Intel Says Welcome to the 'Era of Tera'

The vendor boasts the research chip is the first to offer tera-scale performance from a single, energy-efficient processor.

February 15, 2007

2 Min Read
Network Computing logo

Intel is unveiling a new programmable multi-core processor this week the vendor says marks the beginning of 'Tera-scale' computing. Intel claims the compact 80-core chip is the first to deliver the trillions-of-calculations-per-second performance that will be used in future generations of PCs and servers. Intel is outlining details of the research chip this week at the Integrated Solid State Circuits Conference (ISSCC) in San Francisco.

The company says by providing teraflop performance, and the capability to transfer terabytes of data while using less electricity than the average home appliance, Tera-scale technology has the potential to power the deployment of innovative future applications for a variety of purposes including education, collaboration, artificial intelligence, data mining and entertainment. Intel won't sell this particular processor with floating point cores, but instead says the chip provides its researchers with invaluable data with respect to specialized processor and core functions as well as how these super high-speed processors would work both with other chips and with the computers themselves in a production environment.This particular chip is roughly the size of a fingernail. By comparison, Intel says the first supercomputer to attain teraflop performance'-the ASCI Red Supercomputer, in 1996--occupied more than 2000 square feet, ran on close to 10,000 Pentium Pro processors, and needed over 500 kilowatts of energy. Intel says this single processor runs on just 62 watts of electricity.

Intel says the chip relies on tile architecture to make it easier to put multiple cores in a single, compact chip. The processor uses a mesh-like "network-on-a-chip" design to handle bandwidth-intensive communications between the cores. This design also makes it possible to move terabits of data per second inside the chip. The vendor says ongoing Tera-scale research is concentrating on things such as adding 3-D stacked memory.

Stay informed! Sign up to get expert advice and insight delivered direct to your inbox

You May Also Like

More Insights