Discover
/
Article

IBM’s new chip works like a brain

AUG 08, 2014
Physics Today

New York Times : In a paper published yesterday in Science, a team from IBM and Cornell University describes a new computer chip whose architecture is modeled on the human brain. The chip contains 5.4 billion transistors organized into 4096 cores. The cores in turn are connected to each other in a network that integrates 1 million neuron analogues and 256 million synapse analogues. Both the neurons and synapses are reconfigurable. When it comes to number crunching, the brain chip is no match for conventional chips. However, its parallel architecture requires less power and is suited to tasks, such as pattern recognition, that artificial neural networks perform.

Related content
/
Article
The finding that the Saturnian moon may host layers of icy slush instead of a global ocean could change how planetary scientists think about other icy moons as well.
/
Article
/
Article
After a foray into international health and social welfare, she returned to the physical sciences. She is currently at the Moore Foundation.
/
Article
Modeling the shapes of tree branches, neurons, and blood vessels is a thorny problem, but researchers have just discovered that much of the math has already been done.

Get PT in your inbox

pt_newsletter_card_blue.png
PT The Week in Physics

A collection of PT's content from the previous week delivered every Monday.

pt_newsletter_card_darkblue.png
PT New Issue Alert

Be notified about the new issue with links to highlights and the full TOC.

pt_newsletter_card_pink.png
PT Webinars & White Papers

The latest webinars, white papers and other informational resources.

By signing up you agree to allow AIP to send you email newsletters. You further agree to our privacy policy and terms of service.