Page 7 - ChipScale_Mar-Apr_2020-digital

GUEST EDITORIAL

Post–Moore’s Law electronics: now, until

quantum electronics

By Rao R. Tummala [Georgia Tech]
M the driving engine for in the past with transistor scaling as better computing performance at lower

oore’s Law has been
the primary focus, this new vision
power than current electronics, is yet
science, technology,
m a n uf ac t u r i n g , is referred to as package and system another reason for the end of Moore’s
scaling—its focus is on miniaturization
Law. The human brain is the ultimate
hardware, software, systems, and of active and passive components and system of interconnections for the highest
applications, cont r ibuting to the interconnecting them to form modules performance in the smallest size with
prosperity of thousands of individuals, and systems.
and hundreds of corporations in dozens Just like Moore’s
of countries during the last six decades L a w h a s b o t h t h e
[1]. As Moore’s Law begins to come to an doubling of transistors
end for not only fundamental reasons, but and simultaneous cost
also for computing performance, power, reduction from node
cost, and investments [2] (see Table 1), it t o no d e e ve r y 18 -
is becoming clear that a different vision 24 months, Moore’s
for electronics systems must emerge. L a w f o r p a c k a g e
The first step in this new vision i n t er c o n n e c t i o n s
is a shif t f rom Moore’s Law for c a n d o t h e s a m e .
integrated circuits (ICs) to Moore’s I n ter c o n n ec t i o n s
Law for Packaging by using package have been driven by
interconnections (I/Os) in 2D, 2.5D computing systems,
and 3D between smaller chips with and within computing
the same or higher transistor density, systems, between logic
as well as other components. The next and memory, consistent
step is a shift to optoelectronic package with Von Neumann’s
interconnections. As these transistor- and a rc h i t e ct u re . T h e
interconnections-based electronics begin new era of artificial
to come to an end, quantum electronics intelligence mimicking
is expected to take over. This concept is t h e h u m a n b r a i n ,
shown in Figure 1. with several orders Table 1: Eleven reasons to replace Moore’s Law.
A new vision
This paper describes a shift from
transistor scaling with the most number
of transistors in a single large chip [3],
driven by Moore’s Law, to the package
and system scaling concept that relies
on the most number of interconnections
(I/Os) between smaller chips in the
short term, and to quantum computing
in the long term. So, while transistor
integration to a 20-billion transistor
chip on individual ICs was the basis
of Moore’s Law for ICs to date, it can
be extended by a different vision. In
the short term, this vision involves
new paradigms in electronic and
optoelectronic package interconnections
in 2D, 2.5D and 3D (Figure 2). Unlike
Figure 1: Computing performance driven by Moore’s Law and post-Moore’s Law technologies.

Chip Scale Review March • April • 2020 [ChipScaleReview.com] 5 5