The world’s first supercomputer built with processors from a domestic microarchitecture known as Leonard Euler, nicknamed Teragraph, was developed at Moscow State Technical University (MSTU) NE Bauman. This achievement was reported by the Ministry of Science and Higher Education of the Russian Federation, with detailed commentary from the project’s chief designer, Alexei Popov. Citation: Ministry of Science and Higher Education; project brief by Popov.
The system is designed to process pseudo-graphics. A machine learning researcher, Georgy Polyakov of Kryptonit, explains that graphs represent collections of objects and the connections between them through parameters. In mathematical terms, objects are vertices and the links are edges.
Polyakov offers an example from everyday life: an airline’s intercity flight network can be seen as a graph where airports in cities act as vertices and direct routes between cities are edges.
Valery Andreev, Ph.D., deputy director general for science and development at IVK, discusses how price effects metal values.
Graph-based computation enables discovering relationships in large datasets where connections are not obvious. It can reveal how people are interconnected or how different banking actions influence one another. Establishing strong links between objects and their parameters supports qualitative forecasting. This capability has potential across multiple sectors, including finance, wholesale trade, and forensic medicine.
Alexey Popov, the project’s chief designer, notes that Teragraph’s long-term work with graphics may lead to digital twins of people and experimental treatment modalities. The system could simulate various care options and propose different approaches based on initial input data. This feature is described as the system’s intuition.
The concept is described with a simple illustration: imagine body parameters displayed on a computer screen. A program can monitor the effects of different drugs without risking patient safety. After selecting the most suitable option, the treatment could be applied to the patient through the program’s guidance.
what is the computer made of
Teragraph operates on a Linux-based Intel central processor with three additional Leonard Euler computing modules connected to it. Visually, the configuration resembles video cards linked to a motherboard.
Leonard Euler modules are built with programmable logic chips, while firmware is written for the architecture on spaces left by AMD chips. This arrangement allows the system to run the custom algorithms developed by Russian researchers.
Hardware-wise, Bauman University states that the processors rely on AMD components but execute unique algorithms created by local scientists. A chip cannot be classified as a processor without these underlying algorithms, which are implemented through low-level software specific to the Leonard Euler architecture.
Popov notes that creating a distinct Russian chip based on Leonard Euler is possible if needed, though no immediate plan exists due to funding constraints for producing such chips in Russia.
Each Leonard Euler module contains 24 cores clocked at 200 MHz. Even at this modest speed, the graphics workload can surpass the performance of high-end Intel Xeon servers operating at 3 GHz, according to Popov. He adds that while the architecture excels at graphics processing, it does not match x86 or ARM in traditional arithmetic tasks. Consequently, builders view Leonard Euler as a supplementary component rather than a direct substitute for mainstream central processors.
The strength of Leonard Euler lies in its architecture optimized for graphics, not in raw arithmetic power. In graphic tasks, the processor can handle millions of graphs concurrently rather than processing each graph sequentially, which explains the efficiency advantage.
Other notable benefits include reduced material costs and power consumption. Leonard Euler chips require dramatically less silicon and use far less power than a single Intel Xeon family processor. This makes them attractive for future use in Internet of Things devices.
Graphics issues
Despite progress at the Bauman research group, practical deployment faces challenges. Popov highlights the core difficulty: the system can operate only when initial information is presented as graph data arrays, while most data in databases is organized in tabular form.
Popov identifies two paths to address this gap. One is to graphically assemble new databases, the other is to translate tabular archives into graphical form. Both routes require time and effort, impacting current IT adoption.
Valery Andreev emphasizes that Bauman’s work remains relevant primarily within research settings rather than immediate industry impact. The IT sector is moving toward shrinking transistor sizes and relying more on conventional computing approaches, but he believes the direction could shift as industry limits are reached.
He adds that the university’s efforts should not be dismissed. Sooner or later, the innovations from Russian engineers may prove essential in broader contexts.
Popov notes that even in the prototype stage, the project has attracted interest from the Moscow Department of Information Technologies and the Center for Systems Biology tied to a national health agency, signaling possible government engagement with the Teragraph concept.
The Teragraph system is currently being prepared for a pilot within Bauman University. It is planned to use graph-based analyses to assess threats to state security, with the pilot projected to begin before the end of 2022.