Nvidia Tesla S2050 User manual

TESLA™
U GPU COMPUTING SYSTEM
INSTALLATION GUIDE

NVIDIA Tesla
1U GPU Computing System
TESLA 1U GPU COMPUTING SYSTEM INSTALLATION GUIDE

TABLE OF CONTENTS
TABLE OF CONTENTS
Introduction 1
About This Guide 1
Minimum System Requirements 2
Unpacking 3
Equipment 3
Hardware Installation 5
Safety Instructions 5
Before You Begin 5
Optional Low Profile Bracket 6
Tesla 1U GPU Computing System Installation 7
Mounting the Tesla System in a Rack 7
Connecting to the Host System 10
Software Installation 13
Driver Installation 13
CUDA Toolkit 16
CUDA Software Developer’s Kit (SDK) 16
Verifying Installation if Running X Server 17
Verifying Linux Installation if Not Running X Server 18
References and Resources 21
Getting Driver Updates 21
Registration, Warranty, and Support 22
Registering Your Tesla 1U GPU Computing System 22
Warranty and Support 22

Compliance and Certifications 23
Important Safety Information 30
Precautions for Installing System Unit in Racks 37
Stability 37
Manual Material Handling 38
TABLE OF CONTENTS


INTRODUCTION
INTRODUCTION
The NVIDIA®Tesla™1U GPU Computing System
simplifies adding GPU computing to rack-
mount installations. With Tesla, you can tackle
massive problems with the unprecedented
performance of the many-core architecture
and the C programming enabled by a suite of
developer tools. Tesla enables you to solve
problems faster.
Note: It is important to register your Tesla 1U GPU
Computing System in order to receive NVIDIA Customer
Care online and phone support. To register, please visit:
www.nvidia.com/register.
About This Guide
This installation guide provides instructions to connect the Tesla 1U
GPU Computing System to an NVIDIA-tested host system in a
rack-mount environment. For a list of NVIDIA-tested host systems,
of for information on using the system with applications, refer to the
NVIDIA web site at www.nvidia.com/tesla.


INTRODUCTION
 Minimum System Requirements
Prior to unpacking your new NVIDIA Tesla system, confirm
that your host system meets all the system requirements for a
smooth installation.
>System and Platform
zNVIDIA-tested system or platform
>Operating System
zLinux 32-bit or 64-bit (Red Hat Enterprise
zLinux 3, 4, and 5; SUSE 10.3)
>Processor
zIntel Pentium 4 or Xeon processor or higher
zAMD Opteron processor or higher
>RAM
z2 GB minimum, 4 GB recommended
>System Interface
zPCI Express x16 slot
zPCI Express x8 slot if you are using the optional
x8 interface card


UNPACKING
UNPACKING
Equipment
Check that you have all of the items you need for installing your Tesla
system. Depending on which Tesla product you purchased and from
whom you purchased it, some accessories may have been ordered
and shipped separately. You should have:
One (1) Tesla 1U GPU
Computing System chassis
The 1U chassis includes four (4) Fermi GPUs.
Two (2) NVIDIA Interface Cards
Installs into the PCI Express x16 slot in the host system.
The card comes with the standard ATX-height bracket
attached. A lowprofile bracket is also provided and can
replace the standard bracket on the card. Your solution
provider may have already configured your cards,
brackets and cables to your specific installation needs.
Two (2) Low-Profile Brackets
Used to replace the standard ATX (long) bracket
shipped attached to the interface card. Your solution
provider may have already configured your cards,
brackets, and cables to your specific installation needs.


UNPACKING
 Two (2) Interconnect Cables
Connects from the interface card in the host system to
the Tesla 1U GPU Computing System.
Power Cords
Used to supply power to the Tesla 1U GPU Computing
System. The host system controls the power. Your
solution provider may have already configured your
cards, brackets, and cables to your specific
installation needs.
One (1) Rail Kit
Each rail kit includes two (2) rails for use with a
standard 4-post 19 inch, EIA compliant rack.
WELCOME
Registration Card
Register to gain priority access to NVIDIA’s
Customer Care.
Installation CD
Contains this file (Installation Guide) and all the
necessary drivers to enable the Tesla system. To keep
your drivers current, visit http://developer.nvidia.com/
object/cuda.html and download the latest drivers.
Quick Start Guide
Provides the basic procedures necessary to install and
bring up the Tesla 1U GPU Computing System.


HARDWARE INSTALLATIONS
HARDWARE INSTALLATIONS
Installing the NVIDIA Tesla 1U GPU Computing
System hardware involves opening up your
computer. Follow all of the safety instructions
provided here to ensure that there is no damage
to you, your host system, or the Tesla unit.
Note: All hardware must be installed prior to installation of
the drivers.
Safety Instructions
To reduce the risk of fire, electric shock, and injury always follow
basic safety precautions. Remember to remove power from your host
system by disconnecting the AC main source before installing any
equipment into the host system chassis. For more information
regarding important safety instructions, refer to the Important Safety
Information section. Do not allow anything to rest on the power cord.
Do not place this product where a person can step or trip on the
power cord.
Before You Begin
The Tesla system must be connected to a host system. The host
system accessing the Tesla system must have one or more open PCIe
slots. Any NVIDIA graphics drivers on your host system should be
uninstalled prior to installing the hardware and software associated
with the Tesla system.


HARDWARE INSTALLATIONS
 Optional Low-Profile Bracket
Before beginning the installation, determine if you are going to need
the Tesla interface card with the standard (full-height ATX) bracket or
the shorter low-profile bracket. If you require the lowprofile bracket,
use the following instructions to replace the standard bracket.
Replacing the bracket requires a Philips #2 screwdriver.
1. Remove the two (2) screws holding the bracket to the card.
2. Re-use those screws to attach the new bracket.

HARDWARE INSTALLATIONS 
Tesla 1U GPU Computing System Installation
Mounting the Tesla System in a Rack
Plan your rack configuration to ensure you have the proper cables to
reach from the Tesla system to your host system. The Tesla system
ships with 0.5-meter cables by default. An optional 2-meter cable is
available.
The Tesla system also ships with rails to fit a 4-post 19-inch EIA rack
as the default configuration.
If installed in a closed or multi-unit rack assembly, the operating
ambient temperature of the rack environment may be greater than
room ambient temperature. Therefore, consideration should be given
to installing the equipment in an environment compatible with the
manufacturer’s maximum rated ambient temperature (Tmra). The
maximum operating ambient temperature is 35 ºC.
1 Remove one of the rails from the packing box.
2 Remove the mounting screws attached to either end of the rail.
Set aside as you will need these to secure the rail to the rack at the
end of the steps. Remove any existing graphics card (if there is one).
Save hardware (such as screws and brackets) to use when installing
the NVIDIA graphics card.
Figure 1. Detail of rail for rack mounting
3 Choose either the front or the rear of the rack to start your
rail installation.

HARDWARE INSTALLATIONS
 4 Slide the rail into the side of the rack, and adjust the length
accordingly. Make sure the flare end of the rail is positioned at the
front of the rack.
Figure 2. Detail of putting rail into rack
5 Slide the rail onto rack post. The studs in the front and back should
hold the rail in place temporarily.
6 Use the supplied screws to secure the rails to the rack firmly.
Figure 3. Detail of securing rails in rack
Front of rack
Slide rails to
align the screw
holes
Attach rail to
rear mounting
brackets
Attach rail to
front mounting
brackets

HARDWARE INSTALLATIONS 
7 Repeat Step 1 through Step 7 to install the other rail.
8 Unpack the Tesla system.
9 Align the Tesla system with the installed rails and slide it into
the rack.
Figure 4. Slide the system onto the rails
10 Make sure the status LED on the front panel of the system is located
on the right side for proper installation.
11 Tighten the thumb screws on the front of the unit to secure the Tesla
system to the rack system.
LED display located
on the right side of
the front panel
Align system with
installed rails

 HARDWARE INSTALLATIONS
 Connecting to the Host System
Use the following procedures to connect the Tesla 1U GPU Computing
System system to a host system. Before you begin, make sure you
disconnect the power cord to the host system.
1 Install the NVIDIA Interface Card in an open PCIe slot. If your host
system requires a low-profile bracket on the card, replace the
standard bracket with the low profile bracket provided in the box.
Refer to bottom of page 9 of this document for more info.
2 Connect one end of one of the PCIe cables supplied with your system.
Figure 5. Connecting cable to host system
3 Connect the other end of the cable to the back of the Tesla system.
The thumb tabs on the cable should face outward for the cable to
engage properly.
Figure 6. Connecting PCIe cable to Tesla system
Note: There is no specific order when connecting the
inter-connect cables between the host system and the
Tesla system.
Interconnect cable
connections on the
back of the Tesla
System
Thumb tab should
always be located
on the outside
of the connector

HARDWARE INSTALLATIONS 
4 Connect the power cord to the Tesla system.
Figure 7. Connecting the power cord
5 Plug the power cord from the Tesla system to an AC outlet (it is
recommended to use a power surge protected outlet).
6 Reconnect the power cord to your host system. The host power cord
should have been unplugged at the beginning of the whole installation
process.
7 Power on the host system. At this time, power is also applied to the
Tesla system.
Note: The power-on state of the Tesla system is controlled by
the host system. The Tesla system is powered on only when
the host system is powered on.





DRIVER INSTALLATION
DRIVER INSTALLATION
This section contains the instructions to install
the software drivers within a Linux environment.
In addition to the NVIDIA drivers located on the
Installation CD included with your computing
board, the drivers can be downloaded from the
NVIDIA web site at www.nvidia.com.
The driver follows a unified architecture model in
which a single driver set is used for all supported
NVIDIA GPUs. Your previous NVIDIA unified
driver may not support your Tesla board. In that
case, you will need to install the drivers from the
CD or download from www.nvidia.com.
Driver Installation
Before you begin the installation, you should exit the X server and
close all OpenGL applications (it is possible that some OpenGL
applications persist even after the X server has stopped). You should
also set the default run level on your system such that it will boot to
a VGA console and not directly to X. Doing so will make it easier to
recover if there is a problem during the installation process.
Note: Refer to readme.txt located at /usr/share/doc/NVIDIA_
GLX-1.0/README.txt for more detailed information regarding
the linux driver installation.


 DRIVER INSTALLATION
 1 Download the 32-bit or 64-bit Linux driver (the driver will have a
name similar to “NVIDIA-Linux-x86-100.14.11-pkg1. run” or
“NVIDIA-Linux-x86_64-100.14.11-pkg2.run”).
2 Change to the directory containing the downloaded file.
3 Run as root the NVIDIA-Linux*.run file downloaded in Step 1. As the
root user you can run the following executables:
In the cd.download_directory, choose either NVIDIA-Linux-x86-
100.14.11-pkg1.run or NVIDIA-Linux-x86_64-100.14.11-pkg2.run.
The NVIDIA-Linux*.run file is a self-extracting archive. When
executed, it extracts the content of the archive and runs the contained
nvidia installer utility, which provides an interactive interface to walk
you through the installation. nvidia-installer will also install itself to /
usr/bin/nvidia-installer, which may be used at some later time to
uninstall drivers, auto download updated drivers, etc.
4 Select Accept to accept the License Agreement.
5 Select Yes if a warning window displays indicating that there are
drivers already installed. Selecting Yes tells the installation process
to overwrite the previously installed drivers.
6 Select Yes when the Kernel Interface window displays. When the
installer runs, it will check for a precompiled kernel interface. If it
does not find one, it will attempt to locate an interface on the NVIDIA
ftp site and then download it.

DRIVER INSTALLATION 
7 Select OK to compile a kernel interface. If a kernel interface cannot
be downloaded, either because the FTP site cannot be reached or
because one is not provided, the installer will check your system for
the required kernel sources and compile the interface for you. You
must have the source code for your kernel installed for compilation
to work. On most systems this means that you will need to locate
and install the correct kernel source, kernel headers or kernel
development package.
Linking of the kernel interface (in the case that the interface
downloaded or compiled at installation) required you to have a linker
installed on your system. The linker, usually /usr/bin/ld is part of the
binutils package. If a precompiled kernel interface is not found, you
must install a linker prior to installing the NVIDIA driver.
8 Run the nvidia-xconfig utility. nvidia-xconfig will find the X configuration
file and modify it to use the NVIDIA X driver. If you have an NVIDIA
device that is graphics output capable, you can answer Yes when
the installer asks if it should run it. If you have no device capable of
displaying graphics output, answer No. If you need to reconfigure
your X server later, you can run nvidia-xconfig again from a terminal.
nvidia-xconfig will make a backup copy of your configuration file
before modifying it.
9 Installation is now complete.
Note: The X server must be restarted for configuration
file changes to take effect. More information about nvidia-
xconfig can be found in the nvidia-xconfig manual page by
running: %man.nvidia.config.