HP's Chip Research Extends Moore's Law, Fuels Industry Growth

More than 40 years after the introduction of Moore's Law, which observed that the number of transistors that can be packed into a chip doubles every 24 months, many in the semiconductor industry have begun to wonder how much longer the law could hold true.

Now, Hewlett-Packard's research into a new chip architecture may give Moore's Law a longer life, as well as help sustain rapid-fire growth in the technology industry as a whole, according to analysts.

"We've really been running out of headroom," says Rob Enderle, president and principal analyst of the Enderle Group. "We were looking at Moore's Law breaking down. This kind of breakthrough could keep it going. And Moore's Law is what fuels the processor industry. It drives the tech industry. If it slowed down, it would be harder to come out with compelling new products."

Moore's Law started out as a prediction made in the mid-1960s by Intel co-founder Gordon Moore. He basically said that roughly ever two years the number of transistors on an integrated circuit will double in speed and capacity for the same amount of money. Since then, that statement has held true. Processing power and speed have exploded over the years. The number of transistors in a chip has skyrocketed.

The question lately, though, has been how this kind of rapid acceleration could continue. As transistors grow ever smaller, the miniscule digital roads and gates between them, that stop or allow electrons to pass through, begin to leak energy. It ultimately makes the chips too hot and takes up too much energy diagnosing and solving the problem.

And heat and energy are two main trouble points of microprocessing today.

As a result, some industry watchers have been wringing their hands in recent years, saying Moore's Law simply can't stand up for many more years.

HP's announcement this week, may allay some of those fears.

Company researchers, Greg Snider and Stan Williams, are publishing an article in the journal Nanotechnology next week showing how they redesigned the chip architecture so eight times as many transistors can be built into each chip. They also say the new design uses less energy for any given computation.

The new chip architecture is in the research phase now, and HP announced that it could be "technologically viable" by 2010. HP has great resources in the research realm, but the manufacturing would have to go to companies like Intel, IBM, or AMD.

HP didn't shrink the transistor size, which means the new chips could be manufactured in current fabrication facilities by making only minor modifications, HP reports.

The company explained in a report that the research could lead to the creation of field programmable gate arrays up to eight times denser. The FPGAs are integrated circuits with programmable logic components and interconnects that can be adapted by end users for specific applications. They are used in a wide range of industries, including communications, automotive and consumer electronics.

The technology, according to HP, calls for a nanoscale crossbar switch structure to be layered on top of conventional CMOS (complementary metal oxide semiconductor), using an architecture HP Labs researchers have dubbed "field programmable nanowire interconnect." This is a variation on the well-established FPGA technology.

"It's basically changing the typical transistor design to more of a switch architecture," explains Jim McGregor, a principal analyst and research director at In-Stat. "HP's announcement is a step in the right direction. HP isn't the only one looking at ways to do this. It's very early to tell how these things will play out."

Dan Olds, a principal analyst of the Gabriel Consulting Group, Inc., says this kind of development could extend the life of Moore's Law, though he'll be more convinced of it when he sees it in working silicon.

"They've said we're bumping up against the laws of physics and eventually that will be true," says Olds. "But it seems that smart guys at HP and Intel and IBM keep coming up with workarounds that allow the advances to continue. Will these advances in performance be in lock step with Moore's law? That's the question."

Olds also points out that it's not the law itself that is so important to the industry, but the advances that have been made along the way.

"It's quite important to the industry to do this because I don't see the need for computing going down any time soon," adds Olds. "We're less than 20% towards a digitized, interconnected world. We're not to the point where we can open our garage doors from the office or find out what's in our refrigerator and have it automatically generate a shopping list. But it's coming. That will take a vast amount of computational power. We really need Moore's Law to continue."

This story was modified on Jan. 22 to correctly identify complementary metal oxide semiconductor.