Gigabyte G593-ZD3-AAX1 NVIDIA H100 SXM5 8-GPU Server

The Gigabyte G593-ZD3-AAX1 is a high-density 5U server platform engineered around eight NVIDIA H100 SXM5 GPUs, delivering the full NVLink-interconnected compute fabric required for the most demanding AI training, large language model inference, and high-performance computing workloads at scale. Each H100 SXM5 GPU is built on NVIDIA's Hopper architecture with a 4th-generation Transformer Engine and 3rd-generation NVLink, enabling the server to aggregate up to 3.2 TB/s of GPU-to-GPU bandwidth across the NVSwitch fabric and sustain peak FP8 sparse tensor throughput of approximately 32 petaFLOPS across the full 8-GPU configuration.

The G593-ZD3-AAX1 pairs its GPU subsystem with dual 4th-generation Intel Xeon Scalable (Sapphire Rapids) processors and supports high-capacity DDR5 system memory, providing ample CPU-side headroom for data preprocessing, orchestration, and mixed CPU/GPU pipeline execution. The platform incorporates a high-bandwidth NVMe storage subsystem and multiple PCIe Gen 5 expansion slots for network adapters such as NVIDIA ConnectX-7 InfiniBand or high-speed Ethernet, allowing direct integration into GPU-to-GPU fabric topologies within modern AI supercomputing clusters.

Designed for enterprise datacenters, cloud service providers, and national AI research institutions, the G593-ZD3-AAX1 supports liquid cooling configurations consistent with the thermal envelope of the H100 SXM5 module, enabling sustained full-rack density deployments. The system is positioned within Gigabyte's enterprise server portfolio as a purpose-built AI compute node suitable for deployment in NVIDIA DGX SuperPOD-compatible and third-party GPU cluster architectures serving buyers across UAE, GCC, EMEA, and APAC regions through Omnixon Global.

Ideal for

• Large language model (LLM) pre-training and fine-tuning at billion-to-trillion parameter scale using distributed tensor and pipeline parallelism
• High-throughput generative AI inference serving for production NLP, image generation, and multimodal applications requiring sustained low-latency GPU compute
• Computational fluid dynamics, molecular dynamics, and climate simulation workloads leveraging FP64 double-precision HPC performance on Hopper GPUs
• Deep learning recommender system training for large-scale e-commerce, streaming, and ad-tech platforms requiring high memory bandwidth and NVLink interconnect
• Seismic imaging and reservoir simulation in energy sector HPC environments demanding both FP64 precision and high inter-GPU bandwidth
• Sovereign AI and national research cloud infrastructure deployments requiring dense, rack-optimized GPU nodes with enterprise-grade manageability

Technical specifications

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