Source article

MIT researchers are developing an alternative to solid-state amplifiers. The new amplifier draws upon an old technology — vacuum tubes — and a technology newly developed at MIT called a “ribbon electron beam.” Ribbon beam amplifiers (RBAs) are smaller, generate less heat, require smaller backup batteries, are more electrically efficient, and cost thousands of dollars less than solid-state amplifiers. Installed on cell phone towers, they would reduce the cost of delivering voice and data from the current 50 cents per megabyte to five cents per megabyte. They may be on the market in two years.

3D Chips

Source article

In an effort to avoid a breakdown in Moore’s Law, as convetional two-dimensional chips bump up against the limits of physics, researchers at Rensselaer Polytechnic are developing 3D chips. The idea is hardly new, and some companies today “package” whole chips layered one on top of another. But true “monolithic” 3D chips will have just one base layer of silicon with active wafers layered on top. The layers could include memory, or even bioactive components, and enable functionality not yet imagined. Rensselaer is working with IBM through a government grant, and also has collaborations with other technology companies including Sematech and Freescale semiconductor.

Intel Photonic Chip

Source article

In 1965, when Intel cofounder Gordon Moore Moore coined his famous Law, which held that the number of transistors on a computer chip would double every two years, Intel’s chips contained just a few dozen transistors. Today, its best chip has more than 1.7 billion transistors, and by 2012 will likely have 10 billion. So processing on the chip will be phenomenally fast. Unfortunately, computers won’t be able to move data at anything like that rate along the copper wires that link their many chips internally.

“Enter,” announces Technology Review grandly, “the silicon laser.” In other words, enter a silicon chip that can share its data with other chips photonically rather than electronically, providing essentially the same benefit communication networks get by linking our phones and computers together around the globe with optical fibers rather than copper cables. Intel has made “remarkable progress,” reports the Review, on an all-silicon laser it demonstrated only last Winter, though there is a way to go before we will have them in our PCs. They key breakthrough was in coaxing light out of cheap and easy-to-work-with silicon. That was easy with semiconductors such as gallium arsenide and indium phosphide, but they are very expensive materials. Silicon lasers put optical computing within reach of ordinary consumers.

Intel has achieved data transfer speeds of 10 gigabits per second in its prototypes and hopes to ramp that up to 100 gigabits per second later this year. “There is no question anymore whether we can do this,” Intel’s lead researcher in this area said. “It’s when and how. That’s been the change in the last year.” When the last “how” has been solved, “silicon photonics will be everywhere,” and Moore’s Law will gain a new lease on life.

Molecular Computing

Source article

University of Alberta researchers have demonstrated the feasibility of making a transistor out of a single hydrocarbon molecule bonded to a silicon crystal substrate. Contrast that with the million or so atoms in a single transistor today. The upside is that silicon is cheap and easy to work with. The downside is that the technique as demonstrated requires a million-dollar electron tunneling microscope to control every transistor. No doubt they are working on it.

Nanomech Memory

Source article

A Dutch company’s “Nanomech” memory stores data using thousands of micron-sized electro-mechanical toggles each representing a 1 or a 0, depending on whether the toggle is up or down. The company says its memory uses 100 times less power than standard electronic memory, works up to 1000 times faster, and is more tolerant of temperature and radiation. Its current prototype stores 256 kilobytes (roughly 2 megabits) of data — enough for the simple microcontrollers common in cars and consumer electronics — but by stacking banks of toggles, the company hopes eventually to raise capacity to several gigabytes, perhaps by the end of 2006.

Next Fastest Supercomputer

Source article

Japan has announced plans to build by 2010 a 10 petaflop (10,000,000,000,000,000, or 10^16, floating point operations per second) supercomputer. It will be 73 times faster than IBM’s 136.8 teraflop Blue Gene/L — today’s world record-holder. The un-named machine will be used for drug development, astrophysics, and weather forecasting.