Nvidia DGX H100 User Manual

NVIDIA DGX H100 User Guide
5.3. Conguring the Local Terminal
There are two ways to access the BIOS setup screen:
▶ A direct-attached keyboard and monitor
▶ Serial-over-LAN (SOL) using IPMI tools
To use the SOL connection, you might need to congure your terminal application.
5.3.1. Linux
1. Set the locale and language for your terminal:
sudo localectl set-locale LANG=en_US.UTF-8
2. Set the locale for the current session:
export LANG=en_UTF-8
3. Type xterm to launch the terminal with the set locale.
5.3.2. Windows and MacOS
▶ Congure your terminal application for en_US.UTF-8 support.
5.4. Power on or Reboot the System
1. Reboot the system using one of the following methods:
▶ Connect to the BMC web interface and click power on/reboot.
▶ From an operating system command line, run sudo reboot.
2. Connect to the DGX H100 SOL console:
ipmitool -I lanplus -H <ip-address> -U admin -P dgxluna.admin sol activate
3. Press the Del or F2 key when the system is booting.
The system conrms your choice and shows the BIOS conguration screen.
38 Chapter 5. SBIOS Settings

Nvidia DGX H100 Specifications

General

GPU	8x NVIDIA H100 Tensor Core GPUs
GPU Memory	640 GB HBM3 (80GB per GPU)
System Memory	2 TB DDR5
Form Factor	6U rackmount
Storage	30TB NVMe SSD
Power Supply	10kW
Interconnect	NVLink 4.0

Summary

Introduction to the NVIDIA DGX H100 System

Hardware Overview

Details the components, specifications, and physical layout of the DGX H100 system.

Network Connections, Cables, and Adaptors

Explains network ports, modules, and supported cables for DGX H100 connectivity.

DGX OS Software

Lists the software components included in the DGX OS stack.

Connecting to the DGX H100

Connecting to the Console

Guides on connecting to the DGX H100 console via direct or remote methods.

SSH Connection to the OS

Describes how to establish an SSH connection to the DGX H100 operating system.

First Boot Setup

System Setup

Covers the initial system setup process after powering on or reimaging.

Post Setup Tasks

Outlines recommended tasks after the initial system setup, like updates.

Quickstart and Basic Operation

Installation and Configuration

Provides requirements and instructions for initial installation and configuration.

Turning DGX H100 On and Off

Provides instructions for safely powering the DGX H100 system on and off.

Verifying Functionality - Quick Health Check

Outlines how to use NVSM for system health checks and verification.

Running NGC Containers with GPU Support

Details methods for providing GPU support for Docker containers.

Managing CPU Mitigations

Discusses security updates for CPU vulnerabilities and performance impact.

SBIOS Settings

Accessing the SBIOS Setup

Guides users on how to access the system BIOS setup utility.

Configuring the Boot Order

Provides instructions for changing the system's boot order.

Using the Baseboard Management Controller (BMC)

Connecting to the BMC

Provides steps to connect to the DGX H100's BMC via a web browser.

Overview of BMC Controls

Details the primary controls available in the BMC web interface.

Changing the BMC Login Credentials

Guides on how to change BMC login credentials and manage users.

Using the Remote Console

Explains how to use the remote console (KVM) via the BMC.

Security

User Security Measures

Discusses user-level security practices for the DGX H100 system.

System Security Measures

Covers security measures incorporated into the NVIDIA DGX H100 system.

Secure Data Deletion

Explains how to securely erase data from DGX H100 system SSDs.

Redfish APIs Support

Supported Redfish Features

Lists the Redfish features supported by the DGX H100 system.

Redfish Examples

Provides examples of using Redfish APIs for system management.

Safety

Safety Information

General safety advice for installing and maintaining the DGX H100 server.

Safety Warnings and Cautions

Explains safety symbols and general warnings for personal injury.

Electrical Precautions

Details important electrical safety information for the DGX H100.

System Access Warnings

Advises on safety precautions when accessing the inside of the system.

Compliance

United States

Details FCC compliance for Class A digital devices in the US.

CE

Covers European Conformity (CE) and relevant directives.

Russia/Kazakhstan/Belarus

Details CU TR and FAC compliance for the region.

Third-Party License Notices

Micron msecli

Provides the license agreement terms for the Micron msecli utility.

Mellanox (OFED)

Outlines the terms and conditions for using Mellanox OFED software.

Notices

Notice

Contains general disclaimers, warranty information, and usage restrictions.

Trademarks

Lists NVIDIA and other company trademarks.