NVIDIA GPU

NVIDIA Data Center

to of
Page of



NVIDIA H200

Key Features


> 141GB of HBM3e GPU memory

> 4.8TB/s of memory bandwidth

> 4 petaFLOPS of FP8 performance

> 2X LLM inference performance

> 110X HPC performance

NVIDIA H200 DataSheet


NVIDIA H100

Key Features

> 4th Generation Tensor Cores

> 7x faster dynamic programing algorithms

> 3x faster IEEE FP64 and FP32

> New Thread Block Cluster feature

> New Asynchronous Execution

> New Transformer Engine

> HBM3 memory subsystem

> 50 MB L2 cache architecture

> 2nd-Generation Multi-Instance GPU (MIG) technology

> New Confidential Computing support

> 4th-Generation NVIDIA Link®

> 3rd-Generation NVSwitch

> New NVLink Switch System

> Support PCIe Gen 5

NVIDIA H100 DataSheet


NVIDIA L40S

Key Features
> 4th-Generation Tensor Cores
> 3rd-Generation RT Cores
> 2x FP32 CUDA core
> Transformer Engine
> Efficiency and Security
> DLSS 3
> Accelerate Next-Generation Workloads
> Generative AI
> LLM inference
> LLM fine-tuning and small-model training
> NVIDIA Omniverse™ Enterprise
> Rendering and 3D graphics
> Streaming and video content

NVIDIA L40S DataSheet


{"google":["agGridMaterial"],"custom":["Noto Sans KR"]}{"google":["Abel","Oswald","agGridMaterial","Noto Sans"],"custom":["Noto Sans KR"]}