The collected works of jan-jaap - Page 18

I had two Indigo R3000 boards run out of battery (the infamous 'can't set TOD clock' bug) in a matter of days.

The original batteries (Tadiran, Sonnenschein) are pretty expensive so I decided a CR2032 hack was appropriate. Here's the result:


Chyron Centaur Video option board mounted again. As you can see, clearance is not a problem.


Oh, and they work fine again of course. The CR2032 a somewhat lower voltage (3V versus 3.6V) but considering the price of CR2032 cells I don't really care if they last a little shorter

thegoldbug wrote: How are you attaching the battery holder to that IC?

Two component glue. It's an SMD style battery holder so all it takes is a surface without circuit traces or via's. There are several via's in the PCB area around the original battery location so I opted for the nearest IC large enough to glue it to.

twix wrote: Nice. My soldering skills are totally inadequate, so I went out and bought the Tadiran (ouch!) for my Indigo.
BTW, you don't happen to have a spare keyboard for the Indigo?

Last time I looked (and that was several years ago), cheapest 'genuine' was the Sonnenschein which cost me ~ 14EUR I think + shipping from Germany. Now I don't have to wait two weeks for a battery to arrive, I can simply got to the 'Action' discounter during lunch break and buy five CR2032 for one EUR, or close

It's possible I have a keyboard, I'll have to look.

And I've got sound! Yay! Finally

I improvised with a pair of SGI 'Polk' speakers until now but there's no substitute for a decent pair of speakers. They're powered by a Rotel amplifier and a Cambridge Audio DacMagic Plus.



If you look carefully you can see one of the few mistakes I made during construction: the holes in the wall for the speaker cables are too high...

BetXen wrote:

Code: Select all

ALERT: Error reading monitor ODYSSEY interrupt status 1: no acknowledge
ALERT: Error reading monitor ODYSSEY interrupt status 1: no acknowledge
ALERT: Error initializing the ODYSSEY monitor, no acknowledge
ALERT: Error reading monitor ODYSSEY interrupt status 1: no acknowledge
ALERT: Error configuring ODYSSEY power (XIO 12V BIAS) monitoring: no acknowledge
ALERT: Error configuring ODYSSEY power (XIO 5V) monitoring: no acknowledge
ALERT: Error configuring ODYSSEY power (XIO 2.5V) monitoring: no acknowledge
ALERT: Error configuring ODYSSEY power (XIO 3.3V AUX) monitoring: no acknowledge
ALERT: Error configuring ODYSSEY temperature monitoring: no acknowledge
...
ERROR: I2C:no acknowledge


ODYSSEY is the internal name for VPro graphics. There's only one XIO slot in a Fuel and it's the graphics slot.

It appears to me that the L1 has troubles communicating (via I2C) to the 'monitor' chip on the V12 which monitors various voltages and a temperature reading. It makes sense that this monitor chip is something similar to the notorious environmental monitoring chips used on the Fuel main board.

I'd take out the V12 and inspect it for DS1780 or other Dallas chips. If you can buy / borrow another card that would make it easier to confirm the guilty piece as well.

FWIW: I've got 'no acknowledge' on I2C reads to the CPU of my Fuel . But there's apparently no environmental monitoring going on there so the damage is limited to the CPU serial number disappearing from a 'hinv -mv'

sgifanatic wrote: It would be wonderful if you could share what it looks like now... in the last pics I saw, things were still in transition.

It's still not finished ... the realities of life with a family with two small kids and a full time job, I'm afraid. Weekends are always incredibly busy, so realistically I only spend time on my hobby in the evenings during the working week.

I still need to install the network cables plus the Fibe Channel / FDDI fibres on one side. For quite a while the baby room furniture was stored in my computer room until I had the time to fix it up and sell it. This was massively in the way of everything.

I hope to finish the wiring this winter. Once the wiring is in, I can fill up the large cupboards with spares etc. That should clean up nicely.

Next year I hope to design a comprehensive KVM solution for my workstation rack and get rid of most of the CRT monitors. That will clean up nicely too.

Less is more. How zen

And I need to put some decoration on the walls. I have some failed old Geometry Engine boards of various generations, maybe I can frame some and turn them into art or something.

pentium wrote: I keep thinking that my workshop was an impressive feat but then I remembered you practically built an addition to your home just for your computers.
Now that it has been a while, has there been any issues with ventilation?

It gets warm there during the summer, but not a sauna. I have not installed an airco (yet). When it gets that hot it's a better idea anyway to sit in the garden with a cold beer

bigD wrote: The cost doesn't matter if the quality justifies it.

Lian-Li then.

nongrato wrote: or send a video to an external device:

That cat

I would try that, but the cats are not allowed in my computer room

I've got some on my FTP also: ftp://ftp.jurassic.nl/pub/irix/gnu/irix-4.0/

This is all 10 years old, but I think I first compiled all the packages GCC depends upon using the native tools. This often means rather crude hacks and/or using very old versions of these GNU packages.

Then bootstrap GCC, and finally rebuild everything with the GCC just built. Including GCC itself.

Code generation of GCC v3 wasn't all that terrible compared to the MIPS UCODE compiler. But GCC takes a lot more CPU cycles to do the compilation which can be a problem when all you've got is a 33MHz R3000. I think I used a 200MHz R4400 Indigo2 running a late version of IRIX 4.0.5.

At the very least IRIX 4.0 has a more or less complete ANSI C library. Not any of the fancy POSIX stuff, but still. You should try compiling software for IRIX 3.3: pre-ANSI library, K&R compilers. Now there's a challenge

Start with re-seating the TRAMs. Saved my day twice already (with two different MXE's).

Don't tear things apart unless you have a good reason. You just risk damaging it.

What if you run it without the TRAMs? You can, you know. It'll be an SSI and not an MXI, but if it gives a picture you know it's the TRAMs causing the problem, and not the graphics board itself.

You have to remove both (or none)

Maybe. Maybe not.

That's a userspace implementation of pthreads, using setjmp/longjmp. It's a cool effort, but it's like multitasking in Windows 3.1. A proper threads implementation requires kernel support.

Also, IIRC, some things like per-thread signal masking and handling are not possible (I think, it's been a while).

IRIX has had it's own sproc() based threads since IRIX 3, but again this uses different concepts, and you can't mix it with IRIX pthreads in a single binary, nor can you implement a 100% compliant pthreads implementation using sproc primitives.

hamei wrote: The calculation of what burning fossil fuels is doing to the atmosphere could be done on an Indigo.

Close. What you need is a 4D/380VGX Power Series. My 4D/380VGX, featured in this article when it was brand new:

It still has the original name (`dali') and Fraunhofer asset tags, and it still runs fine. Not too often though, because requires burning lots of fossil fuel

jan-jaap wrote: And I've got sound! [...] They're powered by a Rotel amplifier and a Cambridge Audio DacMagic Plus.

So it turned out that my Dynaudio speakers were a bit too much for my poor Rotel RA-970BX integrated amp. I've had the same problem before with the stereo system I have in my office at work and that was solved with a bigger amp (also Rotel, a pre + power amp combo).

Having good experience with big Rotel amps I got myself an RB-991 power amp. The Dynaudios, being 4Ohm speakers and relatively insensitive ( ~86dB) can handle (and need) a lot of power. Recommendation is 200W I think. The RB-991 puts out nearly twice that and that's continuous output power at low distortion, not some fancy unreal impulse power number. It also has a damping factor of 1000 over the entire audio spectrum -- this thing can drive difficult speakers.

The difference is profound Before, the sound was muddy, with booming bass. Now it's much tighter, with better definition. Amazingly, it sounds like you're in a bigger room, yet full of music. I haven't tied yet, but I'm quite sure this setup can also play so loud it will make an Onyx in 'fastfan' mode completely unnoticeable (and probably make my ears bleed ...)

recondas wrote: Very nice! With IR4 graphics/2 G B of texture memory and eight 800MHz IP35 processors in a single rack it [...]

There's two GB installed (1GB per RM11), but in an IR system this does not result in 2GB of usable texture memory.

Framebuffer memory adds up, and the resulting bigger total amount means more features, or higher bit depth at a given resolution. An RM11 has 2.5GB of frame buffer memory (up from 80MB in RM10), but IIRC the 300MB/s fill rate limit still applies so I'm not sure what's the purpose of this -- not higher display resolutions I think.
Texture memory does not add up, it is used in parallel. So with multiple RMs you get higher fill rates.

Still pretty awesome. Congratulations on an impressive score!

recondas wrote: [...]would easily qualify as my idea of the ultimate Silicon Graphics system for residential use.

only if you own a pair of noise canceling headphones.

recondas wrote: If you have the OEM cables that from from the "TMDS Video Output" ports on the DG5/TVO (in the G-Brick) to the matching ports on the VBoB, and those cables happen to be labeled with an SGI part number, could you post those part numbers?

As a side note: I've used generic high quality dual link DVI cables to wire my Onyx2 TVO option to a DMedia breakout w./ DM5 in the past. I'm pretty sure the DM5 showed up in the Onyx, but can't remember if I ever attached a HD-SDI monitor to look at the output signal. These are the same cables and VBOB I use with the Tezro normally. I think you have to flash something if you move a VBOB between an IR and a VPro system.

Start with the SGI Video BreakOut Box User's Guide: http://techpubs.sgi.com/library/manuals ... index.html

The vbob software is included in the DMedia DM2/DM3 software distribution. If you don't have it I think it's on Ian Mapleson's 'depot' resource and/or Nekochan FTP. It may even still be on SupportFolio

Keep in mind that the whole DMedia package can do two different things: video I/O to decks, cameras, (HD)SDI video monitors etc. using the DM2/DM3 + vbob, and GVO (graphics-to-video). GVO is optional and requires a DM5 in the vbob. You need DVI to connect to the DM5 so an Octane V12 needs a DCD. The vbob put the corresponding (HD)SDI signal on the BNCs, and you can attach the (graphics) monitor to the DM5 output.

You can telnet the vbob but I think it's only useful to run diagnostics on the thing or maybe recover from a bad flash. I normally do not have a serial cable connected to the vbob. You cannot telnet from the PROM.

foetz wrote: they had stdbool.h in 1992

Nope.

But I have a copy of the old info-iris archive on my FTP server: ftp://ftp.jurassic.nl/pub/info-iris/twilight.c

Dated 1992

Nice one!

I've got one of these DeveloperToolbox binders too, the theme is the adventures of two magical Indian characters I think?

FWIW: MIPSpro licenses are not versioned, they will work up to the last version. You need the installation media, but the license will won't be the problem.

I've got FTTH, so I've got an actual glass fibre coming into the house. It terminates in a 'modem' with several ports on it for radio/TV, telephone, IPTV, and one is ethernet. It's 'raw', it has a non-private IP, and I can/must attach the router of my choice. In my case that's an aging EnGenius ESR9850 and it can keep up with it.

I currently have 100/100 Mbit internet service which is enough for me. I think I can get anything up to 500/500 or even gigabit.

I'm in the process of replacing the EnGenius with a small PC (a Jetway JBC372 ), because it has more capabilities and I prefer to be in control of something as essential as my router, but in essence the EnGenius is/was fast enough. Consumer routers will not be able to do high speed VPN routing and other CPU intensive things, but simple routing is no issue.

And then there was IPI-2 which was positioned at the high end. My 1990-vintage 4D/380VGX PowerSeries has two SCSI channels (IO3 board), used for QIC150, CDROM etc, and a Xylogics SV7890 VME IPI-2 controller. In the bottom of the rack sits an 8" Seagate ST81154K (a.k.a. CDC 'Sabre'), which basically eats up half a 19" shelf, consumes 160W (it has it's own 220VAC PSU) and weighs roughly 30kg. It has it's own key panel and LCD screen at the front which you can use to manually key in bad sectors, I believe. Oh, the raw capacity is ~ 1.1GB. It's a bit of a monster. And it still works

IPI-2 has more bandwidth than the synchronous SCSI-1 implementation of the PowerSeries. The ST81154K internally works like a RAID0 (stripe) so it's streaming speeds are faster than you might expect. The PowerSeries PROMs can boot from the Xylogics controller, and 'fx' can format it so you can use it as an IRIX system disk.

IPI-2 had some other unusual features, I think you could attach a disk to two systems for fail-over, similar to Fibre Channel. But that pretty much sums up what I know about IPI-2. Information about it is pretty scarce.

The old router is still connected and answers the DHCP queries faster than the new one?

It's fairly normal that a DHCP client rewrites the resolv.conf, although I don't have much experience with DHCP on IRIX.

Reminds me of Temple OS .

I won't post the animated screenshots here because they're a couple of MB each, but you've got to see this.

Between what I had and what I pulled out of Rusti's basement a couple of weeks ago, I had enough to put together an entire Onyx2

This will be my test rig to test all of my (many more!) Onyx2 parts. I didn't want to use my 'normal' Onyx2 for that because I fear something would inevitably break. This is by no means the maximum config I can put together, it's just how it looks today. A working baseline.



Graphics: IR3, with dual RM10 and DG5-2/GVO:

It has no skins and sits on blocks because I've got no wheelbase either. Ah well


PS: to the right the project I was working on before I got side tracked by Rusti: my 4D/440 and it's broken power supply (which I'm trying to figure out and fix). I have a working PSU for it still, but I'm determined to figure out how these Power One PSU's tick before the last one dies on me... My 4D/70 is awaiting the same treatment.

PPS: The IR graphics are fine. It's the LCD screen which spent too much time in the garage and has clusters of dead pixels.

That's not (just) a logo, that's an honest-to-god 4D/380 VGX Power Series rack. And with a bit of imagination you'll be able to recognize the 42U HP 19" rack behind the monitor

The point of the whole exercise: stuff accumulates over the years. I keep finding Onyx2 parts everywhere whenever I open boxes I have no idea what's inside. I can't bring myself to use my "collection" Onyx2 to test them, I don't want to sell them untested, and I don't want to throw them away either. Something has to be done.

So, I'll test everything, throw away all the broken bits keep a sane amount of spares for my Onyx2 and probably sell off everything else.

For starters: I have a couple of GB worth of 512MB kits for Onyx2/O2000/O200. And enough 256MB kits to fully populate at least 2 nodeboards. Tested good, and in ESD safe packaging. Ready to go

Pontus wrote: The machine sold for 4$

Someone put a LOLbid of 1$/rack and now his wife is shouting at him

Alver wrote: I guess I could even run a proxy on it (but I don't see a need for it).

Privoxy on the router is on my TODO-list. To suppress YouTube ads on the TV

I have more than I will ever need, so I might as well sell a couple:

QTY 3: 030-1305-001, DIVO
QTY 1: 030-1046-002, DIVO

The 030-1046-002 board has extra connectors to install two DVC compressor option boards (which are not included).

All of them have successfully passed /usr/diags/DIVO/bin/divotest.

Price: 30EUR each + shipping from The Netherlands.

isfinite() is C99 and MIPSpro doesn't allow C99 features in C++.

The underlying primitives:

are in libm (I think, could be libc).

You just need the isfinite() macro, then:


(not tested, but you'll figure it out)

Byrd wrote: - The teal Indigo2 sounds like an oddity - from what I can gather the R8K CPU only came out in later models?

The R8000 Power Indigo2 is a very collectable system (IMHO). The R8000 is unique: it is not a CPU, but a chipset . It was the first implementation of the 64bit MIPS IV ISA. It was designed to excel at number crunching, integer performance was basically the same as an R4x00 CPU. The tagline was 'a Cray under your desk' or something.

Because the CPU card was built from several chips and because of the 'streaming' cache memory used, it was very expensive. It was replaced withing a year or two with the R10000 CPU series, which offered similar FP performance to the R8000, but much better integer performance so it was a better all-round performer. R10000 was also easier to scale to higher clock rates.

Compared to R4x00 and R10K Indigo2s, the R8000 is rare. It can also be rather fragile. Enjoy, and take good care of her.

DIVO doesn't need a breakout. It's SDI in and out so you need professional equipment or converters.

chicaneuk wrote: Stupid question but what is it that's missing in the shot of the rear, above the power supply?

That's where the optional PCI cage slides in. Normally it's covered.

Check the elcos in that linear power supply. Reminds of my power amp

chicaneuk wrote: Another scrapper from work, this was absolutely barebones with a 100MHz CPU, 32MB RAM, 8 bit graphics and some extremely loud old drives measured in MB not GB

Yup, had one exactly like that in my first real job ... did real work with it too.

chicaneuk wrote: I guess in it's defence, IRIX was a bit less bloated back then so probably booted and ran a bit quicker!

True, but it came with IRIX 5.1.x which leaked memory like nothing else so it had to be rebooted every couple of days.

With IRIX 5.3 it was perfectly usable, especially after we upgraded the memory with an extra 64MB. Of course, 'usable' was compared to a contemporary PC, usually a 486DX/2 running Windows 95 or OS/2 Warp.

vishnu wrote: Strictly speaking it's not the KDE toolkit it's the K Desktop Environment which is built with the qt toolkit from Troll Tech.

Troll Tech -> Nokia -> Digia, but yeah.

vishnu wrote: The consensus seems to be that qt is the best C++ based toolkit but, and at least this was the case last time I looked at it, it runs a preprocessor over your code which implements their signals and slots architecture so really when you're debugging, you're not debugging the code you wrote, you're debugging the code you wrote that the preprocessor modified, and, at least back then, there weren't any debuggers that could understand that. I'm sure it's all different now I haven't looked at it in a decade at least, and have pretty much stuck with Motif that whole time...

The signal / slot mechanism is actually pretty nifty. It's implemented via macros, but it's not like there's a special processor munging your code -- if your debugger cannot handle it it will have issues with everything in /usr/include as well.

Maybe you're talking about the user interface layout? The .ui files are processed into source code by a processor (moc). Then again, you don't want to debug the layout of your buttons, you want to debug the action taken after the user presses it.

Current versions (> 4.7) are not supported on IRIX.

Pontus wrote: At my previous workplace we had an old DELL running well over 600 days, no UPS.

In my experience, it's more likely to have a UPS break than a power failure.

Not a scientific sample size of course, but I can't remember the last power outage. Must be years ago.

recondas wrote: Eight years ago a number of us bought new 600MHz V12 Octane2s directly from SGI for $600 (no, I didn't leave out any zeros ).

$300 for the Octane, plus $300 for the IRIX media, IIRC. And they wouldn't sell them to Europeans because of ROHS regulations

robespierre wrote: This came across my desk today: https://diafygi.github.io/webrtc-ips/
Firefox and Chrome have been enhanced to support a new protocol called STUN (sounds like the taser, but it's an IETF design for opening connections through NATs, and therefore through firewalls). It has the side-effect of deanonymizing browsers that use VPNs, and exposing their local and public IP addresses to any website. So people who use VPN to view shows that are blocked in their country could soon see them stop working.

Fortunately this can be disabled: http://www.ghacks.net/2015/01/27/sites- ... ng-webrtc/

The sane default would have been to disable WebRTC -- this way it's yet another data point to create a unique browser fingerprint even with cookies disabled.

Alver wrote: The "PITSTOP.EXE" doesn't do much at all either - it launches, attempts to access the A: drive, then exits back to prompt. Guess it needs something on a floppy that I don't have.

I played that game in my first year in university That's probably the copy protection, it needs that floppy and if you wanted to make a copy you needed a special disk duplicator program.

It's probably a 8086. If you're really (un)lucky it might not be 100% IBM compatible.

Seen on fleebay: http://www.ebay.com/itm/271758384526

This thing has no less than 6 HIIPI serial cards installed. Why would you do that? I don't think you can use an O2K as a HIPPI router and one system in 6 different HIPPI networks seems a bit far fetched.