world’s fastest
The power of supercomputers is usually determined by flops (FLOPS, FLoating-point Operations per Second), that is, the number of operations per second on floating point numbers. The first supercomputers had a performance of approximately 1 flop, that is, 1 thousand operations per second. In 1964, a computer with a performance of 1 million flops (1 megaflop) appeared in the United States. Now the fastest supercomputer in the world is Frontier, whose peak performance exceeds 1.6 exaflops, that is, quintillion flops.
Russia is one of the few countries in the world that has its own supercomputers and is part of the global world. top 500 according to their strength. The list includes Chervonenkis (Yandex, 36th place in the world), Galushkin (Yandex, 58th place), Lyapunov (Yandex, 64th place), Cristofari Neo “(Sber, 67th place), “Christofari” (Sber, 119th place). , “Lomonosov-2” (MSU, 370th place) and GROM (MTS, 433rd place).The creation and maintenance of such systems is an indicator of the work of Russian IT companies at the highest level, and almost no one will argue with this.
What do these super machines do? There are many options for using them. Powerful supercomputers have the potential to revolutionize each of these fields: financial analysis, engineering calculations, medicine, energy, government and defense. Therefore, developing a domestic production base to create machines with extraordinary computing power is one of the highest priorities.
Another sector whose development is directly related to supercomputers is artificial intelligence (AI). The prospects and importance of this field are that the artificial intelligence market in Russia will increase by 18% on an annual basis in 2022 and reached 650 billion ruble mark. Interest in technology is determined by the tangible economic impact its use provides. Thanks to the use of these technologies, the operating costs of companies have decreased by 400 billion rubles by 2023, and by 2025 this amount will exceed 1 trillion rubles.
Advances in artificial intelligence require High Performance Computing (HPC) – supercomputers. Global HPC market in 2023 predictions Mordor Intelligence will exceed $56 billion and grow at an average of 11% per year until 2028.
According to academician Dmitry Novikov, director of the Institute of Control Problems of the Russian Academy of Sciences, one of the leading experts in the field of neural networks and artificial intelligence theory, artificial neural networks are successfully used in many fields. It also highlights that the hype around the term artificial intelligence will likely harm the development of technologies in this field.
“The real danger here is civilization and education. In fact, if we adopt a completely model-free approach, then why do we need to study physics, chemistry, biology and even mathematics in school? Instead, as young people now say, we will “create a grid” and they will figure it out on their own. As a result, an extremely worrying structure of knowledge and competence in this field is emerging,” warns Novikov.
Technologies of the future
The academic said that the institute he heads recently organized a hackathon with one of the banks to solve the problems created by the use of artificial intelligence in agriculture.
“Hackathon means that various problems are proposed and the competing teams must solve them within a certain period of time, after which the jury determines the winners based on pre-determined criteria. At first I was very surprised when my colleagues told me who won there, because when solving one of the problems, a team consisting of two high school students and one student won, although the teams were very different from schoolchildren. It is aimed at individuals with high academic degrees. It seems that the professor must beat the student in an intellectual task. But this is not always the case,” the academic said.
Neural networks are becoming a more common part of the lives of ordinary users and professionals every day, and the amount of information that needs to be processed is also increasing. In order to quickly solve existing problems, the power of supercomputers needs to be further increased. stated Russian President Vladimir Putin at the Artificial Intelligence Journey conference dedicated to the development of artificial intelligence.
The development of new supercomputers in Russia is an ongoing process. Thus, testing of a new supercomputer for the Mathematical Institute of SB RAS in Novosibirsk has been completed. Total cluster performance is 54.4 teraflops (54.4 trillion operations per second). will provide institute employees to solve complex research problems in the field of mathematics, physics, biology and other sciences.
Moreover, by 2030, 10 new supercomputers may appear in Russia with many times more power than most similar devices in the Russian Federation. This is stated in the working paper of the Digital Economy ANO, compiled following the results of the E+Data conference that took place on 11 September.
As can be seen from the presentation of Digital Economy ANO, supercomputers were proposed to be created by the Trusted Infrastructure subgroup, headed by Rostelecom Deputy Chairman Boris Glazkov.