Manufacturing cost breakdown for a 16-bit computer

Question

The dynamics of the market, of what platforms end up being used at what times in what contexts, depend not only on what is technically possible, but even more on what can be done for what price, so manufacturing cost can provide insights into why one computer succeeded, only to be replaced by another. To this end, I've found this breakdown of the manufacturing cost of the Commodore 64 highly illuminating: https://en.wikipedia.org/wiki/Commodore_64#Manufacturing_cost

Is there available anywhere, a similar breakdown of the manufacturing cost of any 16-bit computer? I would, for example, be very interested in knowing exactly why the Atari 520ST was many times more expensive than contemporary late-model 8-bit machines from the same company, despite being designed for simplicity at a time when RAM and the 68000 CPU were both cheap.

Edit: I found a reference for the price on the 68000. From https://www.amazon.com/Commodore-Amiga-Years-Brian-Bagnall/dp/0994031025/ hardback edition page 58, on the selection of a CPU by the Amiga team:

'However, Motorola stood out for price. "We were being quoted $12 apiece in large quantity," says Decuir. "That was better than the Zilog people for the Z8000 were offering. And it was much better than Intel was asking for the 8086 family."

Motorola unveiled the chip in 1980, but samples were not publicly available until 1982, just in time for Miner and his engineers to begin work on their new system.'

Answer 1

Re-reading my answer, I see that I've written about the why rather than the what and failed to answer with a specific price breakdown, but I think it's still useful enough to post:

The main visible cost is the CPU. The selling price of a CPU is not linear with performance but resembles more of an exponential curve to reflect the fact that people are prepared to pay quite a lot more for relatively modest gains in performance with only some of the cost being justified in extra engineering effort, and that few are interested in low-performance parts.

However, a CPU is not a computer, and the rest of the system needs to be considered.

A 16-bit CPU has twice as wide a data bus as an 8-bit CPU. ROMs were typically 8 bits wide, and RAMs 1, 4, or 8 bits wide. So you need to double the number of ROM and RAM chips. You might think that you can just halve the capacity of the individual chips to get the same total size and save a bit of money that way, but there are a couple of problems with that. The simple one is that 16-bit code tends to be less memory-efficient than 8-bit code and that one also generally expects the 16 bit machine to do more, so the system needs more memory. A less-obvious one is that chip dies are square, so memory capacities tend to scale up in powers of four rather than two. Half-sizes existed, but were not necessarily sufficiently cost-effective to justify the reduction in capacity.

Normally the I/O and video subsystem would also be upgraded, since who wants their shiny new 68000-based box to be sullied with 8-bit graphics? (Atari ST owners, ba-dum tish.) Better graphics means faster/wider video RAM, higher-quality video DACs, maybe a blitter or fancier sprite hardware. Floppy drives and controllers become a necessity on machines with lots of RAM, which adds further expense.

A 16-bit CPU also needs twice as many data lines as an 8-bit CPU, and normally has more address lines too, which increases PCB complexity, requiring a larger board to handle more traces, and perhaps more board layers as well. The cost of the PCB is often overlooked, but large boards can cost a surprising amount. The larger board usually means a larger case, and moulding large plastic pieces is not cheap. Consider the size of the Commodore 64 versus the Amiga 500.

You could do clever hacks to reduce the bus width and save money that way. This results in the wretched performance of the Sinclair QL and Acorn Electron, both of which did their bit to ruin their respective manufacturers. So don't do that.

After looking at the overall improvements, one can be left wondering why the 16-bit machines were so cheap in comparison to the much worse-performing 8-bit systems.

Answer 2

This doesn't directly answer your question, but I believe your indirect question is really why 16/32-bit computers were more expensive than 8-bit computers sold at the same point in time.

The answer, in addition to what pndc says in his response, is that 8-bit computers, at that time, had had all their sunk costs recovered years before, and could simply be sold for component cost plus markup. The new systems on the other hand had had a lot of research and development expense made (and continuing, as the operating system was continuing to evolve; Commodore, for example, didn't do a thing to the C64 kernel after the 64 was released except to fix a few bugs). All of this time and effort cost money, and for that investment to make sense, that cost needs to be recovered in the sales of the units made as fruit of the R&D effort.