In Russia, ten new supercomputers with a capacity of 10,000–15,000 GPU H100 can be developed within seven years. About it report “Vedomosti” with reference to the working document of the ANO “Digital Economy”.
The H100 is Nvidia’s latest generation 2022 chip. These chips contain 80 billion transistors, which allows for performance of 56–66 teraflops. The American authorities previously decided to ban the supply of these chips to Russia and China.
To date, several powerful supercomputers have been developed in Russia. Seven of them entered the global top 500. In particular, the 27th place in the world in terms of power was taken by the Yandex supercomputer “Chervonenkis”, which has 1592 nodes with Nvidia A100 (previous generation chip).
According to experts, the development of ten supercomputers with a total number of H100 chips of 15,000 units is an application to get to the top of the world rankings.
The general director of the Promobit company, Maxim Koposov, estimated spending on ten machines with 10,000–15,000 GPUs at $6 billion (500,000 billion rubles).
Earlier it became known that the American video card developer Nvidia intends to release the most powerful supercomputer in Israel. It will allow the development of generative artificial intelligence systems. As a prototype and test bed for the NVIDIA Spectrum-X reference designs, Nvidia is building Israel-1.