🚀 go-pugleaf

In article <b0qnybk.tcmay@netcom.com>, tcmay@netcom.com (Timothy C. May) writes:
|> messmer@lut.fi (Willi Messmer) writes:
|> : Yes, anyone heard about iAPX 432 32-bit cpu from early 80's ?
|> : 
|> : If so, a brief introduction & history would be nice.
|> 
|> The 432 project started in 1975 as an effort to build a state of the
|> art 32-bit processor exploiting Intel's latest processing technology,
|> then known as "XMOS" but ultimately called HMOS, for high-performance
|> MOS. (I was at Intel at this time, so excuse the emphasis on the
|> technology side....in other summaries of the 432 I haven't seen it
|> given enough attention.)
|> 
|> The original name was the "8800" and a multi-chip set was planned. As
|> the chips were so large, and CAD techniques then were primitive, new
|> CAD systems were developed (also in cooperation with the 8086/80286
|> team, which also needed them, though not as badly!). (An entire tools
|> group developed design rule check software, continuity verifications
|> tools, etc., all long before such things were commercially available.
|> This played a big role in getting chips like the 8086, 8087, etc. out
|> ahead of Motorola, which was still laying out the 68000 by hand!)
|> 
|> Architecturally, the system was variously called "capability-based"
|> (like some IBM machines, notably the System/38 minicomputer) or
|> "object-oriented" (as in Smalltalk). Others can say a lot more about
|> this, including the performance penalties incurred.
|> 
|> (Despite the various problems, I'll never forget seeing a multiple 432
|> system running UNIX as processor cards were pulled out of the card
|> rack...full fault tolerance, in 1981.)
|> 
|> The 432 set was introduced in 1981 and flopped. Poor performance, a
|> very large die size, and the complete break from the 8086 family were
|> probably factors. I think the last chips were sold around 1984, but I
|> don't think there were ever any notable design wins.
|> 
|> The 432 project and team lived on in the "Gemini" project, which
|> became the Intel-Siemens jointly-owned "Biin Computer." The chips
|> developed had some of the 432 capabilities, but were more efficient
|> and didn't carry the same "everything is an object" penalty. The chips
|> are now sold by Intel as the "960" family.
|> 
|> Biin, which wags immediately dubbed "Billions Invested In Nothing,"
|> could have succeeded, I think, in the fault tolerant market, but Intel
|> and Siemens decided to pull the plug. This happened in 1988 or so.
|> 
|> I hope this summary helps. I'm not a computer architect, so take my
|> opinions with care. I often cite the 8800/iAPX432/Gemini/Biin/960 saga
|> when people say the chip industry, or American industry as a whole,
|> only has a quarter to quarter outlook...this project started when some
|> readers of this group were in diapers!
|> 
|
Tim, wasn't the 432 also ADA Microcoded? -Steve