Delta Sprut

GPGPU platform based on the latest 5th generation Intel® Xeon® Scalable processors with the ability to connect up to 10 NVIDIA H100 accelerators via PCIe Gen5 interface
Delta Sprut

Description

Delta Sprut is a Russian GPGPU platform that enables the creation of comprehensive solutions for artificial intelligence, machine learning, simulation, and 3D VDI infrastructure tasks.
Modular architecture allows scaling the number of accelerators up to 10 FHFL full-size, such as NVIDIA H100, or up to 20 HHHL adapters. 

Specifications

  • Ability to install up to 10 NVIDIA H100 accelerators or up to 20 HHFL adapters, such as NVIDIA L4/A10
  • Increased performance through optional GPU pairing via NVLink
  • High-speed data transfer via modern PCIe Gen5 bus
  • Up to 128 cores Emerald Rapids
  • The maximum RAM capacity of 8TB DDR5 allows for the processing of large amounts of data
  • Up to 8 U2 PCIe 5.0 NVMe drives for hot data storage
  • Extensive monitoring and control functionality in the software Delta BMC and Delta Sokol
Up to 8 FHFL video cards

(NVIDIA A100/H200 или AMD Instinct)

OR
Up to 16 HHHL video cards

(NVIDIA L4/A10)

Appearance

Applications

Artificial intelligence and machine learning

The modular architecture and high level of flexibility of the solution make Delta Sprut an ideal device for building systems with Artificial Intelligence.

HPC

Maximum performance of HPC clusters with the universal Delta Sprut solution that opens up new parallel computing opportunities.

3D VDI

Building a virtual workstation infrastructure for heavy graphics applications based on Delta Sprut.

Use Cases

Modeling
1х Server 1х Delta Sprut
Modeling
AI, machine learning, HPC and modeling
1х Server 6х Delta Sprut (1x switching, up to 5x computing), cascade connection
AI, machine learning, HPC and modeling
AI, machine learning, HPC and modeling
1х Server 1х Delta Firebird 5х Delta Sprut
AI, machine learning, HPC and modeling

Advantages

High density
High density
Capacity for up to 10 NVIDIA H100 accelerators or up to 20 HHFL adapters
Easy maintenance
Easy maintenance
The modular design simplifies access to all its components. No tools needed to maintain
High graphics performance
High graphics performance
Increased performance through optional pairing of GPUs via NVLink

Photo

Any questions left?

Fill out the questionnaire to let us know how we can be useful based on your answers: we will collect information and prepare for the meeting.