The Tyranny of Average Cost - part the twentieth

Here we go again. Today's NYT tells the tale of Intel and the future of 'other' processor architecture. To see how life used to be, go back and re-read "The Soul of a New Machine". The principle lesson from doing so is the realization that, at that time and previously, a hardware engineer could conceive and build a bespoke computer and sell it to make money. The various descrete components were all listed in the catalogs of myriad semiconductor producers; one need only map out the circuit from one's brain to paper, find the necessary components, buy a handful of each, and build a prototype. The single-chip computer (of at least 32 bits, anyway) in commodity quantity, didn't yet exist. While I can't find any specifics about the individual components of the machine (memory says the book talks about that in detail, but that's the best my memory can do and I can't find my copy), here you can see that this ain't today's cpu on a chip!

Not so much these days. TSMC has come to dominate chip production just because of the Tyranny. The world of economics has recognized the notion of the natural monopoly from its beginning; such products existed for millennia before the field of economics was invented. With the unhindered march of capitalized production, the number of descrete low cost production units (whether owned by a monopolist or a few oligopolists) continues to shrink (all the while the unit grows more complex).

This is the most high-larious quote from the Times piece:

Sapphire Rapids began in 2015, with discussions among a small group of Intel engineers. The product was the company's first attempt at a new approach in chip design. Companies now routinely pack tens of billions of tiny transistors on each piece of silicon, but competitors like Advanced Micro Devices and others had started making processors from multiple chips bundled together in plastic packages.

New? Not hardly. Granted the kiddies of today likely have no direct experience with anything that isn't today's X86 chips, but the notion of multi-chip cpu isn't ancient history. The IBM RS/6000 began life as a multi-chip implementation, which is what I chose for my employer about 1990 to run two Progress applications. It was a few years later that the first single-chip RS/6000 came to be. And from there to the PowerPC, and when Steve said, "no mas" oblivion. The reduction of general purpose cpu to just two architecures (X86 and z) would take some time. Now they're in a horse race with a, titular, RISC machine called ARM.

At the time of its introduction in 1996, the P2SC was the largest processor with the highest transistor count in the industry and was a leader in floating point operations.

So, Intel (and others, of course) are reversing course and going with multi-chip implementations. Not that these are built with commodity ICs, by any means. It's as much a matter of yield as density increases. Even if they can build a chip at 5nm, with X billions of transistors per, how many good ones can be gotten from a wafer? Only the Shadow knows.

Will Intel, unlike IBM and myriad others like Data General before, be able to out-duel TSMC in production mojo?? Hard to tell. Once again, the issue comes down to the Tyranny. TSMC still has the biggest piece of the cpu pie. And thus, lowest average cost. Or so we think. Will Intel really become a successful commodity producer? Ask the Shadow.