The Siberian Ring Photon Source (SKIF) in Russia has been equipped with a 390-teraflop computing cluster, which serves as dedicated high-performance computing infrastructure for one of the nation’s major scientific efforts. The system processes experimental data generated by synchrotron research. This system, developed by the Russian technology company RSC, combines liquid-cooled servers, GPU accelerators, and high-capacity storage.
Russia Expands Its Scientific Computing Infrastructure
Russia has finalized the installation and testing of a dedicated data center at the Siberian Ring Photon Source (SKIF) in Koltsovo, Novosibirsk Region. The computing cluster reportedly provides a performance of 390 teraflops in double-precision FP64 calculations, which is equivalent to 0.39 petaflops. The facility is intended to facilitate the continuous collection, processing, and storage of experimental data produced by SKIF’s research stations.
The Russian RSC Group, which specializes in data-center infrastructure and high-performance computing systems, devised and executed the project. The installation has successfully completed testing and is prepared for operation, as indicated in the SKIF project announcement on November 13, 2026. The computing infrastructure, an essential component of the broader scientific system, is currently under development as a fourth-generation synchrotron radiation facility.
The installation is significant because modern scientific facilities require a dedicated computing infrastructure to process experimental measurements. Advanced detectors can generate a significant amount of information, necessitating computing resources that can manage these datasets without disrupting ongoing experiments.
Inside Russia’s 390-TFLOPS Computing Cluster
The SKIF computing infrastructure comprises 24 RSC Tornado servers, which collectively generate 182 teraflops of performance, and two additional servers that are outfitted with GPU accelerators. The combined FP64 performance is reported to be 390 teraflops, with the GPUs contributing an additional 208 teraflops.
The system’s liquid-cooling architecture is a critical component. The 24 RSC Tornado servers employ a cooling solution that RSC has developed. This solution is designed to manage the thermal load generated by continuous computing activity and support reliable operation. Server and networking equipment, data-storage infrastructure, environmental control, and an uninterruptible power supply with a capacity of 300 kW are also included in the installation.
Storage is another critical component of the system. The data center is equipped with 900 terabytes of storage for warm data and an additional 368 terabytes of distributed hot-data storage. Its network infrastructure is capable of transmitting experimental data from detector systems to computing servers at a rate of up to 200 Gbps.
This combination of storage, networking, and computing is tailored to the needs of scientific investigations, as opposed to traditional office or enterprise workloads. Technical report transmitted via Interfax.
Why Does a Synchrotron Need a Supercomputer?
The computing cluster is being developed to provide support for SKIF, an important scientific installation that is intended to generate synchrotron radiation of high intensity. The facility generates radiation that researchers can utilize to investigate the internal structure and behavior of various materials by accelerating electrons to an energy of 3 gigaelectronvolts (GeV).
Synchrotron radiation allows scientists to examine materials at extremely small scales. It is used in the study of biological molecules, the analysis of chemical reactions, the development of new functional materials, and the study of crystal atomic structure. Pharmaceutical development, microelectronics, energy technologies, and industrial manufacturing can also benefit from this research.
When fully developed, SKIF is intended to house 30 experimental stations. The computing cluster is anticipated to manage the data produced by the initial group of stations, as the first-stage infrastructure is introduced gradually.
When considering the volume of information generated by individual experiments, the necessity for significant computational capacity becomes more apparent. The current equipment can generate approximately five terabytes of raw experimental data in a single day using phase-contrast tomography, as stated in the SKIF project announcement.
A Flexible Computing Architecture for Different Experiments
The resource-management architecture of the SKIF computational cluster is one of its most critical features. The system can dynamically allocate processor cores, GPU accelerators, memory, and high-speed storage to individual research groups in accordance with the needs of each experiment, rather than designating fixed computing resources per group.
This enables researchers to customize virtual computing environments to accommodate particular scientific objectives. The allocated resources can be released and reassigned to other users after an experiment has been completed. Experimental data can be transferred from high-speed storage to the primary storage infrastructure for longer-term retention.
This arrangement is especially important due to the fact that the computing requirements of various experiments vary. While some individuals generate relatively small amounts of data but necessitate a significant amount of processing capacity, others generate large datasets that necessitate a greater amount of storage and network infrastructure.
The project also includes connecting SKIF’s computing cluster with supercomputers supporting other Russian megascience facilities through the National Research Computer Network (NIKS). This integration could enable scientific organizations to coordinate data-processing activities and share computing resources across multiple research installations. This section describes the technical aspects of RSC and SKIF.
How Does It Compare With Russia’s and the World’s Supercomputers?
At 390 teraflops, SKIF is not designed to compete with the world’s largest supercomputers. El Capitan and Frontier, which are American systems, operate at the exascale level, achieving performance measured in thousands of petaflops. Additionally, the reported performance of Russia’s largest publicly benchmarked commercial computing systems, such as Chervonenkis from Yandex and Christofari Neo from Sber, is significantly higher.
Nevertheless, it is necessary to qualify these comparisons. The SKIF figure is a representation of the reported FP64 performance, whereas international supercomputer rankings typically utilize measured High Performance Linpack (HPL) results. The two measurements are not directly interchangeable.
The SKIF installation is located in a different location within Russia in comparison to large commercial AI and general-purpose computing clusters. The architecture and resource allocation of the system are specifically tailored to meet the needs of synchrotron experiments. Consequently, its success will be contingent upon the efficiency with which it manages experimental data, provides assistance to scientific users, and ensures continuous operation.
Building Domestic Capabilities in High-Performance Computing
The initiative also illustrates the role of Russian companies in the development of specialized computing infrastructure for national research facilities. RSC has been responsible for the design and implementation of the cluster, which includes the integration of computing, storage, and networking resources, as well as its liquid-cooling solution.
However, domestic system integration should not automatically be interpreted as complete domestic hardware independence. The published project specifications do not guarantee that each processor, GPU, storage component, and networking device is manufactured in Russia. Nevertheless, the acquisition of the necessary skills to design, integrate, chill, and operate specialized high-performance computing systems is a critical component of the development of national technological capabilities.
The scientific infrastructure of Russia is to be strengthened by the broader SKIF initiative. It is expected that its experimental facilities will facilitate research in physics, chemistry, biology, geology, and materials science, while also providing industrial organizations with the opportunity to utilize advanced synchrotron techniques.