to it ...

But if nobody comes up with a photo I'll go digging.

dexter1 wrote: Edit: found it, it was a 4D/480. There is a thesis about the chinook program here: http://citeseerx.ist.psu.edu/viewdoc/do ... 1&type=pdf

I want it, so I can put my 4D/380 to work

vishnu wrote: I'm seriously tempted but hamei's horror stories about how bad his O350 treated him are a significant cautionary tale, do you think he just got unlucky with a lemon? Because on paper anyway thousands of Origins have been running 24/7 in sgi supercomputers for years, and you don't hear any tales of woe coming from them. Unless they're just like google and the second one of them shows a glitch it gets yanked and tossed on the scrap heap...

I bought my O350 more than 5 years ago and it's been running 24/7 since the day I got it. So far, the only issues I've had are:

1. One of the PSUs failed. This turned out to be a design flaw (search the archives). I repaired it and it didn't fail since.

2. The system disk failed. I replaced the disk and recovered from backups.

3. One of the CPUs will throw a correctable L2 cache error every once in a while, especially on a hot summer day. It's been years and it doesn't seem to get worse. If I didn't run this thing 24/7 I might never have noticed. I have a spare CPU board in case it decides to fail entirely but I can't be bothered to take it offline to replace it pre-emptively.

That's it. The NVRAM batteries are probably good as new because the system is always running

I had something like that with a 4D/35 years ago, and believe it or not, someone had the actual chip (HPC?) for sale on eBay so for a couple of bucks the thing was back up and running. I think the seller was 'rborsuk' and I think he's still around.

MrWeedster wrote: The converters that were listed on ebay are long gone.

They're back!

computron wrote:

Code: Select all

Error-- keyboard not responding

Error-- cannot open console "gfx(0)"

initializing tod clock
setting secs=0 min=0 hour=0 day=1 month=1 year=0


System Maintenance Menu

1) Start System
2) Install System Software
3) Run Diagnostics
4) Recover System
5) Enter Command Monitor

Option?

i think i can suspect the keyboard......

Who have a keyboard for me ?

it hang when i ask a hinv via term...... strange ?

Eve

Seems your NVRAM battery ran out -- this is possible since this systems is very old and you didn't use it for a while? I don't know about IP6 by heart, but IP12 (the 4D/35) uses regular CR2032 coin cells.

If the battery runs out, the boot settings are lost and it's possible it defaults to using the serial console instead of the keyboard/mouse/monitor.

You have PROM option [3]: "Run diagnostics". If IRIX is installed on the hard disk, the diagnostics are installed also. Why not try to run them to see what it gives?

If you are certain it's the keyboard that's broken, I have some 4D series keyboards in storage. But they are for the PowerSeries so they have a DB15 connector. I believe the 4D/25 uses a DB9 or DB15 connector? You'd have to replace that connector, otherwise it should work.

These keyboards are quite heavy, so shipping is not for free.

Looks like Microsoft hasn't nailed AI quite yet .

ivelegacy wrote: he said 240V, 30A

Probably the breaker on the PDU, not the actual power consumption.

My 4D/380 consumes ~ 6.5A (1.5kW), but:

You guys got it all wrong. That's obviously $50 for the Indy and $1050 for the EXTREMELY RARE SGI Indy Bike Shirt

mopar5150 wrote: You reminded me of this price sheet. viewtopic.php?f=3&t=16728414&p=7367158&hilit=sgi_prices.pdf#p7367158

My favorite: for half a million bucks . Fortunately, you can save a hundred grand by ordering SC5-EQTRANS-A3KL-1.3:

Yeah, that must have been a big seller. I wonder what that 'E' is for

mopar5150 wrote: I have gone through many Tezros and origin350s, I have never seen a single CPU version.

Me neither, but they did exist.

They have less available PCI slots , and you cannot upgrade the single CPU version to dual or quad CPU .

Hope the buyer realized that...

mopar5150 wrote: You could swap out the main board as that should be the only difference.

Except it isn't, because the interface board (with the PCI slots) is different also:

Tezro Owner's Guide (CH2, Interface Board) wrote: Note:
Figure 2-5 depicts the interface board in a 2- or 4-processor workstation. The
interface board in a 1-processor workstation has 4 PCI-X slots (3 available) and 1 XIO slot.

So the 2-4 CPU Tezro has 8 PCI slots, and 2 XIO slots. The single CPU version has half of everything. And one PCI-X slot is unavailable, probably because it's taken by the IO9. That only makes sense if you eliminate one of the two XIO-to-PCI-X bridges. Maybe cost saving, maybe a necessity. If you look carefully, the dual and quad CPU IP53 boards have *two* high density connectors near the front which connects to them to the interface board. A Fuel has one (and 4 PCI slots and one XIO slot!).

So either:

1. The single CPU IP53 nodeboard doesn't have what it takes to talk to dual XIO-PCI bridges. They produced a reduced cost interface board with less components.

2. The single CPU IP53 nodeboard doesn't have what it takes to talk to dual XIO-PCI bridges but it wasn't worth redesigning the interface board. It's all there, but only half of it works.

I'd like to see some pictures and a hinv from this system. But I wouldn't want to sink my own money into it, nor would I assume upgrading it to a quad config is simply a matter of replacing the nodeboard.

EDIT: OK, Diegel beat me And indeed the INT53 board is different:

So, are the PCI slots populated at all?

Nice!