High Performance

Low Power Consumption

Wide Range of Application Scenarios

Advanced AI Acceleration Features

High Cost-Effectiveness

Built specifically for accelerated virtual desktop infrastructure (VDI)

Cost-effective virtual workstations

High-resolution displays

Encoder throughput increased by over 2x

Higher-quality video

PCI Express Gen 4 (PCIe 4.0)

NVIDIA Ampere Architecture

Professional Features

Remote Collaboration

Enterprise-Grade

3D Visualization

AI-Accelerated Applications

Interactive Rendering Technology

Up to 20x higher inference performance

Higher IVA performance at the intelligent edge

Optimized for servers

Industry-leading AI inference performance across cloud, data center, and edge

Enterprise-ready

Mainstream NVIDIA-certified systems

The leap from Ampere to Hopper

4th-Generation Tensor Core and the FP8 Revolution

Confidential Computing and Energy Efficiency Optimization

Performance Improvement and Application Scenarios

Third-Generation RT Cores

Fourth-Generation Tensor Cores

Advanced Video and Visual AI Acceleration

Deep Learning Super Sampling (DLSS)

Virtualization Ready

Data Center Efficiency and Security

Fourth-Generation Tensor Cores

Third-Generation RT Cores

Transformer Engine

Data Center Ready

Memory Technology: Equipped with 141GB HBM3e memory, delivering bandwidth of up to 4.8TB/s.

4th Generation Tensor Cores: Featuring the same 4th-generation Tensor Core technology as the H100, enabling more efficient deep learning and inference.

Secure Multi-Instance GPU (MIG)NVIDIA

4th Generation NVLink, delivering up to 900GB/s of bidirectional bandwidth.

Inference Performance: Delivers up to 2x the inference performance of the H100 when running large language models such as Llama 2 70B.