NVIDIA H200 NVL 141GB
The NVIDIA H200 NVL is a powerful AI accelerator for professional data center and server environments. Based on the NVIDIA Hopper architecture, it combines 141 GB of high-performance HBM3e memory with 4.8 TB/s of memory bandwidth and is specifically designed for large AI models, generative AI, large language models (LLMs) and high-performance computing.
Its exceptionally fast HBM3e memory enables efficient processing of large models and datasets. This makes the H200 NVL suitable for memory-intensive inference workloads, scientific computing and demanding AI infrastructure.
Technical Highlights
- GPU architecture: NVIDIA Hopper
- GPU memory: 141 GB HBM3e
- Memory bandwidth: 4.8 TB/s
- FP32 performance: 60 TFLOPS
- FP16/BF16 Tensor Core performance: Up to 1,671 TFLOPS with sparsity
- FP8 Tensor Core performance: Up to 3,341 TFLOPS with sparsity
- Interface: PCI Express 5.0
- GPU interconnect: NVIDIA NVLink, up to 900 GB/s per GPU
- Multi-Instance GPU: Up to 7 MIG instances
- Maximum TDP: Up to 600 W, configurable
- Form factor: PCIe, dual-slot
- Cooling: Passive, requires forced server airflow
High Memory Performance for Large AI Models
With 141 GB of HBM3e memory and 4.8 TB/s of memory bandwidth, the H200 NVL provides a powerful foundation for memory-intensive AI applications. The combination of high memory capacity and bandwidth can offer significant advantages for large language model inference, data-intensive machine learning workloads and scientific simulations.
Scalable Server Configurations
The H200 NVL supports NVIDIA NVLink for high-performance connections between compatible GPUs, enabling suitable multi-GPU server configurations. Actual support depends on the server platform, NVLink components and individual system configuration.
Important Cooling and System Compatibility Notice
The NVIDIA H200 NVL does not have integrated fans! This accelerator requires a compatible server system providing sufficient forced airflow to cool the GPU.
Operation in conventional desktop PCs or workstations not specifically designed for this hardware is not intended. Server compatibility, power supply, cooling capacity and available installation space must be verified before installation.