NVIDIA H100
GPU architecture NVIDIA Hopper
GPU accelerator optimized for AI, HPC and data centres
14592 NVIDIA CUDA cores for large-scale parallel computing
456 next-generation Tensor cores for AI model training and inference
80 GB of HBM2e memory with ECC to support very large AI models and datasets
Memory bandwidth up to 2 TB/s
Interface PCIe 5.0 x16 for maximum communication bandwidth with the system
Passive cooling designed for servers and GPU clusters
Maximum power consumption: 350 W
Free shipping from €300
Promocja cenowa na model HDR-15-5
Product intended for professional use only
NVIDIA H100
Description
NVIDIA H100 GPU Accelerator for Generative Artificial Intelligence
NVIDIA H100 Tensor Core GPU is one of the most advanced compute accelerators built for modern data centres and artificial intelligence infrastructure. Based on the NVIDIA Hopper architecture, the chip is specifically designed to support the most demanding workloads for training and running next-generation artificial intelligence models.
The H100GPU is the foundation for many of today's AI platforms used in enterprise, data centre and research environments. With its massive computing power and ultra-high memory bandwidth, it enables it to train models with billions of parameters and support complex data analytics systems.
NVIDIA Hopper architecture - next-generation AI acceleration
The NVIDIA Hopper architecture introduces a new generation of Tensor cores and significant improvements in deep learning and data analytics processing. The H100 accelerator features 14592 CUDA cores and 456 Tensor cores to accelerate the matrix operations that underpin modern artificial intelligence models.
With 80GB of HBM2e memory with ECC error correction and bandwidth reaching 2 TB/s, the GPU enables the processing of massive data sets and the training of AI models with a very high number of parameters.
CUDA cores
Tensor cores
of HBM2e memory
memory bandwidth
GPU designed for generative AI models
The NVIDIA H100 is one of the most widely used accelerators in generative AI infrastructure. The GPU enables both the training of massive AI models and their rapid deployment in production environments.
The accelerator is used in systems such as:
training and language model inference
VLM combining text, image and video
AI using enterprise knowledge
big data and enterprise AI
AI-as-a-Service platforms
advanced scientific computing
GPU infrastructure for data centers
The NVIDIA H100 is designed for installation in modern server systems and GPU clusters with very high computational density. The PCIe 5.0 x16 interface provides very high bandwidth communication between the GPU and the server system.
With passive cooling and a maximum power consumption of 350 W, the card is suitable for professional GPU platforms used in data centres and AI infrastructures.
Why the NVIDIA H100 is crucial for AI development
Modern artificial intelligence models require massive computing power and very high memory bandwidth. The NVIDIA H100 was designed with these requirements in mind and is the foundation of AI infrastructures used by technology companies, research institutes and cloud platforms around the world.
With its Hopper architecture and advanced Tensor cores, the H100 GPU enables AI models to significantly reduce training time and increase the performance of AI systems.
Technical Specification
| H100 SXM | H100 NVL | |
|---|---|---|
| FP64 | 34 teraFLOPS | 30 teraFLOPS |
| FP64 Tensor Core | 67 teraFLOPS | 60 teraFLOPS |
| FP32 | 67 teraFLOPS | 60 teraFLOPS |
| TF32 Tensor Core* | 989 teraFLOPS | 835 teraFLOPS |
| BFLOAT16 Tensor Core* | 1,979 teraFLOPS | 1,671 teraFLOPS |
| FP16 Tensor Core* | 1,979 teraFLOPS | 1,671 teraFLOPS |
| FP8 Tensor Core* | 3,958 teraFLOPS | 3,341 teraFLOPS |
| INT8 Tensor Core* | 3,958 TOPS | 3,341 TOPS |
| GPU Memory | 80 GB | 94 GB |
| GPU Memory Bandwidth | 3.35 TB/s | 3.9 TB/s |
| Decoders | 7 NVDEC, 7 JPEG | 7 NVDEC, 7 JPEG |
| Max Thermal Design Power (TDP) | Up to 700 W (configurable) | 350-400 W (configurable) |
| Multi-Instance GPUs | Up to 7 MIGs @ 10GB each | Up to 7 MIGs @ 12GB each |
| Form Factor | SXM | PCIe dual-slot air-cooled |
| Interconnect | NVIDIA NVLink™: 900 GB/s, PCIe Gen5: 128 GB/s | NVIDIA NVLink: 600 GB/s, PCIe Gen5: 128 GB/s |
| Server Options | NVIDIA HGX H100, Partner and NVIDIA-Certified Systems™ with 4 or 8 GPUs | NVIDIA DGX H100 with 8 GPUs, Partner and NVIDIA-Certified Systems with 1–8 GPUs |
| NVIDIA Enterprise Add-on | Included | Included |
| Note | *With sparsity | |
