IBM RS/6000 320/32H specific notes

PrevUp

The 320/32H are of particular interest to me, since they were the first RS/6000s which appeared at my workplace, GSI. Introduced by IBM in early 1990 they were available at GSI soon after as playfields to sneek into UNIX. Until then, we had only "proprietary" systems such as MVS on S/370 mainframes and VAX/VMS.
Those early 320s looked more like PCs, in fact they shared the Microchannel Architecture. They came with only 8MB or 16MB of RAM, which was definitely not enough. Especially linking large (for the time) programs easily could lead to turn around times of about half an hour. But equipped with enough RAM (32MB was about right and not too expensive) they were real racing horses which left all of the competition in the dust for quite some time.
With the usual L&F (==educational) discounts one could purchase a 32H with 1/2 the raw horsepower of an IBM mainframe's CPU, but at only 1/20 of the price. Its price/performance ratio also beat that of the then popular VAXstations by a factor of three or more.
Riding this wave, in 1991 a 32H became my first RS/6000, my first workstation and my first personal UNIX box altogether. Unfortunately it got scrapped later on when it became obsolete (in fact I wasn't too much interested in keeping old stuff back then).
So my current 320 and 32H machines are a reminiscence of those old days. There's almost no difference, except they are clocked differently, at 20 and 25 MHz, respectively. The 320 came rather incomplete (no disk, no network, no internal SCSI cable, and most unfortunate: no key), so I'll concentrate on the 32H. It had no disks either, but at least the necessary internal cable. And it still had the key, which is necessary not only for opening the case but also for installing the AIX OS.

Pixx

front view rear view Unfortunately the case has a little damage at the rear side.

empty view View inside with CPU and RAM removed.

inside view View inside, working condition.

CPU

CPU card The original POWER CPU was not a singÎe chip, but comprised a collection of functional units on one card. Nevertheless it was by far the fastest workstation CPU when it came out.

RAM

RAM card This is the RAM card: S1, FRU=59F4433, (53F3089/91 if it comes with 8/16MB installed), P/N=41F0521 (?). holding 8 single RAM modules (ECC, not just parity). Originally it had 2MB modules (Samsung KMM540512BG-8, 512Kx40-8P, P/N 68x6271). The S1 maxes out at 32MB, for which 4MB modules are needed, for example THM401010ASG-80 68x6356. I also tried IBM P/N 68x6357 FRU 70F9976, which have 8MB. They work, but only 4MB are recognized.
Eventually I could spot enough of the 68x6356 modules to max out the S1 board, and even more, a third party board, the Cambex P/N 360-028-700 A gave me a whopping extra 64MB.

SCSI card

SCSI comes on a dedicated MCA board. It has an extra fuse (12.5V/1A) which was missing in one of the machines, so the board appears to be dead and I had to swap it. Unfortunately both, the internal and the external connectors are IBM proprietary, but this wasn't so unusual for the times. For the internal connection a special cable is needed (see below). The external connector needs an adapter cable.

SCSI disk

Disk holder The disk holder with the special internal cable.
The proprietary end connects to the SCSI card, the other one is standard 50-pin, and any 50-pin SCSI2 disk should work. In fact the cable has two 50-pin outlets, corresponding to the two disks one can stuff into the holder. The cable has its own terminator, which makes it a bit difficult to plug this end into a flat disk's connectors.

Key Mode Secrets

Diagnostic sequence for an 888 code

Press reset once.
    Is "102" displayed on LED?
       YES --> Software prob. press reset once for crash code,
               then again for dump code. Reboot. END
       NO  --> Continue.
    Is "103" or "105" displayed on LED?
       YES --> Hardware prob. press reset 2x for SRN code
               then again for FRU code
               press reset 8x for location code. Reboot. END 

Last updated: 30-Dec-2011, M.Kraemer