Image credit: Cerebras Systems
Cerebras Systems and the United Arab Emirates' technology group G42 jointly announced the launch of Condor Galaxy, a network of nine interconnected supercomputers. The network provides a new AI computing method that is expected to significantly reduce the training time of AI models.
The first supercomputer in the network, called the Condor Galaxy 1 (CG-1), has 54 million cores and 4 exaFLOP processing power.
Cerebras and G42 plan to deploy two more such supercomputers, CG-2 and CG-3, in the United States in early 2024. With a total planned capacity of 36 exaFLOP, this unprecedented supercomputing network is expected to drive breakthrough advances in artificial intelligence worldwide.
Located in Santa Clara, California, CG-1 connects 64 Cerebras CS-2 systems into an easy-to-use AI supercomputer with an AI training capability of 4 exaFLOP. The two companies offer CG-1 as a cloud service that simplifies customer access to high-performance AI computing without the need to manage or distribute models on physical systems.
CG-1 accelerates innovation by enabling G42 and its cloud customers to quickly and easily train large, breakthrough models.
A complete rendering of the Condor Galaxy 1 AI supercomputer.
Andrew Feldman, CEO of Cerebras Systems, said: "CG-1 provides AI computing with 4 exaFLOPs on FP 16, greatly reducing AI training time while eliminating the pain of distributed computing. "Many cloud companies have announced massive GPU clusters that cost billions of dollars to build, but are extremely difficult to use.
"Distributing a model across thousands of micro-GPUs takes months for dozens of people with rare expertise. CG-1 eliminates this challenge. Setting up a generative AI model takes minutes instead of months and can be done by a single person. CG-1 is the first of three 4 exaFLOP AI supercomputers deployed across the United States. Next year, we plan to expand this deployment with the G42 and support efficient, purpose-built AI computing with up to 36 exaFLOP. ”
With the Condor Galaxy Supercomputing Network, the two companies are democratizing AI, giving people simple and easy access to industry-leading AI computing. G42's collaboration with different datasets in healthcare, energy, and climate research will enable system users to train new, cutting-edge foundational models.
CG-1 is the first of three 4 exaFLOP AI supercomputers (CG-1, CG-2, and CG-3) built and located in the United States. The three AI supercomputers will be interconnected in a 12 exaFLOP, 162 million core distributed AI supercomputer consisting of 192 Cerebras CS-2 and powered by more than 218,000 high-performance AMD EPYC CPU cores.
G42 and Cerebras plan to launch six more Condor Galaxy supercomputers in 2024, bringing the total computing power to 36 exaFLOP.