Chipping away at a revolution

Tucked away in a lab in the A.V. Williams building on campus is a four-square-centimeter computer chip. Despite its chaotic appearance - with several exposed wires and countless other microscopic parts - it is a single-chip computer developed by university professors who say it represents the future of computing.

Uzi Vishkin, 54, a professor at the A. James Clark School of Engineering, and his team of 12 completed this prototype known as the Explicit Multi-Threading (XMT) computer in December after nine years of work. It is intended for personal use and able to work 100 times faster than typical desktops.

Vishkin said he has imagined programs that with XMT will be able to simulate a virtual meeting place for friends, organize research points into a well-edited grade-A paper and design custom prescription medication based on an individual's medical history and genetic code.

While typical desktops use serial processing, which allows one operation to be performed at a time, Vishkin's creation uses a process called parallel processing. Vishkin, who is in his nineteenth year as a university professor and works in the Department of Electrical and Computer Engineering, explained why serial processing is inferior to parallel processing.

"Suppose it takes one cleaner 300 minutes to clean your home," Vishkin said. "Parallel processing means that using 100 cleaners, we can clean your home in 3 minutes, reducing the interruption to your life, as well as reducing your cost since you will be paying the same hourly rate."

Although Vishkin has made waves with his prototype, he said it could be three to five years before it reaches the market. The next step is to acquire the funding necessary to build an industry-grade chip to try to sell to a large company, such as Intel, that can mass-produce it, Vishkin said. This chip would be less than two-thirds of a square inch - smaller than the prototype.

"I believe that the work we are doing is truly revolutionary," said George Caragea, a fifth-year doctorate student in computer science who worked with Vishkin on the prototype for three years. "It remains to be seen if it will overcome the barriers set by the big corporations which control the market," Caragea said.

Patrick O'Shea, chairman of the Department of Electrical and Computer Engineering, praised Vishkin's innovation and said that his work represents the motto of the department, "we bring you modern life."

"Professor Vishkin's work is an ideal embodiment of our role as a leading educational institution," O'Shea said. "Here, faculty, staff and students work together on cutting-edge research to produce results that have good and useful outcomes for humanity."

Vishkin has showcased his patented invention, which was given about $2 million in funds by the National Science Foundation and the Department of Defense, at various conventions for computer manufacturers and researchers such as the Association for Computing Machinery's International Conference on Supercomputing in Seattle in June. He also held a demonstration on how to program the computer in local high schools, most recently Montgomery Blair High School.

A Tel Aviv native, Vishkin said he began to develop the necessary theory to program parallel processing in 1979 while earning his doctorate at Technion-Israel Institute of Technology. It took him about 15 years, but he said his steady approach of working with theories before mechanics allowed him to progress.

"Parallel processing requires a very different way of thinking that is different from doing one thing at a time," Vishkin said.

In the 1990s, Vishkin said he discovered that the time for a computer to perform one function, known as clock speed, would not improve in the new millennium because the electronic signals that allow for computations cannot travel faster than the speed of light. Therefore, the only way to improve a computer's productivity was to place many more transistors, which conduct electricity and are fundamental in the operation of all modern electronic devices, on a single chip. A single chip with billions of miniature transistors can utilize parallel processing because many chips can perform the same operation, allowing for a faster performance.

It is important for the next generation of computer engineers to understand how to program a computer that uses parallel processing, Vishkin said. This spring, he is teaching two university courses about his developments and theories: "Towards a New Era of Supercomputing" and "Parallel Algorithmics."

To promote his new technology, Vishkin is holding a contest, open to anyone in the world, to choose a name for the single-chip supercomputer. The deadline for submissions is Sept. 30 and the winner will receive $500.

dbknews@gmail.com