(English) Tech-friendly Russian digital lending giant Tinkoff Bank has developed one of Russia’s most powerful super computers.
Named after the eminent Russian mathematician Andrey Kolmogorov, the machine has been ranked 8th in the Top 50 supercomputers in Russia – at the highest place among participating businesses. It boasts the highest per-node performance in Russia (41.9 TFlop per second), according to the ranking.
As explained by the bank, “fast connections between computer nodes improve the efficiency of hardware resources in distributed training on huge data sets.”
Thus, Kolmogorov offers a “much faster solution to machine learning and AI-related tasks.” These tasks include, in particular, distributed training of neural network models for speech recognition, speech synthesis and processing of natural language; as well as the training of conventional machine learning models for scoring, acquisition and predictive analytics.
The Kolmogorov cluster takes the speed of training neural networks “hundreds of times faster,” the bank claims: “For example, it took us just 24 hours to retrain a sales probability forecasting model on the entire 13-year set of accumulated data as part of our outgoing calls optimization effort. A conventional approach to retraining would have taken us around six months, according to our estimates. The cluster enables business to test hypotheses, improve services and bring new products to the market more quickly and efficiently.”
Tinkoff Bank uses its supercomputer as part of its machine learning and artificial intelligence platform. “The purpose of this supercomputer is to foster a culture of working with data, lower the threshold for entry into this area for our teams and make machine learning accessible for every analyst and developer at Tinkoff,” said the bank’s CIO Vyacheslav Tsyganov.
“We didn’t plan to build a system that would be called ‘super’. It is quite a small part of our infrastructure, but we have reached a performance level that brings us to the top of Russia’s supercomputers,” he added.
Several global banks, as exemplified by JP Morgan and Royal Bank of Scotland, have developed supercomputers; but less than 10 financial institutions – from China and the UK – have their machines featured in the latest global Top 500 supercomputer ranking.
Kolmogorov boasts 658.5 TFLOPS of peak double-precision floating-point (FP64) performance. The system includes 10 nodes with cutting-edge NVIDIA Tesla V100 accelerators, powered by tensor cores delivering exceptional AI performance. The computational nodes of Tinkoff’s supercomputer are connected with advanced 100 Gb RoCE (RDMA over Converged Ethernet) enabled network. Combining the latest technologies, the cluster reached a 418.9 TFLOPS performance in the Linpack test to secure a leading position in the national supercomputer ranking. Kolmogorov uses the same HPC accelerators as the world’s fastest supercomputer Summit (OLCF-4). It is also the most powerful supercomputer among ranking participants in terms of per-node performance. It means that each of its servers is itself a very powerful unit (41.9 TFLOPS).