NVIDIA DGX Station™ A100 brings AI supercomputing to data science teams, offering data center technology without a data center or additional IT infrastructure.
Powerful performance, a fully optimized software stack, and direct access to NVIDIA DGXperts ensure faster time to insights.
PERFORMANCE
Iterate and Innovate Faster
High-performance training accelerates your productivity, which means faster time to insight and faster time to market.
TRAINING
BERT Large Pre-Training Phase 1
Over 3X Faster
DGX Station A100 320GB; Batch Size=64; Mixed Precision; With AMP; Real Data; Sequence Length=128
INFERENCE
BERT Large Inference
Up to 1.25X Higher Throughput for AI Inference
DGX Station A100 320GB; Batch Size=256; INT8 Precision; Synthetic Data; Sequence Length=128, cuDNN 8.0.4
DATA ANALYTICS
ResNet-50 V1.5 Training
Up to 83X Higher Throughput than CPU, 2X Higher Throughput than DGX A100 320GB
DGX Station A100 320GB; Batch Size=192; Mixed Precision; Real Data; cuDNN Version=8.0.4; NCCL Version=2.7.8; NGC MXNet 20.10 Container
实例
Top Use Cases Deployed on DGX Systems
Building Leading Edge AI Across Industries
CASE STUDY
See how customers are using DGX Station
MGH and BWH Uses AI to Improve Radiologist Efficiency
Mass General Hospital (MGH) & Brigham and Women’s Hospital (BWH) Center for Clinical Data Science is using DGX Station to power generative adversarial networks (GANs) that create synthetic brain MRI images, enabling the team to train their neural network with significantly less data. DGX Station serves as a dedicated AI resource to ensure their radiologists can keep moving projects forward.
SBB Uses AI to Safeguard Railway Integrity
Swiss Federal Railway (SBB) has 15,000 trains that provide 1.2 millions rides per day. The power of DGX Station enabled more accurate and automated fault detection in railway tracks and reduced the time for onsite inspections. The optimized AI software in DGX Station lets their engineers focus on gathering the right data instead of testing and configuring components.
Avitas Systems Uses AI for Smarter Inspection Services
Avitas Systems uses AI-powered robots that detect corrosion, leaks, and other defects imperceptible to the human eye with incredible accuracy and go places unfit for humans. They use deep neural networks developed on NVIDIA DGX servers in the data center and easily extended to NVIDIA DGX Station in the field for inferencing on the data closest to where that data is being created.
主要特点
Data Center Performance Anywhere
With DGX Station A100, organizations can provide multiple users with a centralized AI resource for all workloads—training, inference, data analytics—that delivers an immediate on-ramp to NVIDIA DGX™-based infrastructure and works alongside other NVIDIA-Certified Systems. And with Multi-Instance GPU (MIG), it’s possible to allocate up to 28 separate GPU devices to individual users and jobs.
DGX Station A100 is a server-grade AI system that doesn’t require data center power and cooling. It includes four NVIDIA A100 Tensor Core GPUs, a top-of-the-line, server-grade CPU, super-fast NVMe storage, and leading-edge PCIe Gen4 buses, along with remote management so you can manage it like a server.
Designed for today’s agile data science teams working in corporate offices, labs, research facilities, or even from home, DGX Station A100 requires no complicated installation or significant IT infrastructure. Simply plug it into any standard wall outlet to get up and running in minutes and work from anywhere.
NVIDIA DGX Station A100 is the world’s only office-friendly system with four fully interconnected and MIG-capable NVIDIA A100 GPUs, leveraging NVIDIA® NVLink® for running parallel jobs and multiple users without impacting system performance. Train large models using a fully GPU-optimized software stack and up to 320 gigabytes (GB) of GPU memory.
Resources
DGX Station A100 Data Sheet and Whitepaper
