HPE Nvidia H200 NVL 141GB HBM3e PCI-Express 5.0 x16 Graphics Card - P72159-001

The HPE NVIDIA H200 NVL is a next-generation AI and HPC accelerator featuring 141GB of HBM3e memory on a PCIe 5.0 x16 interface. Built on NVIDIA's Hopper architecture, it delivers exceptional memory bandwidth and capacity for large language model (LLM) inference, AI training, and scientific computing workloads in HPE ProLiant server environments.

Key Features

• 141GB HBM3e memory — the largest GPU memory capacity available for LLM inference
• PCIe 5.0 x16 for maximum host-to-GPU bandwidth
• NVIDIA Hopper architecture with Transformer Engine and FP8 precision
• NVLink Bridge support for multi-GPU configurations
• HPE OEM variant optimized for ProLiant server integration

Product Specifications

• Part Number: P72159-001
• Manufacturer: HPE / NVIDIA
• Model: NVIDIA H200 NVL
• Memory: 141GB HBM3e
• Interface: PCIe 5.0 x16
• Architecture: Hopper (GH100)
• TDP: 700W

Processor Specifications

• GPU Architecture: Hopper (GH100)
• CUDA Cores: 16,896
• Tensor Cores: 528 (4th Gen)
• NVLink: Yes — NVLink 4.0 for multi-GPU
• Memory Bandwidth: Up to 4.8 TB/s
• ECC Memory: Yes

Performance Specifications

• FP8 Tensor Performance: Up to 3,958 TOPS
• FP16 Tensor Performance: Up to 1,979 TFLOPS
• Memory Bandwidth: 4.8 TB/s (HBM3e)
• Power Connector: PCIe 5.0 power (server platform supplied)

Use Cases

• Large language model (LLM) inference requiring maximum GPU memory capacity
• Generative AI training and fine-tuning on HPE ProLiant AI servers
• Scientific simulation and HPC workloads requiring high memory bandwidth
• Multi-GPU AI clusters via NVLink for scaled-out model parallelism

Compatibility

• Compatible with HPE ProLiant servers supporting PCIe 5.0 x16 GPU slots and 700W GPU TDP
• Supports NVLink 4.0 for multi-GPU configurations within compatible HPE server platforms
• Managed via NVIDIA Management Library (NVML) and HPE iLO GPU monitoring

Frequently Asked Questions

What is the difference between H200 NVL and H100?

The H200 NVL upgrades from H100's HBM3 to HBM3e memory, increasing capacity from 80GB to 141GB and memory bandwidth from 3.35 TB/s to 4.8 TB/s — critical for LLM inference with large context windows.

What power supply is required?

The H200 NVL has a 700W TDP. HPE ProLiant servers configured for H200 NVL include appropriate power infrastructure — verify server configuration before ordering.

Does it support NVLink for multi-GPU?

Yes — the H200 NVL supports NVLink 4.0 for high-bandwidth GPU-to-GPU communication in multi-GPU server configurations.

The HPE NVIDIA H200 NVL delivers the most advanced AI acceleration available, with 141GB HBM3e and 4.8 TB/s memory bandwidth for the largest LLM and HPC workloads. Order today and power your AI infrastructure with next-generation Hopper performance.