A brand new pc chip has been designed to run an synthetic intelligence (AI) supercomputer on a scale by no means seen earlier than.
That’s the declare from the crew behind the Wafer Scale Engine 3 (WSE-3), which is the newest era of a platform designed to run methods equivalent to OpenAI’s ChatGPT.
The chip made by Cerebras is a silicon wafer that measures 21.5cm x 21.5cm and contains 900,000 AI cores. Its predecessor included 2.6 trillion transistors however the brand new model is jam-packed with 4 trillion of them.
Click right here for extra of the newest science information from the Daily Star
Andrew Feldman, the chief government and co-founder of Cerebras, stated: “When we started on this journey eight years ago, everyone said wafer-scale processors were a pipe dream.
“We could not be more proud to be introducing the third-generation of our groundbreaking wafer scale AI chip.
“WSE-3 is the fastest AI chip in the world, purpose-built for the latest cutting-edge AI work … We are thrilled to bring [it] to market to help solve today’s biggest AI challenges.”
(Image: Cerebras)
The new chip makes use of the identical quantity of energy as its earlier incarnation however is twice as highly effective, in accordance with its producer.
To put the brand new chip into perspective, one of the highly effective chips getting used to coach AI fashions in the meanwhile is the Nvidia H200 graphics processing unit (GPU). It has 80 billion transistors, however that determine is 57 occasions lower than the brand new Cerebras one, reviews LiveScience.com
The WSE-3 chip will probably be used to energy the Condor Galaxy 3 supercomputer, which is at present being constructed and will probably be primarily based in Dallas, Texas. The supercomputer can hit eight exaFLOPs of efficiency and is the third set up of the Condor Galaxy community of AI supercomputers.
For the newest breaking information and tales from across the globe from the Daily Star, join our publication by clicking right here.