Nvidia DGX H100 - Introduction to the NVIDIA DGX H100 System

92 pages

To Next Page

To Previous Page

Chapter 1. Introduction to the NVIDIA

DGX H100 System

The NVIDIA DGX H100 System is the universal system purpose-built for all AI infrastructure and work-

loads, from analytics to training to inference. The system is built on eight NVIDIA H100 Tensor Core

GPUs.

Main Page

Table of Contents3
Introduction to the NVIDIA DGX H100 System7
Hardware Overview8
DGX H100 Models and Component Descriptions8
Mechanical Specifications9
Power Specifications9
Support for 4+2 Redundancy9
DGX H100 Locking Power Cord Specification9
Using the Locking Power Cords10
Environmental Specifications11
Front Panel Connections and Controls11
With a Bezel11
With the Bezel Removed13
Rear Panel Modules13
Motherboard Connections and Controls14
Motherboard Tray Components15
GPU Tray Components16
Network Connections, Cables, and Adaptors17
Network Ports17
Compute and Storage Networking18
Network Modules19
Supported Network Cables and Adaptors20
DGX H100 System Topology21
DGX os Software21
Customer Support22
Connecting to the DGX H10023
Connecting to the Console23
Direct Connection23
Remote Connection through the BMC25
SSH Connection to the os27
First Boot Setup29
System Setup29
Post Setup Tasks31
Obtaining Software Updates31
Enabling the SRP Daemon31
Quickstart and Basic Operation33
Installation and Configuration33
Registration33
Obtaining an NGC Account34
Turning DGX H100 on and off34
Startup Considerations34
Shutdown Considerations34
Verifying Functionality - Quick Health Check34
Running the Pre-Flight Test35
Running NGC Containers with GPU Support36
Using Native GPU Support36
Using the NVIDIA Container Runtime for Docker37
Managing CPU Mitigations38
Determining the CPU Mitigation State of the DGX System38
Disabling CPU Mitigations39
Re-Enabling CPU Mitigations39
SBIOS Settings41
Accessing the SBIOS Setup41
Configuring the Boot Order42
Configuring the Local Terminal44
Linux44
Windows and Macos44
Power on or Reboot the System44
Using the Baseboard Management Controller (BMC)45
Connecting to the BMC45
Overview of BMC Controls46
Changing the BMC Login Credentials49
Using the Remote Console49
Setting up Active Directory, LDAP, or E-Directory49
Configuring Platform Event Filters50
Uploading or Generating SSL Certificates51
Viewing the SSL Certificate51
Generating the SSL Certificate52
Uploading the SSL Certificate53
Updating the SBIOS Certificate53
Security57
User Security Measures57
Securing the BMC Port57
System Security Measures57
Secure Flash of DGX H100 Firmware58
Encryption58
NVIDIA System Manager Security58
Secure Data Deletion58
Prerequisites58
Procedure59
Redfish Apis Support61
Supported Redfish Features61
Redfish Examples62
BMC Manager62
Firmware Update63
BIOS Settings65
Telemetry66
Chassis66
SEL Logs67
Virtual Image67
Collect BMC Debug Data67
Safety69
Safety Information69
Safety Warnings and Cautions69
Intended Application Uses70
Site Selection70
Equipment Handling Practices71
Electrical Precautions71
Power and Electrical Warnings71
Power Cord Warnings72
System Access Warnings72
Rack Mount Warnings73
Electrostatic Discharge74
Other Hazards74
California Department of Toxic Substances Control74
Nickel74
Battery Replacement74
Cooling and Airflow75
Compliance77
United States77
United States/Canada77
Canada78
Australia and New Zealand79
Brazil79
Japan79
South Korea81
China82
Taiwan84
Russia/Kazakhstan/Belarus85
Israel86
India86
South Africa87
Great Britain (England, Wales, and Scotland)87
Third-Party License Notices89
Micron Msecli89
Mellanox (OFED)90
Notices91
Notice91
Trademarks92
Copyright92

Nvidia DGX H100 - Introduction to the NVIDIA DGX H100 System

Table of Contents

Other manuals for Nvidia DGX H100

Related product manuals