On September 1, the rector of Lomonosov Moscow State University (MSU) Viktor Sadovnichy announced the launch of the supercomputer "MSU-270" with an AI performance of 400 Pflops (the accuracy of the calculations is not specified).
As reported by the press service el salvador whatsapp resource of Moscow State University, the start of testing tasks by the new machine has already been announced. The complex will become part of the united network of scientific supercomputer centers of Russia and will allow the creation of Russian language models similar to ChatGPT.
since 2020 with the participation of the Faculty of Computational Mathematics and Cybernetics of Moscow State University. The system includes about a hundred of the most modern accelerators, TASS clarifies. The machine uses an unnamed 200-Gbps interconnect, which also covers the storage system. A network with a throughput of 100 Gbps is used for control and integration with external infrastructure. In addition, the machine received new engineering systems, and Russian-made units and components were widely used in the creation of the entire complex. "MSU-270" will form a single computing cluster with the previously commissioned supercomputer "Lomonosov-2".
No other technical specifications were given, so we can only guess what accelerators are used in the new system. If we assume that we are talking about 400 Pflops in FP8 calculations (with sparseness), then about a hundred of the latest NVIDIA H100 accelerators in the SXM form factor or, probably, unsanctioned H800 will be enough to achieve this level of performance.
As reported by the TASS agency, citing Sadovnichy, the tasks of "MSU-270" will mainly be related to the development of artificial intelligence (AI) technologies and the training of personnel in this field. According to the press service of MSU, the researchers will be engaged in "developing mathematical methods of machine learning for processing large volumes of scientific text information, intelligent image analysis for high-performance plant phenotyping and precision farming, predicting the quality of heterogeneous channels in data transmission networks based on probabilistic models and machine learning methods" and research in other areas, for example, in the field of materials science.
The construction of MSU-270 is envisaged by the MSU development program until 2030. According to Igor Sokolov, Dean of the Faculty of Computational Mathematics and Cybernetics at MSU, MSU-270 has already begun to perform the first test tasks, and the results of the work will be summarized in December. In particular, they are related to image analysis and medical research, and in the future, the model may help speed up the emergence of a means for more effective control of the attention of younger students in class. The supercomputer will be used to support courses in the field of AI, develop master's programs, automate and digitalize the educational process, and hold hackathon competitions.
The supercomputer has reportedly been in development
-
- Posts: 477
- Joined: Thu Jan 02, 2025 7:23 am