Some time ago I bought an old microchannel RS/6000 39H as a hobby toy. I mostly deal with newer Sun/Oracle equipment so I'm a bit out of my element.
The machine was running, AIX 4.1.3 I believe. Using it via IBM terminal. I stored it for what was apparently too long, and now it stalls out during built in testing.
I'm wondering if any IBM guys can help me solve this one.
Anyway, it dies out with flashing eights and I read out the error codes as 888-103-125-911-ccc-104-205-c09-888. I checked the old microchannel service documentation and I'm either not smart enough or am missing something as the codes didn't seem to translate correctly for me. I'm stuck.
I did all the obvious, reseated the boards, reseated the simms. checked the various power connections. My best guess is that it is some sort of memory issue.
If anybody has any insight I'd greatly appreciate it.
The important thing is what the LED displays when you turn it on: and what it displays during the boot process, without that information we have no clues on where your system is blocked...e.g. did you see displaying values 100-199 at the beginning? then what were the next values...
N.B
A server not running for a (very ) long time may not start because the boot disk fails to start... Ivw seen that a few times, and trying to get it started can be by tapping on it, using a hair dryer to heat the grease etc... there are some chances of success...
So listen carefully when you turn the key id you hear the boot disk start...
From what I remember LEDS flashing 888 is not good sign... Normally you would press on the reset button and by what you wrote in post #1 it displayed 103 which means possible software but also ( and more probable...) hardware issue (e.g. soft could be corrupted bootstrap...)
Now I remember hter were here at this stage action where you would press again the reset button like pressing twice, but I have all forgotten...
What I remember is that you should be able to get more info by this means, and 2 cases should appear:
1) its Soft and you have generated a Number for IBM : Service Request Number
2) you got a FRU (Field Replaceable Unit ) which would give you the faulty device...
You are stuck in the BIST...So its HW... I would strip down to the motherboard and give a serious clean, and check each plugin after... Could be cards or cables...
That said I have no idea (cant remember ..) what a 39H is... I am no IBM person, but did put my hands on Bull escala series and I remember of an IBM 43P (which was no more MCA...)
I have taken every card out, cleaned them as best I can using compressed air. Tempted to try some defluxing spray. I'm not totally sure which cards are critical to system operation. There is the POWER2 CPU and RAM board (I'm assuming I need that) a power supply board, and there is an additional RAM board that is fully loaded. I have tried with and without that additional RAM board.
Maybe the AIX guys will know more about these. I was tempted to post there but wanted follow the rules. I was pretty sure it was hardware from the start.
Sorry for not answering earlier, i simply missed this thread.
First: this type of hardware rarely goes broken. At the time this baby was built, IBM used only stainless steel and reinforced concrete. Nothing short of an atomic strike will be able to break these babies once you got them going. ;-))
Second: You need an "ioplanar" a "sysplanar" and maybe a SCSI connection to attach the disk(s). I am not sure if the 39H had a SCSI adapter integrated on the ioplanar or not. This should be enough to connect via a serial terminal (set it to "9600,8,N,1", unlike newer system which use 19200 baud, use either a "wyse60" emulation or a "IBM3151[-III]").
The flashing "888" is a very bad sign. To help you search further, it is called "three-digit-code" in IBM-lingo. Subcode 103 is, if memory serves correctly, something about a dump, AFAIR, and "125-911" will be subcodes for this 103-condition (if it is really about dumping probably the reason for the dump).
//edit: 103 is hardware, according to this. This documentation is about another MicroChannel system but should be the identical to yours in this regard.
I will see if i can retrieve further information from the heap in my basement (:-)) ). In the meantime this link might help you, maybe this link too.
Well, after digging through the IBM documentation for an hour or two and trying a diagnostic procedure of taking ALL the ram off the CPU board that i wouldn't have guessed of trying, it seems the CPU board itself is the source of my issue. I am fairly confident that thats what it is at least.
The good news is that if I can't fix it and there isn't a lot on the board I think i can fix short of the clock crystals I can source a new one for $120.
A guy on ebay wants $1300 for one of these 39H systems. I doubt it would sell for anywhere near that price... but I paid $30 for it and I feel okay putting $120 into it.
I read on some accounts that a failing battery might cause similar (if not the same) effects, so i suggest you start by replacing the battery first. This is probably the least expensive part of it.
If you do not absolutely need a 39H for some reason but would be satisfied with any RS/6000 you might consider buying one of the many 43Ps still flying around. It won't run any OS newer than 5.1 (Model 130/230) or 5.3 (Mod. 140/240) but the 39H will not run anything newer than 4.3.3 either (if it even does this, there might be microcode issues preventing even 4.3.3).
You should get even a well-equipped 43P for under $500,-.
I did replace the battery, I wired up a AA holder to the pin header plug. I think it was a 4.5 volt battery if memory serves, so 3 AA cells. Was one of the first things I tried but sadly didn't make a difference. The old battery was very dead. Floated at around 2 volts and would slowly drop on the meter.
I definitely don't need the 39H and probably will end up picking up a newer system. I'd still like to fix it though. I'm a bit of a collector and its not generic PC hardware after all. I think its worth keeping running.