--- title: "SKIF Synchrotron Receives Domestic Computing Cluster with 390 TFLOPS Performance" description: "The SKIF User Center has completed the installation and testing of a computing complex based on RSC Tornado servers and GPU nodes: the system offers 390 TFLOPS of performance to process data from seven synchrotron stations." date: 2026-11-12T10:58:23.028Z lang: en url: https://xab.info/en/posts/skif-synchrotron-rsc-cluster-390-tflops tags: [skif-synchrotron, rsc-tornado, supercomputer, data-center, russia-science] author: "Alexander Buran" publisher: "XAB.info" --- # SKIF Synchrotron Receives Domestic Computing Cluster with 390 TFLOPS Performance ![Metal server rack with perforated doors at the SKIF data center facility](https://xab.info/media/2026/11/12/skif-synchrotron-rsc-cluster-390-tflops/skif-synchrotron-rsc-cluster-390-tflops-1.webp) ## Key Points - SKIF has completed the installation of a computing complex based on domestic RSC Group hardware. - The system offers 390 TFLOPS: 182 TFLOPS from 24 liquid-cooled RSC Tornado servers and 208 TFLOPS from two GPU servers. - The data center will process data from seven first-stage stations, including high-speed imaging at up to 1 billion frames per second. - The complex is designed to support the full scientific cycle for 2,000–2,500 researchers per year. At the Siberian Ring Source of Photons (SKIF) User Center, the installation and testing of a computing complex based on domestic hardware from the RSC Group have been completed. The stated performance of the new system reaches 390 TFLOPS, providing the computing infrastructure required to process data streams generated by the synchrotron’s experimental stations. ### Cluster Composition and Performance The SKIF data center has been equipped with 24 "RSC Tornado" servers featuring a liquid cooling system; their total performance is 182 TFLOPS. Additionally, two GPU servers have been integrated, contributing another 208 TFLOPS. The National Computing Complex (NCC) also includes the necessary network nodes and a data storage system, although their exact configuration is not specified in open materials. The total capacity of the CPU and GPU segments forms the claimed 390 TFLOPS. ### Data Processed by the Data Center The new computing center is designed to store and process experiment results from the seven first-stage stations of the synchrotron. These include modules for "Microfocus," "Structural Diagnostics," "Fast-Running Processes," "XAFS Spectroscopy and Magnetic Dichroism," "High-Energy X-ray Diagnostics," "Electronic Structure," and a station for "Basic Methods of Synchrotron Diagnostics for Educational, Research, and Innovation Student Activities." The "Fast-Running Processes" station is particularly computationally demanding: it is capable of X-ray imaging at up to 1 billion frames per second, which requires instantaneous stream buffering and subsequent 3D model reconstruction. ### Scale of Scientific Tasks The capacity of the new data center is estimated to cover the full cycle of scientific work for 2,000–2,500 researchers annually who will conduct experiments at the synchrotron. SKIF itself is a circular accelerator with a 158-meter perimeter, where electron beams rotating at near-light speed reach energies of 3 GeV. These accelerator parameters and the nature of the emitted radiation dictate the volume and speed of data flow for which the computing complex was designed. ### Significance for Domestic Science The completion of the installation and testing of the RSC-based complex means that Russia's flagship synchrotron has acquired its own computing infrastructure using domestic server hardware. This allows for a complete cycle — from signal registration at the experimental station to processing and storage — on a proprietary hardware base, which is important considering the requirements for import substitution in critical scientific infrastructure. ## Fact-Check & Sources - [Сибирский синхротрон «СКИФ» получил НРС-кластер РСК с производительностью 390 Тфлопс](https://3dnews.ru/1151196) - Source: Cluster composition (24 RSC Tornado servers, 182 TFLOPS; 2 GPU servers, 208 TFLOPS), seven stations, accelerator parameters (158 m, 3 GeV), 2,000–2,500 researchers. - [Сибирский синхротрон «СКИФ» получил НРС-кластер РСК с производительностью 390 Тфлопс](https://servernews.ru/1151196) - Matches the primary source regarding cluster composition and performance. - [Путин открыл СКИФ: что сможет уникальный российский синхротрон](https://fedpress.ru/article/3447548) - Contextual source on synchrotron purpose; excluded from the article to avoid duplication and dating issues. ## FAQ ### What is the performance of the new SKIF computing complex? The system's stated performance is 390 TFLOPS: 182 TFLOPS from 24 RSC Tornado servers and 208 TFLOPS from two GPU servers. ### What hardware was used in the cluster? The complex is built on domestic RSC Group hardware: 24 liquid-cooled RSC Tornado servers, two GPU servers, and undisclosed network nodes and storage systems. ### What data will the data center process? The data center will handle experiments from seven first-stage synchrotron stations, including the 'Fast-Running Processes' station capable of capturing up to 1 billion frames per second. ### How many researchers is the complex designed for? The new data center's capacity is designed to support the full scientific cycle for 2,000–2,500 researchers per year.