Scientists at Rice University may have overcome a fundamental barrier to the continued miniaturization of computer chips, clearing the way for faster and smaller chips than have ever been possible with traditional silicon.
For a while, it seemed like the proverbial good times would never end when it came to computer development. Everyone wants things smaller and faster, and manufacturers have been more than happy to provide.
In 1965, Intel’s co-founder Gordon Moore famously claimed that the number of transistors that can be realistically and inexpensively placed on an integrated circuit should double approximately every two years. The observation became known as Moore’s law, and for over four decades, it has driven the computer industry to continue to push to increase capacities while shrinking computer chips. But eventually things have to stop. The limitations of the current generation of technology are beginning to show, and soon manufacturers will hit a fundamental block based on physics and expenses. Enter a new generation of technology.
Scientists at both Rice University and HP are separately reporting that they have both found ways to overcome that fundamental, and fast-approaching barrier. And they’re not alone. Intel, IBM, and others are also expanding their research into new ways to make sure the “good times” continue to roll.
According to the NY Times, researchers from Rice claimed to have discovered a way to build small, vitally needed digital switches at scales that have previously been thought impossible, by using silicon oxide. Silicon oxide is a common chemical compound used primarily as an insulator by the industry. Its ready availability and ease of manufacturing make it commercially appealing.
Working with a Texas-based startup called PrivaTran, Rice scientists have begun to make prototype chips. The new chips use filaments that are five nanometers in width — a nanometer is one billionth the size of a meter. That makes the filaments already thinner than what the computer industry hopes to produce within the next ten years using current manufacturing techniques. The discovery was made by Jun Yao, a graduate researcher at Rice, who said he accidentally discovered the switch. At the moment, the chips can only hold 1,000 bits of data, but within five years scientists promise that these new chips could rival any current chips for capacity.
HP is also planning an announcement of a partnership with an unnamed semiconductor company regarding similar, but competing technology. Both HP and Rice are developing what are called memristors (memory resistors), switches that can retain information without power in the same way that flash memory can retain data. The technology being developed by Rice and HP has existed in theory since the 1960s, but there has not been a practical way to realize it until now.
Currently chip manufacturers have made finding or developing next-generation technology a priority. Each new advanced chip-making factory can cost in excess of $4 billion to properly stock it with the tools and equipment needed to make the current generation of chips. One current method for expanding on existing technology is to increase the density of chips by layering them, and stacking circuits on top of each other. The process is costly, it is producing results, and it still has room for growth — but it is a stop gap measure, and will eventually reach its limits.
Not to be out done by Rice or HP, IBM, Intel and others are developing competing technology called phase-change memory, which relies on heat to transform a glassy material into a crystalline state, then back again. The technology has shown promise, especially for flash chips which retain memory without power.
Don’t expect any major new technologies to hit the shelves for a while yet, but the future for computer chips once again looks limitless.
No comments:
Post a Comment