John Baliga (27 February, 2004). Chips Go Vertical. Spectrum Online, Weekly Feature. Retrieved March 11,
2004, from http://www.spectrum.ieee.org/WEBONLY/publicfeature/mar04/0304chip.html.
ever-increasing technology in microchip design is now forcing chip engineers to
become more creative when creating new chips.
The performance of the wires connecting the transistors has become the
limiting factor for microchip performance instead of the number of transistors
on the chips. To overcome the
inherent obstacles of placing so many transistors and wires together the newest
wave of chip design is to stack the individual chips to make three-dimensional
integrated circuits. The article
discusses some front-running design techniques and the companies using them.
the 1960ís the development of microchip technology has been screaming along at
an impressive pace. We have gone
from computers the size of buildings to computers that fit in the palm of your
hand. In recent years the number of
transistors on one chip has reached 410 million and will soon be at 1 billion.
This has caused the recent design problems to be the wires connecting the
transistors (as thin as 90 nanometers) to reach combined lengths of 7 kilometers
per square centimeter. Combine this
wire length with the capacitance added by the insulation on the wires and it all
adds up to unwanted power consumption and heat production.
There are three basic strategies for handling this problem:
three-dimensional chips, X Architecture (using 45-degree angles to connect
transistors) and the network on a chip approach (see article for explanation).
The three-dimensional approach stacks multiple dies with direct
connections tunneling through them. This
reduces interconnect lengths and increases the number of transistors within one
cycle of each other. There are
several different ways to produce the three-dimensional chip and several
different companies are starting to fund the various programs working on this
technology. Tru-Si Technologies of Sunnyvale, California is being funded
by Intel and Ziptronix Inc is being funded by Xilinx for example.
Each company has its own development philosophies and the article
discusses many different ways that chip stacking could be addressed and some
philosophical ideas behind them which I donít think helped the article very
The article concludes by saying that cost and yield will decide whether
these new chip designs ever make it to the market which I feel is a pretty basic
understanding of how corporations work and was just a cheap way to end an
Overall, the article is very detailed as I would expect from Spectrum
Online which is for technically savvy readers.
I do believe that in the middle of the article he deviates from the topic
for a paragraph when he starts talking about pixel arrays and infrared heat
sensing and chemical detection. The
article gives nice historical background and covers the topic very thoroughly
without taking sides as to whether this technology will succeed or fail.