Offered in both four- and eight-GPU configurations to accommodate a wide range of computing requirements.
Products
PRODUCTS: GPU
NVIDIAH200
Supercharges generative AI as the first GPU with HBM3e, H200’s faster, larger memory fuels the acceleration of generative AI and LLMs while advancing scientific computing for HPC workloads.
141 GB
MEMORY
HBM3e per GPU
4.8TB/s
MEMORY BANDWIDTH
HBM3e per GPU
1.6X
GPT-3 175B inference
Performance vs H100
1.9X
Llama2 70B Inference
Performance vs H100
2.0X
HPC Simulation
Performance vs H100
NVIDIA H200 TENSOR CORE GPU
Key Features
NVIDIA H200 TENSOR CORE GPU
TECH SPECS
HBM3e Memory
The H200 is the world’s first GPU to feature HBM3e memory, delivering 4.8TB/s of bandwidth—a 43% boost over the H100—and expanding memory capacity to 141GB, nearly double the H100’s 80GB. This advancement dramatically improves data throughput for generative AI and high-performance computing (HPC) workloads. For large language models (LLMs) like GPT-3, the H200 delivers up to 18X the performance of the A100 and 11X faster performance compared to the H100.
Unmatched Performance
The H200 delivers 32 petaflops of FP8 deep learning performance, making it a powerful solution for the most demanding AI workloads. It also triples the FLOPS of double-precision Tensor Cores, providing 67 teraFLOPS of FP64 computing power for high-performance computing (HPC). For AI-accelerated HPC tasks, the H200 supports TF32 Tensor Core precision, enabling nearly one petaFLOP of single-precision matrix-multiply performance—without requiring any code modifications.
Advanced Architecture
Powered by NVIDIA Hopper™ architecture, the H200 is designed for continuous performance gains through ongoing software updates.
Compatibility
Fully compatible with existing HGX H100 systems, enabling seamless integration and performance upgrades without the need for infrastructure modifications.
Versatile Deployment
Designed for versatile deployment across data center environments, including on-premises, cloud, hybrid-cloud, and edge infrastructures.
GPU Memory
141GB
GPU Memory Bandwidth
4.8TB/s
FP8 Tensor Core Performance
4 PetaFLOPS
Form Factor
SXM | PCIe
Server Options
NVIDIA HGX H200 partner and NVIDIA-certified systems with 4 or 8 GPUs
Nvidia Enterprise 5.0
Included
PRODUCTS: ACCELERATED COMPUTING PLATFORM
NVIDIAHGX200
The NVIDIA HGX H200™️ AI supercomputing platform brings together the full power of NVIDIA GPUs, NVLink®️, NVIDIA networking, and fully optimized AI and high-performance computing (HPC) software stacks to provide the highest application performance and drive the fastest time to insights.
8xH200
GPUs
8x NVIDIA H200 SXM
32 PFLOPS
Tensor Core
FP8/FP6 TENSOR CORE
900 GB/s
Bandwidth
NVSwitch GPU-GPU Bandwidth
7.2TB/s
Agg. Bandwidth
Total aggregate bandwidth
NVIDIA HGX 200 COMPUTING PLATFORM
Key Features
NVIDIA HGX 200 COMPUTING PLATFORM
TECH SPECS
Scalable Configuration
Eight-Way Configuration
An eight-GPU HGX H200 system delivers more than 32 petaflops of FP8 deep learning performance and offers a total of 1.1TB of high-bandwidth memory.
Seamless Integration
Compatible with current HGX H100 systems, enabling effortless and cost-effective upgrades.
NVIDIA NVLink and NVSwitch
High-speed interconnects support the development of powerful scale-up servers, ensuring fast and efficient communication across multiple GPUs.
INT8 Tensor Core
32 POPS
FP16/BFLOAT16 Tensor Core
16 PFLOPS
TF32 Tensor Core
8 PFLOPS
FP32
540 TFLOPS
Nvidia Enterprise 5.0
Included
PRODUCTS: SOFTWARE
NVIDIA AI ENTERPRISE 5.0
NVIDIA®️ AI Enterprise is an all-encompassing, secure AI software platform designed to accelerate the data science pipeline and streamline the development and deployment of AI in production environments. As an end-to-end solution, it offers enterprises a robust, stable, and cloud-native platform packed with over 100 frameworks, pretrained models, and tools, covering a wide range of AI applications including generative AI, computer vision, and speech AI.
5x
Data Processing
Improve data processing time by up to 5 times
4x
Cheaper Operations
Reduces operational costs by 4 times
40x
Acceleration
Accelerate application performance up to 40 times
100+
Frameworks
Includes 100+ frameworks, pre-trained models, and tools
NVIDIA AI ENTERPRISE 5.0 SOFTWARE PLATFORM
Benefits for enterprises
NVIDIA AI ENTERPRISE 5.0 SOFTWARE PLATFORM
Top Use Cases
Secure + Stable Platform
Supports a wide range of AI workloads—including generative AI, computer vision, speech AI, and more—through a secure, stable, cloud-native platform featuring over 100 AI frameworks, pretrained models, and development tools.
Data preparation
Boost data processing speeds by up to 5× and cut operational costs by 4× using the NVIDIA RAPIDS Accelerator for Apache Spark.
AI Training
Build precise custom models in hours—not months—using the NVIDIA TAO Toolkit combined with pretrained models.
Optimization for Inference
Boost application inference performance up to 40× compared to CPU-only platforms using NVIDIA TensorRT.
Deployment at Scale
Streamline and enhance large-scale AI model deployment in production environments with NVIDIA Triton Inference Server.
LLM/Generative AI with Nemo
Extract, transform, and load(ETL)/Data Processing with RAPIDS Accelerator for Apache Spark
Inference with TensorRT, Triton Inference Server and Triton Management Service
Speech and Translation AI with Riva
Cybersecurity with Morpheus
Healthcare with MONAI and Parabricks
