Thin is in--A CPU update

Recently, I have been making homemade pizza. A heavy pizza stone, a
500-degree oven, and a continuing battle to stretch the dough very thin without
tearing it (dough has a mind of its own!) are helping me to approximate the
crisp, very thin crust that I really enjoy. I guess the folks at IBM's
semiconductor research labs must have similar tastes, because at work, they have
now begun stretching silicon!

Brought to our attention by RCFoC reader Don McArthur, the scientists have
demonstrated that by stretching, or ‘straining’ silicon's atomic lattice by
one percent (while being careful not to break the atomic bonds), they can reduce
the resistance that electrons encounter. And that allows these chips to run up
to thirty-five percent faster, while consuming less power! The icing on this
silicon cake is that they're producing these strained silicon chips using
conventional semiconductor manufacturing technology, rather than esoteric lab
processes.

The yield for strained silicon devices does have to improve before they're
ready for the commercial market, but this is expected to occur by 2003.

The intriguing thing about this semiconductor advance is that it demonstrates
that there are ways to increase semiconductor performance at a given level of
miniaturization-this technique yields a 35 percent speed improvement without
shrinking the transistors at all (the ‘traditional’ way of speeding things
up)! Which implies that as Moore's Law continues to shrink the transistors as
well, we will get compounded improved performance. Not bad.

Of course there may be a small fly in IBM's ointment-RCFoC reader Harry
Hardman points out that another company, AmberWave Systems is claiming that THEY
invented strained silicon, and will happily license their intellectual property
to one and all Who owns which, and owes what to whom, can get very complex in
the world of high tech patents…

210 Gigahertz!

But regardless of how that plays out, IBM has some other irons in their
silicon foundry fires. According to a recent announcement, “…they have
developed a new silicon-germanium transistor that has reached speeds of 210
Gigahertz (!), while sipping a mere milliamp of current. This transistor is
expected to yield specialized communications chips that will run at 100GHz
within two years-five times faster and four years sooner than recently-announced
competitive approaches."

The silicon-germanium combination can reach such speeds, in part, because
electrons can move more quickly through this material than through silicon
itself.

But what intrigues me the most about this announcement are comments made by
IBM Fellow and VP Bernard Meyerson in the June 25 New York Times es), where he
says that this development is, "...nowhere near the physical limits of the
technology... The current transistors are about 100 atoms thick, but there was
no reason that it would not be possible to cut that dimension in half, which
would yield a tremendous increase in speed."

Especially on the heels of Intel's recent reflection that “the pace of
silicon development is accelerating" these latest developments represent
another reminder, and just the tip of the iceberg, that as we get better at
tinkering with things at the molecular and at the atomic levels, the results
will have an enormous impact on the things around us.

Today...

Oh, and on the more pedestrian today front, Intel is about to launch the next
increment in P4 speed-1.8 GHz chips should be available next week. Amazing as it
might have seemed just one year ago, the day of the 2 gigahertz commodity CPU is
fast approaching...

...And tomorrow

I can easily imagine a day, when looking back at today's chips will seem like
looking back at a vacuum tube radio-we would not be able to imagine how we could
have survived with such crude, huge, and power-hungry devices. Just the tip of
the iceberg...