The world’s fastest new supercomputer with a power of 2 exaflops was introduced + photo

G42 Technology Holding, which is based in Abu Dhabi, UAE, has one of the fastest supercomputers in the world named Condor Galaxy-1 (CG-1) unveiled This computer currently offers a power equivalent to 2 exaflops, but by the end of 2023, its power will reach 4 exaflops. In the future, the total number of these supercomputers that will be connected to each other will reach 9 and increase the power of this system to a stunning number of 36 exaflops.

According to Zdnet, G42 has agreed with the American company Cerebras to build the world’s largest supercomputer. The Condor Galaxy supercomputer takes its name from a cosmic system 212 million light-years away from Earth. The first version of this supercomputer named CG-1, with 32 Computer for artificial intelligence called CS-2 Made up of a total of 27 million cores Computer, 41 terabytes of memory and bandwidth of 194 trillion bits per second.

The current power of this supercomputer 2 exaflops It is considered to be one of the most powerful supercomputers in the world, i.e. Frontier with a power of 1.19 exaflops. However, the system will soon double its configuration by the end of 2023. This means the number of CS-2 computers to 64 number with 54 million cores and will have 82 terabytes of memory and 388 trillion bits per second of bandwidth. CG-1 is made in America.

The world’s largest supercomputer is built in the form of a network

Cerebras, in the fourth phase of its cooperation with G42, which will probably be implemented in the second half of 2024, a system of 9 supercomputers It will activate itself in different countries, each with 4 exaflops of power, to build the largest networked supercomputer for artificial intelligence with a total power of 36 exaflops to present

Due to the growth of artificial intelligence systems, many companies have been busy building large clusters of graphics processors to train these models. But G24 and Cerebras want to provide this processing power through a network of supercomputers to speed up the development of artificial intelligence models.