On Sat, 28 Apr 2001, Rohit Khare wrote:
> It would be interesting to imagine what happens when chips get fast
> enough that *RAM* is "a few weeks away by stagecoach".
blah blah won't happen unless you want it to. there's always a way to
get the memory closer, as long as you don't cling to the idea of RAM as
chips that plug into slots. All aboard; no, all achip. Designing it on
the die kills modularity (for commercial sale/upgrading) and makes power
issues even more nightmarish wrt the layout, but at .05 and lower
(biotech or whatever your latest personal ultrabips wet dream is), who's
gonna have a board fast enough in relative terms to do it otherwise?
Allegro and the like are going to run out of steam, yes. There's only so
much you can do on a board. The interfaces are always brutal. Even when
we get biocomps they have to have I/O that's detectable on some mechanical
level and then translated up to where we can see the output. I don't care
about the speed if the bio-flash output to fiber from your ganglion takes
1 million cycle's time to recharge.
(life would be easier if timing didn't
matter and packaging was a formality)
This archive was generated by hypermail 2b29 : Sun May 06 2001 - 08:04:36 PDT