NVIDIA DGX
NVIDIA DGX STATION A100
Workgroup Alliance for the Age of AI
The NVIDIA DGX Station A100 brings AI supercomputing to data science teams, delivering data center technology without the need to build out data centers or IT infrastructure. Designed to allow multiple users to connect simultaneously, DGX Station A100 utilizes server-grade components in an office-friendly form factor. It's the only system with four fully interconnected Multi-Instance GPU (MIG)-enabled NVIDIA A100 Tensor Core GPUs and up to 320GB of total GPU memory that plugs into any standard power outlet, A powerful AI appliance that can be placed anywhere.
Product spec
|
NVIDIA DGX Station A100 320GB |
NVIDIA DGX Station A100 160GB |
GPUs |
4x NVIDIA A100 80GB GPU |
4x NVIDIA A100 40GB GPUs |
GPU memory |
Total 320GB |
Total 160GB |
Performance |
2.5 petaFLOPS AI 5petaOPS INT8 |
|
System power consumption |
1.5kW (at 100-120Vac) |
|
CPU |
Single AMD 7742, 64 cores, 2.25GHz (base) ~3.4GHz (maximum boost) |
|
system memory |
512GB DDR4 |
|
Network |
Dual port 10Gbase-T Ethernet LAN Single port 1Gbase-T Ethernet BMC management port |
|
storage |
4GB GPU memory, 4x Mini DisplayPort |
|
System acoustic characteristics |
less than 37dB |
|
software |
Ubuntu Linux OS |
|
system weight |
43.1kg (91.0lbs) |
|
System package weight |
57.93kg (127.7lbs) |
|
system size |
Height: 639mm (25.1in) Width: 256mm (10.1in) Length: 518mm (20.4in) |
|
Operating temperature range |
5-35ºC (41-95ºF) |
NVIDIA DGX A100
The world's first AI system built on NVIDIA A100
NVIDIA DGX ™ A100 is a universal system for all AI workloads for training, inference and analytics, delivering unprecedented compute density, performance and flexibility in the world's first 5 petaflops AI system. NVIDIA DGX A100 features the world's most advanced accelerator, the NVIDIA A100 Tensor Core GPU, delivering 20x the performance of VOLTA generation GPU systems.
Product spec
|
NVIDIA DGX A100 (640GB) |
NVIDIA DGX A100 (320GB) |
GPUs |
NVIDIA A100 8x80GB GPUs |
NVIDIA A100 8 x 40GB GPUs |
GPU memory |
640 GB total |
320 GB total |
Arithmetic performance |
5 petaFLOPS in AI 10 petaOPS on INT8 |
|
NVSwitch |
6 |
|
CPU |
Dual AMD Rome 7742,128 cores total, 2.25GHz(base), 3.4GHz (max boost) |
|
system memory |
2TB |
1TB |
Network |
single port Mellanox ConnectX-6 VPI 200Gb/s HDR InfiniBand x 8 dual port Mellanox ConnectX-6 VPI 10/25/50/100/200Gb/ Second Ethernet x 2 |
single port Mellanox ConnectX-6 8 VPIs 200Gb/s HDR InfiniBand dual port Mellanox ConnectX-6 1 VPI 10/25/50/100/200Gb/ seconds Ethernet |
storage |
OS: 1.92TB M.2 NVME 2 drives Internal storage: 30TB (3.84TB x 8) U.2 NVMe drives |
OS: 1.92TB M.2 NVME 2 drives Internal storage: 15TB (3.84TB x 4) U.2 NVME drives |
software |
Ubuntu Linux OS others: Red Hat Enterprise Linux Cent OS |
|
weight |
123.16 kg (maximum) |
|
packing weight |
163.16 kg (maximum) |
|
size |
Overall height: 264.0mm Width: 482.3 mm Depth: 897.1mm |
|
Operating temperature range |
5ºC to 30ºC |
NVIDIA DGX-2
The world's most powerful AI system to tackle complex AI challenges
The world's first 2-petaflops system, NVIDIA® DGX-2™, delivers 10x deep learning performance with 16 fully interconnected GPUs, breaking the speed and scale barriers of AI.
Powered by NVIDIA® DGX™ software and a scalable architecture based on NVIDIA NVSwitch technology, it enables you to tackle the world's most complex AI challenges.
Product spec
|
DGX-2 |
GPUs |
Tesla V100 × 16 |
GPU memory |
512 GB (total of 16 GPUs) |
Arithmetic performance |
2 petaFLOPS |
NVIDIA CUDA® Core |
81920 |
NVIDIA Tensor Cores |
10240 |
NVSwitch |
12 |
CPU |
Dual Intel Xeon Platinum 8168, 2.7GHz, 24 cores |
system memory |
1.5TB |
Network |
100 Gb/sec × 8 or 8x Infiniband/100GigE Dual 10/25Gb/sec Ethernet |
storage |
OS: 2 x 960GB NVME SSDs Internal storage: 30TB [3.84TB × 8] NVME SSDs |
software |
Ubuntu Linux OS |
Maximum power consumption |
10 kW |
weight |
154.2kg |
size |
Overall height: 440mm Width: 482mm Depth: 795 mm (without front bezel) 834 mm (with front bezel) |
Operating temperature range |
5℃-35℃ |