A lost IBM PC/AT model? Analyzing a newfound old BIOS

Some people not only have a very particular set of skills, but also a very particular set of interests that happen to align with those skills perfectly. When several unidentified and mysterious IBM PC ROM chips from the 1980s were discovered on eBay, two particular chips’ dumped contents posed particularly troublesome to identify.

In 1985, the FCh model byte could only mean the 5170 (PC/AT), and the even/odd byte interleaving does point at a 16-bit bus. But there are three known versions of the PC/AT BIOS released during the 5170 family’s lifetime, corresponding to the three AT motherboard types. This one here is clearly not one of them: its date stamps and part numbers don’t match, and the actual contents are substantially different besides.

My first thought was that this may have come from one of those more shadowy members of the 5170 family: perhaps the AT/370, the 3270 AT/G(X), or the rack-mounted 7532 Industrial AT. But known examples of those carry the same firmware sets as the plain old 5170, so their BIOS extensions (if any) came in the shape of extra adapter ROMs. Whatever this thing was – some other 5170-type machine, a prototype, or even just a custom patch – it seemed I’d have to inquire within for any further clues.
↫ VileR at the int10h.org blog

I’ll be honest and state that most of the in-depth analysis of the code dumped from the ROM chips is far too complex for me to follow, but that doesn’t make the story it tells any less interesting. There’s no definitive, 100% conclusive answer at the end, but the available evidence collected by VileR does make a very strong case for a very specific, mysterious variant of the IBM PC being the likely source of the ROMs.

If you’re interested in some very deep IBM lore, here’s your serving.