IEEE Spectrum February, 2008 - 7

"it would be bloody wonderful if
it worked" -medical imaging expert Keith
Rogers regarding Fermiscan's technology

briefly collaborated, have
now tried and failed to
reproduce her results. "Some
very eminent people have
tried to reproduce it," says
Rogers. "The irritating
thing is I so much want to
believe her, but I can't find
the proper evidence."
Halfway through its
recent trial, Fermiscan
claimed that hair analysis
has an 82 percent probability
of detecting breast cancer in
a person who has the disease,
nearly matching the best
results from mammography
trials. However, hair analysis
has only a 77 percent chance
of correctly confirming that
people without the disease
really don't have it, making the test less reliable than
some mammography trials.
If it succeeds, Fermiscan
could wedge itself into a
lucrative market. About
1 million mammograms
were conducted in 2007
in the United States
alone, according to the
U.S. Food and Drug
Administration. Companies
such as GE, Hologic, and
Siemens are encouraging clinics to upgrade
and purchase digital and

three-dimensional X-ray
mammography machines at
a cost of hundreds of thousands of dollars per unit.
The Fermiscan test will
cost the consumer about
US $200 per sample, says
David Young, Fermiscan's
managing director. That's
more than the $50 to $150 it
costs for a mammogram in
the United States, but with
Fermiscan the clinic wouldn't
need to purchase and maintain its own imager. The
Fermiscan test should take
just 24 hours to complete,
including the time it takes to
ship the hair to the Advanced
Photon Source at Argonne
National Laboratory, in
Illinois, where Fermiscan has
its tests done. Young says that
the company will first market
its services in Southeast
Asia, through a joint venture
with regional healthcare provider Avia Reed
International, in Singapore.
Although most scientists
in the field have given up on
the technique, some remain
hopeful that it will work.
"I wish [them] really well
because it would be bloody
wonderful if it worked,"
says Rogers. -Morgen E. Peck

14

Na * Ieee Spectrum * february 2008

Long the ugly stepchild of computer arithmetic,
division is getting a much needed makeover

w

hat does
the mortgage
crisis have to
do with microprocessor
architecture? It turns out
that calculating prices for
those financially dubious
mortgage-backed securities is a division-intensive
process, and division has
long been the weak link in a
microprocessor's arithmetic
operations. With Intel's
new crop of 45-nanometer
processors, code-named
Penryn, the company is
making the first substantial
upgrade in its processors'
divider since the original
Pentium came out in 1993.
The speedup doubles the
number of bits calculated
with each tick of the
processor's clock and will
make a substantial difference
to financial and scientific
computing. And because
Intel powers so much of
the computer market, the
development could tempt

programmers to retreat from
the less accurate but faster
software tricks they've used
as a substitute for division.
"Divide had become the
long pole in the tent," says
Steve Fischer, the lead
architect for Penryn. "We
tried at least to chop the
pole in half. It's still long
compared to some functions. But it's a lot better."
The new divider is a
variation on the old one,
known as SRT Radix-4. SRT
(for Sweeny, Robertson,
and Tocher) is basically a
souped-up version of long
division that generates two
bits of the answer with each
step. The new Radix-16
divider works fundamentally the same way but
computes four bits in each
step. Getting those other
two bits was no simple task.
"I would say that the divider
really pushed the edge
in terms of the max clock
frequency performance for
www.specTrum.ieee.org

Intel

eye-opening: X-ray diffraction patterns from the hair of women with
[left] and without [right] breast cancer. photo: FermIsCan

intel makes
a big Jump in
Computer- math
http://www.specTrum.ieee.org
Table of Contents for the Digital Edition of IEEE Spectrum February, 2008
