When the
first computer chip design pioneers huddled around in their
low-tech, under-financed labs, one of the decisions they had to make
was whether to stick in a clock.
They needed something that would regulate all the components of
the microprocessor in a reliable way, if they were to build
something people could depend on. They decided there was no way
around it, they had to go with a clock.
Design historians say they made the right choice at the time. In
any case, chip designers ever since have been trained to design
microprocessors to work to the steady, if lightning-fast, beat of a
clock.
It's worked great so far. But there will come a time soon when
those tiny, oscillating crystals in chips will bump head on into the
laws of physics. That's why a handful of designers today are working
on chips free from the constraints of man-made time.
.gif)
No More Speed
Wars?
It's difficult to conceive of chips that don't use clock speeds
as measures. The speed wars between Intel, AMD and others have been
on the high-tech media radar for years. Clock speeds are something
that people who otherwise know very little about computers look for
when they scan the shelves or catalogues for a new machine.
Even the most information-resistant computer user knows, for
example, that a 1 gigahertz processor is faster than a chip with a
clock speed of 500 megahertz. What the user may not know, however,
is that a 1 GHz chip is not twice as fast as the 500 MHz.
That's because of the fact that as the speeds of chips increase,
the clock itself uses more and more power, in order to be able to
synchronize the ever increasing number of transistors on the most
advanced, sophisticated chips, such as Intel's Pentium 4.
Nowadays, a clock can consume almost a third of the chip's
computing capacity.
Small Group, Big
Believers
Electronic pulses that skitter across chips carrying information
barely make it between ticks of the 1 GHz chips. When 2 GHz chips
start to come out in a couple of years, however, there will be a
problem. The clock's traditional role will begin to break down.
The believers in clockless chips, known as "asynchronous" or
"self-timed circuits," say it is inevitable clockless chips will
eventually replace conventional ones.
Most of the big firms, like IBM, Intel and Sun Microsystems, have
devoted portions of their research and development to asynchronous
chips. Universities have started teaching students about the
clockless concept, and there are a number of small startups devoted
to the concept.
Weak Links
Asynchronous chips use power only when there is a reason, when
there is a computing task at hand. The transistors on an
asynchronous chip can exchange data independently, as opposed to
conventional chips, which must wait for every component to do its
job before a task can be completed.
A conventional chip can run no faster than its slowest component.
Asynchronous chips can run at the average speed of all components.
The advantages are numerous. Not only are they faster, the
efficiency of their electrical systems are much higher, which means
longer battery life.
In addition, they can perform encryption better because they give
off no regularly timed signals that hackers can identify and
exploit.
Also, asynchronous chips emit very low levels of electromagnetic
noise, which means they are far less likely to interfere with other
devices, one of the biggest problems for today's mobile chips.
Prototypes
Exist
With their potential to revolutionize the chip industry, the
logical question is: why haven't these chips become more widely
adopted?
Intel, Sun Microsystems and IBM all have prototypes that are two
to three times faster than their conventional counterparts, yet the
test chips never made it out of the lab.
Philips Electronics has an asynchronous chip-powered pager that
runs almost twice as long as others on the market. Few other
companies use machines powered exclusively by asynchronous chips,
though there are some companies that have incorporated asynchronous
concepts -- the Pentium 4 Intel released this year, for example.
Will Catch on
Slowly
The answer lies in the nature of the semiconductor industry
itself. Over the course of 20 years, the chip industry has invested
billions in streamlining testing, developing and manufacturing.
With no mass market to exploit, companies have no financial
incentives to develop the "tools" needed to manufacture clockless
chips. It would take much longer and be more expensive to get the
chips to market, and few profit-driven companies will go to the
expense. Also, it has also proven difficult for companies to find
asynchronous designers, since the concept is contrary to what most
schools have taught for decades.
Still, analysts expect the new chips to catch on slowly,
especially in the mobile market, until conventional chips can no
longer do the job. 
September 24, 2001 
See more news
Front Page
Talkback: 
.gif){kind=icon}
Virus
Outbreaks: Are They All Microsoft's Fault?.gif){kind=small}
