3-Node H200 GPU Compute Cluster – 24x NVIDIA H200 SXM5, Dual AMD EPYC Per Node
Highlights
 Cluster Nodes: | 3 (8U each, total 24U in 48U rack) |
| GPU Configuration: | 24x NVIDIA H200 SXM5 141GB HBM3e |
| Processor Total: | 6x AMD EPYC 9754 (128-core) |
| Networking: | NVIDIA QM9700 64-port NDR InfiniBand + 6x ConnectX-7 |
Starting Price: Contact for Price
The 3-Node H200X24 GPU Cluster from Hyblox is a fully integrated multi-node AI system engineered to deliver breakthrough performance for large-scale deep learning and high-performance computing. Each of the three 8U servers houses 8x NVIDIA H200 SXM5 GPUs, combining for a total of 24 GPUs across the cluster. With support for 141GB HBM3e memory per GPU, this solution is designed to accelerate large language model training, generative AI, and simulation workloads at data center scale.
Built for Scalable AI and HPC Infrastructure
At the core of each node are dual AMD EPYC 9754 processors with 128 cores, paired with DDR5-6000 ECC memory to provide up to 3TB of memory per node. High-speed PCIe Gen5 NVMe storage ensures rapid data throughput, while NVIDIA ConnectX-7 adapters and a QM9700 InfiniBand NDR switch deliver ultra-low latency interconnects across all nodes. The system is powered by enterprise-grade 3000W Titanium PSUs, with integrated in-row cooling and rack-optimized airflow for maximum reliability.
Key capabilities include:
- Scalable AI GPU cluster with 24x NVIDIA H200 SXM5 GPUs for next-generation training workloads
- Optimized for LLM training, AI inference, and scientific HPC simulations
- Ultra-fast InfiniBand NDR networking with NVIDIA ConnectX-7 for multi-node efficiency
- Enterprise-ready deployment with 48U rack integration, redundant power, and advanced cooling
Compute Components (Total for 3 Nodes)
Component | Specification | Quantity |
GPU | NVIDIA H200 SXM5 141GB HBM3e | 24 |
Server Chassis | Supermicro AS-8125GS-TNHR (8U) | 3 |
CPU | AMD EPYC 9754 (128-core, 2.25GHz) | 6 |
Memory | 128GB DDR5-6000 RDIMM ECC | 72 |
Boot Storage | Samsung PM1743 2TB PCIe Gen5 NVMe U.2 | 6 |
Model Storage | Western Digital SN861 8TB PCIe Gen5 NVMe U.2 | 18 |
Network Infrastructure
Component | Specification | Quantity |
InfiniBand Switch | NVIDIA QM9700 64-port NDR 400Gb/s (MQM9700-NS2F) | 1 |
Network Adapter | NVIDIA ConnectX-7 NDR 400GbE Single-port OSFP | 6 |
IB Cables | MCP4Y10-NO01 1m OSFP Passive DAC | 6 |
Management Switch | 48-port 10GbE managed switch | 1 |
Management Cables | Cat6a cables, various lengths | 10 |
Power Infrastructure
Component | Specification | Quantity |
Rack PDU | Eaton HDX G4 60kW 415V 3-phase | 1 |
Power Supply | 3000W Titanium PSU | 18 |
Power Cables | C19 to C20, 2m | 18 |
UPS System | 50kVA 3-phase UPS with runtime module | 1 |
Environmental & Rack
Component | Specification | Quantity |
Server Rack | 48U standard depth | 1 |
Rack Space Used | 24U (3 servers × 8U) | – |
In-Row Cooling | Vertiv Liebert CRV 45kW | 1 |
Cable Management | Vertical and horizontal organizers | 1 set |
Blanking Panels | Various sizes for airflow optimization | 1 set |
About
Built for Scalable AI and HPC Infrastructure
Key capabilities include:
- Scalable AI GPU cluster with 24x NVIDIA H200 SXM5 GPUs for next-generation training workloads
- Optimized for LLM training, AI inference, and scientific HPC simulations
- Ultra-fast InfiniBand NDR networking with NVIDIA ConnectX-7 for multi-node efficiency
- Enterprise-ready deployment with 48U rack integration, redundant power, and advanced cooling
Specifications
Compute Components (Total for 3 Nodes)
Component | Specification | Quantity |
GPU | NVIDIA H200 SXM5 141GB HBM3e | 24 |
Server Chassis | Supermicro AS-8125GS-TNHR (8U) | 3 |
CPU | AMD EPYC 9754 (128-core, 2.25GHz) | 6 |
Memory | 128GB DDR5-6000 RDIMM ECC | 72 |
Boot Storage | Samsung PM1743 2TB PCIe Gen5 NVMe U.2 | 6 |
Model Storage | Western Digital SN861 8TB PCIe Gen5 NVMe U.2 | 18 |
Network Infrastructure
Component | Specification | Quantity |
InfiniBand Switch | NVIDIA QM9700 64-port NDR 400Gb/s (MQM9700-NS2F) | 1 |
Network Adapter | NVIDIA ConnectX-7 NDR 400GbE Single-port OSFP | 6 |
IB Cables | MCP4Y10-NO01 1m OSFP Passive DAC | 6 |
Management Switch | 48-port 10GbE managed switch | 1 |
Management Cables | Cat6a cables, various lengths | 10 |
Power Infrastructure
Component | Specification | Quantity |
Rack PDU | Eaton HDX G4 60kW 415V 3-phase | 1 |
Power Supply | 3000W Titanium PSU | 18 |
Power Cables | C19 to C20, 2m | 18 |
UPS System | 50kVA 3-phase UPS with runtime module | 1 |
Environmental & Rack
Component | Specification | Quantity |
Server Rack | 48U standard depth | 1 |
Rack Space Used | 24U (3 servers × 8U) | – |
In-Row Cooling | Vertiv Liebert CRV 45kW | 1 |
Cable Management | Vertical and horizontal organizers | 1 set |
Blanking Panels | Various sizes for airflow optimization | 1 set |
