A Smaller, Faster Future For ChipsA Smaller, Faster Future For Chips

The economic downturn and anemic computer sales haven't halted advances in semiconductors. That's because demand for faster, cheaper PCs and servers hasn't diminished.

Two weeks ago, scientists at Hewlett-Packard and the University of California at Los Angeles patented a technique for building molecular-scale logic chips, which they believe will, within a decade, work side by side with silicon semiconductors and eventually replace silicon chips.

Replicating earlier work they'd done on molecular-level memory chips, the scientists invented a process that uses wire grids each only a few atoms thick and linked by tiny, molecular-size electronic switches.

The scientists demonstrated in the earlier work how rare earth metals can form nanoscopic parallel wires when exposed to a silicon substrate. They also showed how a one-molecule-thick layer of electrically switchable molecules, called rotaxanes, could be sandwiched between regular-size wires to trap molecules, which could then receive downloaded signals. Those techniques created simple logic circuits one-hundredth the size of today's chips, but the electrical signals on the tiny grids interfered with each other where the wires crossed. In the most recent work, HP and UCLA researchers solved the interference by breaking the wires into smaller lengths using insulators placed at intersections on the grid. The chips are so small that scientists create them and alter their structure using chemical and electrical processes instead of physical means.

These efforts are eventually expected to lead to computer chips hundreds or thousands of times smaller than today's processors. However, leading chip vendors have already made real-world gains in semiconductor research and manufacturing techniques resulting in smaller, faster chips.

The recent advances will be most significant to businesses that are constantly seeking better performance. In particular, Intel's progress in processing speed and power consumption will appeal to buyers of servers that have typically favored higher-performing and more reliable servers based on Unix architectures, says Brooks Gray, an analyst at Technology Business Research.

In early January, Intel began selling the 0.13-micron version of its Pentium 4 chip for PCs and its Xeon chip for workstations and servers. It's been shipping 0.13-micron versions of its lower-end Celeron chip since May; by midyear, Intel plans to offer 0.13-micron versions of its top-end Itanium chip.

The 0.13-micron designation means that each transistor on the surface of a semiconductor is 0.13 microns wide; the smaller size means more transistors can be squeezed onto a chip without increasing its size. For example, a 0.13-micron Pentium 4 has 55 million transistors, compared with 42 million transistors on a 0.18-micron Pentium 4. It also means Intel can produce more finished chips from a standard-size silicon wafer, lowering production costs and ultimately reducing prices for finished products.

According to Intel, the improvement from 0.18 micron to 0.13 micron means the chips operate faster and require less power, resulting in smaller, faster, and less-expensive computers. For instance, the Pentium 4 at 0.13 microns has a top clock speed of 2.2 GHz vs. 2.0 GHz for the 0.18-micron Pentium 4.

Power consumption is critical to Intel's chips. "We see [power] becoming a fundamental part of our design process," says Wilfred Pinfold, technology director of microprocessor research at Intel Labs. In addition to obvious benefits such as longer battery life, reduced power consumption lowers operating temperatures, which means computers will require fewer components to cool them.

In November, Intel unveiled two technologies that together form a new structure called a "TeraHertz transistor.'' The first technology involves building transistors in ultrathin layers on top of a layer of fully depleted silicon, which lets the transistor switch on and off more than a trillion times a second and reduces current leakage when the transistor is turned off. The second employs a new material designed to boost performance and cut power consumption.

Advanced Micro Devices Inc. will begin shipping its own 0.13-micron chips this quarter on its Athlon XP for desktops and its Athlon 4 for notebook computers, the company says. It's also working on a new line of chips, code-named Hammer, that will incorporate 64-bit technology for performance that will be at least 50% greater than current models. The Hammer line will use a new, high-speed internal bus architecture called "hyper-transport" that's 10 times faster than current architectures and an on-chip memory controller that also boosts performance. The new chips also will incorporate "silicon on insulator" technology that AMD says will have a 30% advantage over all-silicon technologies.