A domestic neural network accelerator with record performance was presented in Moscow. The development was brought to market by the company "HighTech" - this is the first Russian platform of this level, built on its own microprocessor architecture.
Peak performance reaches 960 TOPS - this is the maximum figure among Russian solutions and a level comparable to leading global analogues. A key feature is the ability to simultaneously process more than 100 neural network models, which makes the system effective for data centers and complex computing tasks.
The solution is already being implemented in practical scenarios: from medicine, where neural networks analyze CT, MRI and genomic data, to robotics, unmanned transport and security systems. The accelerator is also designed for use in embedded computing systems and automated complexes.
The development strengthens Russia's position in the field of artificial intelligence and high-performance computing, opening up opportunities for large-scale digital transformation of key industries.
Read more on the topic:
-
Russian neural networks will not be able to develop outside of competition with the global market, according to analyst Murtazin -
The Ministry of Digital Development has published a draft law on sovereign AI: neural networks must be created in Russia and trained on Russian data -
Russian Scientists to Get Free Domestic Photonic Chips from Bauman Moscow State Technical University
