iAPX432 : Gordon Moore, Risk and Intel’s Super-CISC failure

I found this via OSNews:

From the article:

If we can learn a lot from studying technology failures, then the Intel iAPX432 should be a rich source of lessons. The most common reaction to Intel’s ‘flagship’ processor development program of the late 1970s and 1980s is probably “just what were they thinking?”

1 Like

One thing that hurt performance quite a bit and isn’t mentioned in the article is that the 432 processors connected to each other and to memory using a 16 bit wide packet bus. To read a single 32 bit word from memory you had to send the command and the top 8 bits of the address in the first cycle, the lower 16 bits of the address in the second cycle, then load the first 16 bits of data in the third cycle and finally load the other 16 bits of data in the fourth cycle.

As the Acorn people found by playing around with adding different processors via the “tube” in the BBC Micro, the performance of non trivial von Neumann computers is proportional to the memory bandwidth independent of the internal details. So just expanding the 432’s packet bus to 32 bits might have nearly doubled its performance.

A 432 system could have multiple packet busses, but that is for reliability instead of performance. Even so performance would be helped since multiple processors could then access memory without stepping on each other’s toes so much.

1 Like

In 1981 (or thereabouts), Intel gave the UC Santa Barbara Computer Science Department some 432 development systems. Although I was an undergraduate I was privileged to sit in a seminar led by Prof. Paul Eggert, who went on to do many interesting things, dedicated to figuring out what to do with them.

The bottom line was: we couldn’t. We didn’t have the resources to create major system software completely from scratch. Nobody could see a way to port Unix because of the capability-based addressing. I think the systems just languished.

Ironically, I moved to Oregon in the late 80s to work for an Intel spinoff called Sequent. I worked there with several alumni of the 432 project. If you brought it up, they’d pretty much just laugh and shake their heads.