Вы используете устаревший браузер, который не поддерживает такие современные технологии как HTML5 и CSS3, используемые на этом сайте.
Просьба воспользоваться современным браузером Chrome, Opera, Firefox

The Krylov State Research Centre is equipped with the most powerful supercomputer in Russian shipbuilding with a processing capacity of 56.13 TFlops.

Krylov’s experience gained from numerical simulations for marine engineering applications demonstrates that these techniques make it possible to

  • reduce the test scope required for hull shape optimization and thus to save the cost and time of design efforts,
  • take well-informed decisions regarding the choice of design solutions,
  • improve predictions of full-scale performance,
  • raise the efficiency and, therefore, competitiveness of the prospective ship or marine structure under design.

Supercomputer case studies:

  • Preliminary optimization of hull lines, i.e. calculations to examine a range of hull shape options for obtaining the best or prescribed hydrodynamic characteristics, followed as a rule by manufacturing of ship model and model tests.
  • Calculations of parameters which are impossible to obtain in physical experiment (or when a physical experiment is not practicable).
  • Estimation of scale effects, i.e. comparative calculations for correlation of model and full scale conditions.
  • Computer simulation fully substituting the physical experiment, a rare case at present, when the accuracy of estimated characteristics is not required to be too high (within 10 to 20 %) to meet the design purposes or when the physical experiment provides the same accuracy.

Key numerical methods:

  • Modeling of multi-phase and multi-component flows, including thermodynamic effects;
  • Detached eddy simulation (DES) method for resolving of large scale turbulence;
  • Adaptive mesh refinement for saving the computational resources and maintaining the high accuracy at the same time;
  • Simultaneous fluid-structure interaction (FSI);
  • Flow visualization, including 3D video.

Krylov’s Supercomputer Specifications

Total peak efficiency:

56,12 TFLOPS

Total peak efficiency without GPU

44,42 TFLOPS

Total volume of RAM:

9280 Gb


4 nodes 256 Gb each
& 1 node 2048 Gb

Communication network:

QDR Infiniband

100 nodes – AMD Opteron 6174 200 items (2400 cores)


6400 Gb

45 nodes – Intel Xeon E5-2697v2 90 items (1080 cores)


2880 Gb DDR3
supports ECC 1600 MHz


NVIDIA Kepler K20, 10 items

Operational system:



Developed/Owned by


Foreign software



Dassault Systemes



Fluent, CFX



Siemens PLM Software


Open-source software (GNU General Public License)

OpenCFD Ltd (ESI Group)


In-House Software of Krylov Centre

Glimpse of Future

Background work is under way with a plan to set up an industrial supercomputer centre with a computing cluster of 250 TFlops to address a wider scope of issues at the Krylov Centre.