EL-HP

A high-performance fault-tolerant server designed for tasks that require a lot of computing power. Hybrid architecture based on AMD EPYC CPU and NVIDIA A100 GPU allows you to achieve a performance density of 624 teraflop/s on 2U (in tasks using BFLOAT16/FP16 number representation formats).

Applications:

  • Working with Big Data – generating analytical reports, processing disordered sets of large volumes of information.
  • Mathematical modeling and simulations in science, medicine, design – decoding DNA, studying the structure of substances and materials, medicines.
  • High–precision forecasting - making weather forecasts, assessment and search for new mineral deposits.
  • Development of artificial intelligence applications – separate AI programs and robotics.
  • Cryptography – ensuring the confidentiality and integrity of data.
  • Statistical data processing – evaluation and grouping of observations, calculation of averages and indices.
  • Working with complex graphic objects – animation rendering, special effects creation, film production.
  • The presented server solution is easy to install and convenient to use due to the logical arrangement of components and power supplies. Scaling is possible both by installing a larger number of data center servers, and by using more powerful and capacious components.

Specifications

  • Dual AMD EPYC 7002/7003 Series Processors
  • 8TB Registered ECC DDR4 3200MHz, 32 DIMMs
  • 4 PCI-E 4.0 x16, 1 PCI-E 4.0 x8
  • Dual-port 10GbE
  • 4 SATA3 ports
  • 2U Form Factor
  • 4 hot-swap 2.5" drive bays
  • 2200W Redundant Power Supplies
  • 4 hot-swap heavy-duty fans