Crystal Group FG2 3U GPU server

Equipped with up to two dual-slot GPUs, Crystal Group’s FG2 3U GPU server brings versatile, scalable compute power for demanding AI and high-density graphics applications into the field. A low-latency path between GPUs with a large amount of bandwidth deliver real-time compute capabilities by meshing the GPU engines for optimal data reconstruction.

The system is designed to deliver acceleration for data processing, machine learning, AI training and AI inference. Weighing less than 16.8 kg, the compact, rugged server tackles graphic-intense workloads demanding more than 20K CUDA cores and thousands of RT cores. Mission-critical applications like real-time sensor fusion, video processing and object recognition require A100 or A30 GPUs, while ray tracing platforms can use tensor cores in either the A40 or A10.

The integrated system includes one or two 3rd Gen Intel Xeon Scalable or AMD processors delivering the latest PCIe Gen4 interconnect bus. It can also be configured with ultrafast, low-latency data processing units (DPUs). Complementing GPUs with DPUs allows users to achieve high levels of data and application security, data centre and cloud application performance, and ultralow latency and lossless networking, in an edge appliance.

A dedicated air plenum and custom heat sinks optimise thermal performance for seamless, real-time operation under high computational loading. The overall configuration of the certifiable system is validated for performance, manageability, security and scalability. Engineered to meet strict MIL standards, accelerate compute-intensive workloads and provide advanced thermal management, the server is suitable for use in demanding and unpredictable situations.