SHARK A.I. 10-GPU Server

Description

SHARK A.I. 10-GPU server – Universal Carrier-Grade Platform

SHARK A.I. is a general purpose carrier-grade platform designed for edge computing applications. By combining a server node, a PCIe expansion with PCIe switch, the DEVKIT can support a combination of up to ten NVIDIA® GPUs depending on the application requirements. It is a configurable edge platform which supports different workloads and requirements. The GPU servers support multiple topologies and bandwidths between GPUs and CPUs with simple cable management settings. In addition, Infiniband support enables easy scaling to multiple GPU clusters.

Framework flexibility for diverse AI applications

The SHARK A.I. supports both single and dual root complexes for different AI applications. For deep learning applications, a single-root complex can use all GPU clusters to focus on large learning processes and the CPU for small tasks. For machine learning, a dual root complex can allocate more tasks to CPUs and arrange less distributed learning between GPUs. The flexible framework of the SHARK A.I. 10-GPU server makes it an extremely flexible AI platform. Enabling a switching option to configure specific PCIe lanes of GPUs for specific I/O and CPU cores improves the overall flow of information to and from multiple virtualized applications. This provides the developer with a wide range of options for configurability and manageability without having to stack systems, consuming valuable space, power and cooling.

Edge High Performance Computing (EHPC)

High performance computing systems for GPU acceleration are a critical enabler for specific solutions that need to optimize high performance computing (HPC) applications and remote virtualization. The SHARK A.I. 10-GPU server can significantly increase flexibility and agility in the cloud domain. It offers the flexibility to implement different head nodes and freely choose the number of GPUs per virtual machine (VM). It is an ideal hardware system that supports a variety of configurations through software implementation.

Reliability and serviceability

The hardware structure of SHARK A.I. is hot-swappable, has redundant fan modules and redundant power supplies. GPU cards can be easily installed after removing the top cover. The SHARK A.I. promotes efficient maintenance while providing optimal performance.