HP’s New Design - A Breakthrough in Chip Technology

The way to even smaller and cheaper electronics may be just around the corner now, thanks to a breakthrough in chip design by scientists at Hewlett-Packard. Scientists at HP may have found a way to wire a computer chip that can make this kind of a technology breakthrough possible.

While the breakthrough technology is still in the simulation stages, spirits in the company are running high, with the director of the Quantum Science Laboratory at HP Labs, Stan Williams predicting a working copy within the next one year.

About 30 years back, Gordon Moore of Intel had come up with Moore’s Law, which states that the number of transistors that can be placed on a chip doubles roughly once every 18 months. Scientists have recently been struggling to keep this theory going, as spurts in innovation have led to scientists shrinking the size of the chip till the maximum possible limit already.

The technology breakthrough at HP may enable scientists to skirt the issue now. Williams said, “Moore’s Law is on a collision course with the laws of physics.” HP’s breakthrough may enable scientists to shrink the size of the wires connecting transistors, which would mean that there would be no more need to shrink the size of the transistors.

Basically, the working unit of the chip is the transistor, which works based on the binary system, and is like a gate. It exists in two states, open and closed. A transistor registers ‘one’ when open and ‘zero’ when closed. The more transistors on a chip, the more powerful the chip becomes. In the context of an application, this logic extends to the chips themselves.

Until now, the only way scientists have been able to make chips smaller, faster, and cheaper has been to reduce the size of the transistor. However, beyond a point in time, the cost factor works in reverse, it becomes costlier to make smaller transistors. HP’s novel approach could be just the solution scientists were looking for: shrink the size of the wires, and not the transistors.

To understand this better, one can take the analogy between a chip and a city. The components of a chip are arranged like the buildings in a city. The wires run between the transistors on a chip, connecting them just as the roads connect the buildings in a city. So long, the wires were running between the transistors, taking up valuable space. HP’s new technology enables them to have the wires above the transistors, thereby enabling the transistors to be packed tighter and creating valuable space.

Williams said the earlier scheme of wiring resulted in wires occupying over 80 percent of space on a chip. However, with the new technology, this number would fall drastically.

According to Williams, the new technology involves two layers of conducting material that are arranged perpendicular to each other in a grid above the transistor. Between these two wires is a third component that acts as an insulator. However, applying voltage to the top layer causes the insulator to act as a conductor itself. This is akin to throwing a switch. To turn off the switch, the same voltage is applied to the bottom wire.

The new technology would work wonders for field-programmable gate arrays (FPGAs), a versatile chip that is currently used in cutting-edge electronic technology such as in big-screen TVs.

One has to understand that these ‘wires’ are basically invisible, as their thickness is measurable in atoms. This is a very important statistic as, in the words of Ivo Bolsens, chief technology officer at Xilinx, a chip manufacturing company based in San Jose, “If you're talking about the difference between something that's supposed to be 100 atoms thick and it's 102, that's not a big difference. If it's supposed to be four atoms thick and it's two, that's another matter.”

Bolsens, who is familiar with the current work at HP, however, had an interesting observation, something which HP itself has acknowledged – the incredibly small size of the wires could mean that a lot of the switches may not work (HP has accepted that the defect rate could be expected to be relatively high). However, Bolsens said, the special nature of the chips themselves meant the defects could be tolerable.

As can be deciphered from the name, FPGAs can be reprogrammed after installation. So even if the connections were unreliable enough to render a part of the chip useless, routing around the problem could be a feasible option.