NVIDIA DGX systems

NVIDIA DGX systems are supercomputers for problem solving of machine learning and AI. It is currently the most powerful hardware and tuned software stack that contains the most widely used machine learning frameworks in the form of application containers.

NVIDIA DGX-1

Essential instrument of AI research from NVIDIA designed for data centers.

More info

NVIDIA DGX-2

The world’s most powerful AI system for the most complex AI challenges with power 2 PetaFLOPS.

More info

NVIDIA DGX Station

The first personal supercomputer for machine learning and artificial intelligence designed for your office.

More info

Hardware

Let’s take a look at the NVIDIA DGX in detail, first from hardware point of view.

Parameter	NVIDIA DGX-2	NVIDIA DGX-1	NVIDIA DGX Station
GPUs	16× NVIDIA Tesla V100 32GB	8× NVIDIA Tesla V100 32GB	4× NVIDIA Tesla V100 32GB
Performance	2 PetaFLOPS	1 PetaFLOPS	0,5 PetaFLOPS
GPU memory	512 GB in total	256 GB in total	128 GB in total
CPU	2× Platinum 8168, 2.7 GHz (24 cores)	2× E5-2698 v4 2.2GHz (20 cores)	E5-2698 v4 2.2GHz (20 cores)
NVIDIA CUDA cores	81 920	40 960	20 480
NVIDIA Tensor cores	10 240	5 120	2 560
GPU interconnect	NVSwitch, non-blocking, 2.4 TB/s	NVLink, hypercube topology	NVLink
RAM	1,5 TB	512 GB	256 GB
HDD	2× 960GB NVME SSD, 8× 3.84TB NVME SSD	1× 480GB SSD, 4× 1,92TB SSD	4× 1,92TB SSD
Network	2× 10/25Gb Ethernet, 8× 100Gb Infiniband/Ethernet	2× 10Gb Ethernet, 4× 100Gb Infiniband/Ethernet	2× 10GbE
Power consumption	10 kW	3 500 W	1 500 W
Case	rack, 10U	rack, 3U	tower, watter cooling of GPU, CPU
	spec sheet	spec sheet	spec sheet

NVIDIA offers an even more powerful version NVIDIA DGX-2H next to DGX-2. The NVIDIA DGX-2H is equipped with two Intel Xeon Platinum 8174, 3.1 GHz (24 cores) processors, tuned to 2.1 PetaFLOPS peak power and a maximum power input of 12kW. Other DGX-2H parameters are the same as DGX-2.

All NVIDIA DGX systems feature the latest and fastest accelerators today — NVIDIA Tesla V100 32GB. DGX Station contains four cards, DGX-1 eight cards and DGX-2 even sixteen accelerators! Main benefits of NVIDIA Tesla cards are specialized Tensor cores for accelerating machine learning applications or large memory (32 GB for each card) secured by ECC technology. NVIDIA Tesla cards are also equipped by interface for high bandwidth card communication — NVLink. NVLink can reach speed up to 300 GB/s. NVIDIA DGX-2 additionally offers super powerful NVSwitch that connects sixteen NVIDIA Tesla V100 with 2.4 TB/s bisectional bandwidth in non-blocking architecture.

How to leverage Multi-GPU systems?

This is the one of the most often question we can hear from our customers. There are couple of techniques you can use as described in webinar to Multi-GPU topic. Or you can attend Fundamentals of Deep Learning for Multi-GPUs workshop organized by us together with NVIDIA Deep Learning Institute (DLI).

Software

What is more interesting, however, is the already mentioned software package offered by NVIDIA DGX machines. All of these offer pre-installed and performance-tuned environments for machine learning (e.g. Caffe, resp. Caffe 2, Theano, TensorFlow, PyTorch, nebo MXNet) or an intuitive environment for data analysts (NVIDIA Digits). All of this is elegantly packed in Docker Containers. Such a tuned environment provides 30% more power for machine learning applications against applications deployed purely on NVIDIA hardware. The main advantage of the pre-installed environment is the deployment speed, which is in units of hours. The base DGX system image contains Ubuntu operating system, NVIDIA GPU drivers and Docker environment for application containers downloadable from NVIDIA GPU Cloudu (NGC). NVIDIA also supports to run these Docker images in Singularity environment.

NVIDIA DGX systems SW stack

NVIDIA GPU Cloud

NVIDIA GPU Cloud (NGC) represents repository of the most used frameworks for machine learning and deep learning applications, HPC applications, or NVIDIA GPU cards accelerated visualization. Deploying these applications is a question minutes — copying a link of the appropriate Docker image from NGC repositry, moving it on the DGX system, and downloading and running the Docker container. Content of Docker images — versions of all the libraries and frameworks or setting environment parameters — is updated and optimized by NVIDIA specialists for deployment on DGX systems. https://ngc.nvidia.com/

NVIDIA GPU Cloud (NGC)

Support

The strength of the NVIDIA solution is to support the entire system. Hardware support (in case of failure of any of the components) is a matter of course. Software support for the entire environment is critical if something does not work. The customer has hundreds of developers ready to help. Support is part of NVIDIA DGX purchase. It is available for 1, 3 or 5 years and can be further extended after this time.

NVIDIA support includes:

Access to NVIDIA GPU Cloud (NGC) Portal
NVIDIA Cloud Management
DGX Software Upgrades
DGX Software Updates
DGX Firmware Updates
Hardware Support
Hardware SLA (replacement parts) 1 day
Software support — DGX OS image and full AI software including ML frameworks
Enterprise Support Portal
24×7 Phone Support
Access to NVIDIA Knowledgebase

You can also use consultancy with specialists from M Computers in Czech, Slovak and English.

With a combination of tuned hardware, software and NVIDIA support, NVIDIA DGX systems delivers significantly higher performance and acceleration in learning phase:

Acceleration against traditional CPU computing environment, GPU environments vs. NVIDIA DGX-1.

The difference between tuned DGX system solution for fas and powerful machine learning and DIY variant (Do It Yourself) is evident from the following video:

We delivered our first Nvidia DGX-2 system to IT4Innovations Supercomputing Center VSB in Ostrava, Czech Republic. You can see the details of the installation in the short reference video.

Reference Architectures

NVIDIA DGX systems represents really huge computing power. When designing an architecture, it is necessary to consider their involvement in the overall IT infrastructure and its tuning to achieve maximum performance. NVIDIA has introduced NVDIA DGX POD Reference Architecture, including networking and storage disc arrays. Here you will find individual design proposals from the key storage vendors that describe the overall infrastructure solution for running ML and AI applications. We are the most experienced with ONTAP AI solution with NetApp storage.

NetAPP ONTAP AI reference architecture	Pure Storage AIRI reference architecture
DDN A3I reference architecture	IBM Spectrum AI reference architecture

ONTAP AI reference architecture with NetApp storage

NVIDIA Deep Learning Institute

The NVIDIA Deep Learning Institute (DLI) offers both online and hands-on trainings for developers, data scientists, and researchers looking to solve challenging problems with deep learning and accelerated computing. More info.

GTC 2020

In March 2020, the premier GPU Technology Conference (GTC) will take place in San Jose. You were able to get acquainted with NVIDIA DGX systems and examples of their deployment and listen to the visionary lectures of the most important people in the fields of artificial intelligence and machine learning, including the charismatic CEO of NVIDIA Jensen Huang.

GTC 2020

NVIDIA offers special programs for DGX systems and Tesla accelerators for EDU organizations or start-up companies. We are pleased to provide you information about current promotions.

Get an offer

Testing

We have the NVIDIA DGX Station available to test performance and speed of deployment of ML and AI applications. Thanks to the NVIDIA Tesla Test Drive program we have also 2 x NVIDIA Tesla V100 and 2× NVIDIA T4 available for testing. If you are interested please fill out this form.