Intel unveils teraflop programmable processor

By Ann Steffora Mutschler -- Electronic News, 2/12/2007

The innovations just don’t stop at Intel Corp. Researchers at the Santa Clara, CA-based chip leader have developed what the company believes is the world’s first programmable processor capable of delivering supercomputer-like performance.

The 80-core chip, not much larger than the size of a fingernail, uses less electricity than most of today’s home appliances, Intel said, which is the result of the company’s tera-scale computing research aimed at delivering teraflops performance, meaning, it can perform trillions of calculations per second. The research focus is aimed at improving performance for future PCs and servers.

Intel expects tera-scale performance and the ability to move terabytes of data to a pivotal role in future computers by powering new applications for education and collaboration, as well as allowing the rise of high-definition entertainment on PCs, servers and handheld devices.

Specifically, the company says applications such as artificial intelligence, instant video communications, photo-realistic games, multimedia data mining and real-time speech recognition could become real.

While Intel’s Jerry Bautista, director of the tera-scale research program says the company has no plans to bring this exact chip designed with floating point cores to market, chips with 20 to 40 cores could hit the market within a decade.

Also, Intel’s tera-scale researcher are investigating individual or specialized processor or core functions, and how software will need to be designed to best leverage multiple processor cores.

This teraflops research chip gave Intel researchers specific insights into silicon design methodologies, high-bandwidth interconnects and energy management approaches, Bautista said.

Intel noted that the first time teraflops performance was achieved in 1996, it occurred on the ASCI Red Supercomputer built by the company for the Sandia National Laboratory. That computer took up more than 2,000 square feet, was powered by nearly 10,000 Pentium Pro processors, and consumed over 500 kilowatts of electricity. Intel says its research chip achieves this same performance on a single, multi-core chip.

Interestingly, this 80-core research chip achieves teraflops performance while consuming just 62 watts, which is less than many single-core processors today.

The chip was designed with a new “tile” approach in which smaller cores are replicated as tiles, aimed at making it easier to design a chip with many cores.

The chip, designed mainly at Intel’s Oregon and Bangalore India locations also contains a mesh-like “network-on-a-chip” architecture meant to allow super-high bandwidth communications between the cores, and capable of moving terabits of data per second inside the chip.

Intel said its research also investigated methods to power cores on and off independently, so only the ones needed to complete a task are used, thus providing more energy efficiency.

Future research will focus on the addition of 3-D stacked memory to the chip as well as developing more sophisticated research prototypes with many general-purpose Intel Architecture-based cores, the company noted.