[libre-riscv-dev] interesting article about abolishing motherboards and replacing with silicon

[Samuel Falvo II]

On Wed, Sep 25, 2019 at 9:27 PM Jacob Lifshay <programmerjake at gmail.com>
wrote:

> PCBs won't completely die anytime soon. also, the silicon wafer used
> instead of a PCB should be relatively inexpensive, because it doesn't need
> high resolution (few um rather than few nm).
>

Despite all that being true, I distinctly recall that it took many decades
after the widespread adoption of surface mount technology in commercial
manufacturing before it became commonplace in the homebrewer's lab.  Even
then, most products utilizing surface-mount tech uses pick-and-place
services from commercial fabs.  I'm not even talking about production runs,
either.

Some problems I see with Si-IF which heavily offsets its benefits:

1) Fabricating silicon wafers takes a lot longer than fabricating PCBs.
This won't affect large-scale production runs, but it will deleteriously
impact home-brew and one-off/R&D design flows.
2) Si wafers are (for all intents and purposes to most people) shards of
brittle glass.  PCBs are fiberglass (usually).  To prevent damage from
flexure, devices built with Si-IF will need more robust housings, which
will drive costs up.  Whether the cost of Si-IF + more robust housing is
lower than traditional approaches remains to be seen.
3) Si-IF has no capacity what-so-ever for bodging.  So, if you get a few
test IFs back and they don't work because you've mistakenly wired something
wrong, you can't just cut a trace and solder a wire or two.  You have to
rework the entire design and wait for a new Si-IF wafer to arrive all over
again.  The investment of producing that batch of Si-IF wafers will be sunk.


-- 
Samuel A. Falvo II