Gregory's ServerToday's die photo: the Intel iAPX 432 processor (1981). This "micro-mainframe" was so complex, the processor needed to be split across two chips: this is the 43202, which executes microinstructions. The 432 was slow, late, and a failure.
 9 comments
The instruction set was very complicated, object-oriented, and not byte aligned, with instructions from 6 to 321 bits long. It took a whole chip just to decode instructions. This chip then executed them. Here's the die with the main functional blocks labeled. My favorite part of the layout is in the corner where the bus lines turn from vertical to horizontal with a diagonal transition. That left a gap in the corner, so they kind of threw the clock buffer circuits into the unused space. Thanks to Tom Lyon for providing the wafer. With a big die and a small 4" wafer, each wafer only held a few chips, especially when many of the ones around the edge were truncated. 
[DATA EXPUNGED]
@rotopenguin Those are test dies to check the characteristics of the wafer. They need to be spread out across the wafer to get accurate measurements. Pushing them to the edges would defeat the purpose.  @kenshirriff Do you have any idea what the five very different chips are - the top, bottom, left, right and centre ones? I guess some kind of test chip?   |
The iAPX 432 was a follow-on to the Intel 8008 and 8080, originally called the 8800. It was supposed to be Intel's flagship processor but that didn't work out. This chip has the label 8802 since it is the second chip.