Bunnie Huang
COLUMN
SPARK
Bunnie Huang
Andrew ‘Bunnie’ Huang is a
hacker by night, entrepreneur
by day, and writer by
procrastination. He’s a
co-founder of Chibitronics,
troublemaker-at-large for the
MIT Media Lab, and a mentor
for HAX in Shenzhen.
ne of the most remarkable
aspects of Moore’s Law is
how smooth the rate of
improvement had been
up until a few years ago.
(or five decades, maLor
industries could count on regular
improvements to performance and cost.
In hindsight, this smoothness was a
result of do\ens of foundries Lockeying for
pole position. If one foundry stumbled,
another would take its place. *owever, as
transistors became much smaller than the
si\e of the photons used to draw them,
shrinking dimensions was no longer Lust a
matter of better
optics or clever
computational optics
tricks. #t nm,
something
fundamental had to
change: the industry
moved to double-
patterning and
FinFET transistors.
This was a difficult
transition that fundamentally changed the
structure of transistors. 7p until nm,
things had gotten better and cheaper, so
Lust about every conceivable circuit rode
the Moore’s Law train for free. *owever,
from nm onwards, transistors got
denser and better, but stopped getting
cheaper. Thus, only applications that could
Lustify the higher cost continued to ride
the Moore’s Law train, marking the
beginning of the end of Moore’s Law.
In recent years, the number of players
has continued to dwindle to Lust three:TSM%, Samsung, and Intel. Progress has
been halting; instead of a continuum of
performance improvements, the industry
is forced to contend with quantised
leaps in performance. Intel’s absence
in the 1 nm and nm nodes has been
notable, leaving P% manufacturers
struggling to give consumers reasons
to upgrade their existing computers.
4ecently, TSM% has filled the leadership
vacuum in the nm node, and everyday
consumers have felt this in the form
of AMD’s Ryzen renaissance.
9hat’s beyond nm? The industry is
finally about to embark on a huge
paradigm shift from
using 1 nm deep
ultraviolet light, to
1.5 nm extreme
ultraviolet (EUV) light.
EUV is hard in so
many ways; it’s
reported to gobble a
megawatt of power to
produce Lust
watts of '7V light,
and it requires landing a bullseye – not
once, but twice – with a powerful %1^2 laser
on droplets of ultra-pure tin that are 25
microns in diameter, falling at a rate of
5, times per second. The big news is
that TSMC has recently gone into
production with '7V, marking the
beginning of a new quantum for Moore’s
Law. This should hopefully unlock a trickle
of incremental improvements over the
next couple of years – and then we’ll have
to wait again, until the next quantum of
Moore’s Law.The quantisation
of Moore’s Law
It’s 2019 – where are our faster chips?
O
EUV is hard in so many
ways; it’s reported to
gobble a megawatt of
power to produce just
200 watts of EUV light@bunniestudios