From: Eugene.Leitl@lrz.uni-muenchen.de
Date: Mon Jan 22 2001 - 14:39:50 MST
Damien Broderick wrote:
> < Immediate prospects include a hugely expanded Internet, driven by
> plummeting computer chip prices--well under a dollar apiece by 2000, around
> 5 cents by 2005, a cent by 2010. >
>
> How did that work out, so far?
He doesn't say what kind of chip. You wouldn't get very much in an
easily attainable (effective power at the user end, using mainstream
development tools and vanilla developers) absolute performance for
a single dollar in 2000. If he says "Internet", he doesn't say
whether he means the matrix (switches, routers) or the connected
nodes running the clients. Currently, there is a large difference
between these two.
The actual hardware prices are relatively irrelevant for switches
and routers, as the complexity of the switching bare metal is low.
The sofware layer is the dominant part, and being speciality
items, the things have a high price markup. Instead of a trunk
and tree/leaf mentality, using low end hardware on locally-connected
grids would scale better and have a better price/performance ratio.
Of course the performance of individual routers grows, but I very
much doubt they constitute the bottleneck of Internet growth. I
suspect the bottleneck to be present in the missing last mile (same
tree/leaf thinking again, but the problem is real as long people
believe in it), and too much copper instead of fiber, and
in politicking (if I have surplus bandwidth I won't sell it to my
competitors, same thing vice versa).
The desktop bloat can absorb any amount of hardware thrown at
it. The current lull in domestic market PC sales seems to
indicate that these people who intended to buy a PC now already
have one, and that it does fit their current and near future needs.
This is not true for large parts of the world. We certainly have
this more than compensated on server, infrastructure, gaming and
embedded market. Reality simulation is expensive, so as long people
buy gaming gear, hot silicon will sell there. New killer applications
might emerge, such as real voice recognition, which will require
order of magnitude more crunch than current machines can provide.
The PDA and the cellphone will see convergent evolution. We might
be almost ready for the Dynabook, even. The wearable market does
not even exist yet. The home automation market ditto. As is vehicular
navigation and entertainment. And the Internet, of course, as we're
still very far remote from high-resolution high-QoS streaming video.
This archive was generated by hypermail 2.1.5 : Sat Nov 02 2002 - 08:05:15 MST