Gigabyte G593-ZD2-AAX1 8-GPU H100 SXM5 Server SKU

The Gigabyte G593-ZD2-AAX1 is a purpose-built 8-GPU AI and HPC server platform engineered around NVIDIA H100 SXM5 accelerators, delivering a dense, high-throughput compute node designed for the most demanding enterprise AI training, large-scale inference, and scientific simulation workloads. The system integrates eight H100 SXM5 GPUs interconnected via NVLink 4.0 and NVSwitch technology, enabling GPU-to-GPU bandwidth that far exceeds PCIe-based configurations and allowing tightly coupled all-reduce operations essential for distributed deep learning at scale.

Each H100 SXM5 GPU in this platform provides 80 GB of HBM3 memory per accelerator, yielding up to 640 GB of aggregate GPU memory across the node. The Transformer Engine with FP8 precision support accelerates next-generation large language model (LLM) training and inference workloads, while the fourth-generation NVLink fabric delivers up to 900 GB/s bidirectional bandwidth per GPU. The G593-ZD2-AAX1 is built on a dual-socket Intel Xeon Scalable (Sapphire Rapids) processor platform, supporting high-bandwidth DDR5 system memory and multiple PCIe Gen 5.0 expansion paths to feed the GPU subsystem with sufficient CPU-to-GPU data throughput.

Designed for deployment in hyperscale data centers, AI factories, and enterprise GPU clusters, the G593-ZD2-AAX1 conforms to a standard 5U rack form factor and incorporates a high-capacity power and cooling infrastructure to sustain continuous full-load GPU operation. The platform is well-suited for integration into multi-node GPU clusters interconnected via NDR InfiniBand or high-speed Ethernet fabrics, supporting frameworks such as PyTorch, TensorFlow, NVIDIA NeMo, and Megatron-LM. Omnixon Global supplies this SKU to enterprise buyers across the UAE, GCC, EMEA, and APAC regions.

Ideal for

• Large language model (LLM) pre-training and fine-tuning at scale, leveraging all-to-all NVLink 4.0 GPU interconnect for efficient distributed gradient synchronization
• High-throughput generative AI inference serving for enterprise applications requiring low-latency, high-concurrency response from models exceeding 70 billion parameters
• Computational fluid dynamics (CFD), molecular dynamics, and physics-based HPC simulation workloads requiring dense FP64 double-precision throughput
• Enterprise AI pipeline acceleration for computer vision, speech recognition, and multimodal model training across large proprietary datasets
• Multi-node GPU cluster node for AI supercomputing environments interconnected via NDR InfiniBand, enabling workloads that span hundreds to thousands of GPUs
• Sovereign AI and private cloud GPU infrastructure deployment for government, financial services, and healthcare organizations requiring on-premises data sovereignty

Technical specifications

Available from Omnixon Global. Submit an RFQ and our team will confirm configuration and availability for your order.