The supercomputer is designed to solve machine learning problems in Yandex data centers. Chervonenkis runs on NVIDIA A100 graphics accelerators with InfiniBand interconnect based on Mellanox switches.
It is based on the NVIDIA HGX A100 architecture, optimized for machine learning tasks facing Yandex. This allowed us to increase the cluster size and train the largest ML models about twice as fast as with the standard architecture.
Characteristics: number of nodes - 199, power - 21,530 teraflops, cores - 25,472, GPU - 1,592 NVIDIA A100 80G, RAM - 199 TB, power consumption - 583 kW.
Ranked 19th in the world according to the Top500 rating (November 2021).
Other developer projects: