SGI Onyx 10000 RealityEngine²

I'm a sucker. I couldn't help myself, the price was right and it was only a 14 hour round trip (!) -

Putting the Onyx back together again was a bit of a chore. The process consisted of me first stripping down the machine and then moving it to its new location, my office, and reassembling it. By the way, even stripped the units are damned heavy, and they're a real bitch to haul up those stairs (below).
On the back of the truck:

After moving all the bits and pieces into the office, including a brief pictorial of the path the Onyx took into the office:

What the rest of my office looked like (it was tidied up shortly thereafter):

Boards in the machine - two of an unidentified VME board (anyone know what it is?), Raster Manager (RM5T) and it's side connector, two of the Geometry Engine (GE10) - yes, that's 12 i860s; two of the processor board (4x 195mHz R10000 MIPS); two of the memory board:

(I'm missing a picture of the DG2, the Display Generator, from the above lot which is part of the RE² GFX subsystem board set).

Reassembly of The Beast:

Subsequent re-disassembly to get to the CC3 midplane and have a decent look at the power supply wiring. What I'm working on here is how to convert the three-phase unit to single-phase.
Overview of the removed parts and panels - the power supply 'cage' is in front; Looking at the business end of the OLS cage, where the OLS' slip in (top cover removed); Looking down into the OLS cage showing wiring; The wiring going into the OLS cage from the mains; Looking into the lower front of the Onyx, where the OLS cage and SCSI boxes usually reside - what you see is the "front" of the CC3 midplane - The rectangular hole at the bottom is where the wiring to the board comes in from the OLS'; A closeup of the CC3 midplane, showing the CC3 Power Boards (512 and 3x 505), the ribbon cables that carry the chassis 48VDC supply to CC1/CC2 midplane and the pins where the OLS' connect to the midplane.

Three pictures of the TwistLock plug BEFORE I mangled it; Two pictures of my Frankenstein wiring for "converting" it to single phase and running on my stove outlet; The CC3 midplane behind the OLS' cage, my fan hack (see below).

The story: I picked up this Onyx at Boeing surplus in St Louis, Missouri. The unit came without CC2 Power Boards, a System Controller and Graphics/Main I/O breakout panels. Thanks to the internet I found a chap on who was hawking pieces of some scrapped Challenges, and got all of the above for a song.

I have made a walk-around tour of this system and posted the video:

Rackmount Onyxes ("Terminator") come with either two or three power supplies (Online Switchers, or OLS) depending on their configuration. A machine with the optional CC3 cardcage, or with an optional second SCSI Box - or both, like mine - comes with a third OLS, and is wired for three phase. I do not have three phase in my office where the unit stands, so I needed to 'convert' it to Single Phase.
I first researched the differences between the single and three phase wired Onyxes. The short answer is they, obviously, only have two OLS, the wiring is slightly different in the OLS' cage tray and in a CC3-less machine there is a large PCB inside a frame carrier, where the CC3 midplane usually is, that carries voltage from the OLS to the CC1/CC2 midplane. I purchased this PCB/carrier from the same chap on, but upon reflection and some good advice from the Classic Computing listserve ( I realised that the OLS' really couldn't care too much about the phase of the power input, and that if I simply rewired the plug to a 220v single-phase outlet it would probably just work.
And with that, I did. The above photos show the NEMA10-50R stove plug I purchased and replaced the TwistLock for, and my "Four into Three" wiring. This brings up two of the three OLS, which would be a problem if my CC3 was stacked, or my second SCSIBox filled with devices, but in my case there just isn't enough devices to worry. I had to remove the OLS that was not receiving AC - the Onyx "knows" it is there, and halts the boot process with a POWER FAIL.
After removing the OLS, much to my amazement, the unit powered on. The system made a good go of bootstrapping off its System Controller PROMs and starting until it discovered that the second fan ("Blower") was deceased, BLOWER B FAIL. Unfortunately I don't have a spare blower, but I've got a chap going to find one for me. Nevertheless, I had gotten this far, I wanted to see if the daffy machine would power on at all - or whether there was something else wrong with it.
To this end I made the above wiring hack, 'joining' the RPM lines of both blowers so the Onyx could read RPM coming off the tach line on Blower B. Sure enough, the Onyx came up to PROM!
The hinv reports:

              System:  IP25
           Processor:  194mHz R10000, 2M secondary cache
Primary I-cache size:  32 Kbytes
Primary D-cache size:  32 Kbytes
Secondary cache size:  2 MBytes
           Processor:  194mHz R10000, 2M secondary cache, (cpu 1)
Primary I-cache size:  32 Kbytes
Primary D-cache size:  32 Kbytes
Secondary cache size:  2 MBytes
           Processor:  194mHz R10000, 2M secondary cache, (cpu 2)
Primary I-cache size:  32 Kbytes
Primary D-cache size:  32 Kbytes
Secondary cache size:  2 MBytes
           Processor:  194mHz R10000, 2M secondary cache, (cpu 3)
Primary I-cache size:  32 Kbytes
Primary D-cache size:  32 Kbytes
Secondary cache size:  2 MBytes
         Memory size:  256 Mbytes
           SCSI Tape:  scsi(0)tape(2)
          SCSI CDROM:  scsi(0)cdrom(0)
           SCSI Disk:  scsi(1)disk(2)
           SCSI Disk:  scsi(1)disk(3)
           SCSI Disk:  scsi(1)disk(4)
            Graphics:  Reality-Engine 2 12GEs 4RMs
The Boot Messages log contains:
BLOWER B FAIL           11:03:01    06 NOV 2007
POWER FAIL WARNING      07:56:13    31 OCT 2007
SYSTEM OFF              13:07:49    19 MAY 2003
SCLR DETECTED           12:42:04    19 MAY 2003