List compiled by David Beem. Sources: PS/2 Assistant, empirical evidence.
CPU Type Encoding
The middle digit of the sub-model identification indicates type and clock
speed of the installed CPU, except for the early PS/2 systems (all models 25,
30, 50, 60, 70, P70, P75, and 80) where it has a different meaning.
|3||386SX-16 (duplicate of '1'), Theory: working FLOAT pin?|
|6||486SLC2-40 (CPU daughtercard on 8556/8557)|
|7||Theory: 486SLC3-60? (CPU daughtercard on 8556/8557)|
|B||486SLC2-50 (9533/9553/9556/9557 planar or CPU daughtercard on 9556/9557)|
|E||486SLC3-75 (9556/9557 planar or CPU daughtercard on 9556/9557)|
|U, X||486SX-33 (Theory: one PGA, the other PQFP?)|
Note: We are making a few assumptions here (marked
as "Theory"), to see if it can be fully pinned down...
Possible missing encodings: POD63/POD83, IBM 486DLC2, 486SL.
IBM even had little stickers included with the CPU daughtercards to put over the sub-model label.
What About Zero?
What about 0 for the Type 0 and 0- 386DX-20 complexes in the Model
The 8590-402 was a '4' "Special Bid" unit that makes use of the
sub-model middle digit CPU encoding for other uses - There is no '0' CPU as far
as I know.
Another late night thought is the Intel 486DX3-75 and classic Pentium 100
MHz CPUs not on any factory PS/2 configuration. I believe IBM attempted to
cover every Intel and IBM CPU from the 386SX to classic Pentium at all offered
clock speeds. IBM was anal enough to have a separate encoding of whether the
same IBM 486SLC2 CPU daughtercard was running at 40 MHz (from an 8556/8557
planar host) or 50 MHz (from a 9556/9557 planar host or 9533/9553 planar)
I also thought about checking the CPU encoding for the PS/2 N45 ("PS/2
Note") sub-models since it uses the Intel 486SL-25 CPU (there is also an Intel