Difference between revisions of "Jetson TK1"

From eLinux.org
Jump to: navigation, search
(Added shutdown info)
(Projects using Jetson TK1: Added TR4Tegra)
Line 273: Line 273:
A user is using CUDA accelerated computer vision to find balloons (or other red round things). [http://pleasingsoftware.blogspot.com/2014/06/identifying-balloons-using-computer.html This article explains the algorithm and has a link to the source.]
A user is using CUDA accelerated computer vision to find balloons (or other red round things). [http://pleasingsoftware.blogspot.com/2014/06/identifying-balloons-using-computer.html This article explains the algorithm and has a link to the source.]
|'''Project Jedroid'''
|[[File:Jedroid v1.0.jpg|300px]]
'''Project Jedroid'''
Android KK on Jetson TK1.
Android KK on Jetson TK1.
Pure AOSP, Pure open source code.  
Pure AOSP, Pure open source code.  
[http://jetson.co/wiki/main_page/android-on-jeston Preview Image Ready! Enjoy!]
[http://jetson.co/wiki/main_page/android-on-jeston Preview Image Ready! Enjoy!]
[[File:Jedroid v1.0.jpg|300px]]
|Play Android 2K14 NBA on JetsonTK1
NBA 2K14 Android game using Jedroid on JetsonTK1
[https://www.youtube.com/watch?v=SjgPy0R1fms  Demo video]
[https://www.youtube.com/watch?v=SjgPy0R1fms  Demo video]
|[[File:TR4Tegra.jpg|231px|Transformative Reality]]
TR4Tegra. Smart-glasses with a depth camera, connected to Jetson TK1 in a backpack, detects object location and notifies a blind user with a ~25x25 neural implant
[https://plus.google.com/u/0/+WaiHoLi/posts/UxtarcqvUJb info]

Revision as of 00:27, 29 October 2014

About this site

This is the official Wiki for embedded Tegra & the Jetson TK1 board, maintained by both the community and NVIDIA.

The other embedded Tegra community sites with official NVIDIA support are:

  • The forum for discussing embedded Tegra & Jetson TK1 issues with the community & NVIDIA.
  • The blog to stay updated with the latest news & plans for embedded Tegra & Jetson TK1 from NVIDIA.

About Tegra K1

Tegra K1 is the world's first chip to have the same advanced features & architecture as a modern desktop GPU while still using the low power draw of a mobile chip! The Jetson TK1 board therefore allows embedded devices to use the exact same CUDA code that would also run on a desktop GPU (used by over 100,000 developers), with similar levels of GPU-accelerated performance as a desktop!

About Jetson TK1

Jetson TK1, "the world's first embedded supercomputer!"

Jetson TK1 is NVIDIA's embedded Linux development platform featuring a Tegra K1 SOC (CPU+GPU+ISP in a single chip), selling for $192 USD ($1 per CUDA core!). Jetson TK1 comes pre-installed with Linux4Tegra OS (basically Ubuntu 14.04 with pre-configured drivers). There is also some official support for running other distributions using the mainline kernel, discussed further in the Distributions and Mainline kernel sections below.

Besides the quad-core 2.3GHz ARM Cortex-A15 CPU and the revolutionary Tegra K1 GPU, the Jetson TK1 board includes similar features as a Raspberry Pi but also some PC-oriented features such as SATA, mini-PCIe and a fan to allow continuous operation under heavy workloads:

Hardware Features

  • Dimensions: 5" x 5" (127mm x 127mm) board
  • Tegra K1 SOC (CPU+GPU+ISP in a single chip, with typical power consumption between 1 to 5 Watts):
    • GPU: NVIDIA Kepler "GK20a" GPU with 192 SM3.2 CUDA cores (over 300 GFLOPS!)
    • CPU: NVIDIA "4-Plus-1" 2.32GHz ARM quad-core Cortex-A15 CPU with Cortex-A15 battery-saving shadow-core
  • DRAM: 2GB DDR3L 933MHz EMC x16 using 64-bit data width
  • Storage: 16GB fast eMMC 4.51 (routed to SDMMC4)
  • ROM: 4MB SPI boot flash (routed to SPI4)
  • mini-PCIe: a half-height single-lane PEX slot (such as for Wifi, SSD RAID, FireWire or Ethernet addon cards)
  • SD/MMC card: a full-size slot (routed to SDMMC3)
  • USB 3.0: a full-size Type-A female socket
  • USB 2.0: a micro-AB female socket (for connecting to a PC, but can also be used as a spare USB 2.0 port using a micro-B male to female Type-A adapter that is sometimes included)
  • HDMI: a full-size port
  • RS232: a full-size DB9 serial port (routed to UART4)
  • Audio: an ALC5639 Realtek HD Audio codec with Mic in and Line out jacks (routed to DAP2)
  • Ethernet: a RTL8111GS Realtek 10/100/1000Base-T Gigabit LAN port using PEX
  • SATA: a full-size port that supports 2.5" and 3.5" disks, but is not hot-pluggable. (Turn off the power before plugging in SATA disk drives)
  • JTAG: a 2x10-pin 0.1" port for debugging
  • Power: a 12V DC barrel power jack and a 4-pin PC IDE power connector, using AS3722 PMIC
  • Fan: a fan-heatsink running on 5V (to allow safely running intense workloads continuously, but can usually be replaced by a heat-spreader or heatsink)

The following signals are available through the 125-pin 2mm-pitch expansion port:

  • Camera ports: 2 fast CSI-2 MIPI camera ports (one 4-lane and one 1-lane)
  • LCD port: LVDS and eDP Display Panel
  • Touchscreen ports: Touch SPI 1 x 4-lane + 1 x 1-lane CSI-2
  • UART
  • HSIC
  • I2C: 3 ports
  • GPIO: 7 x GPIO pins (1.8V). Camera CSI pins can also be used for extra GPIO if you don't use both cameras.

Hardware-accelerated APIs supported:

  • CUDA 6.0 (SM3.2, roughly the same as desktop SM3.5)
  • OpenGL 4.4
  • OpenGL ES 3.1
  • OpenMAX IL multimedia codec including H.264, VC-1 and VP8 through Gstreamer
  • NPP (CUDA optimized NVIDIA Performance Primitives)
  • OpenCV4Tegra (NEON + GLSL + quad-core CPU optimizations)
  • VisionWorks

Buying Guide

Where can I get one and for how much?

  • Visit the NVIDIA store then choose a distributor in USA, or click on "International Orders" to see other countries it is available from including UK, Germany, France, Italy, Japan, Russia, China, Singapore, and Australia.
  • Jetson TK1 costs $192 ($1 per CUDA core!) in USA.
  • The package includes a power supply (with a detachable US mains cord), a USB micro-B cable for connecting it to a PC, and usually a USB micro-B to female USB-A adapter allowing you to have 2 regular USB ports.

Basic setup steps to access the board and access internet

Whether you want to sew your embedded Tegra into a backpack or put it in a robot or simply use it as an ultra powerful media center, the first thing you should do with a new Jetson TK1 board is attach it to a HDMI monitor & keyboard & mouse to make sure it works and get familiar with it for a few minutes. The device can be accessed in 2 possible ways, depending on whether you want to plug a keyboard & mouse & monitor directly into the Jetson TK1 board or you want to plug an Ethernet cable between your device and a PC or laptop or router and access it through a network:

Direct access to a Jetson board using its own keyboard & mouse & monitor

The Jetson TK1 Quick Start Guide (included as a booklet with your Jetson TK1) shows how to use the Jetson TK1 board as a mini standalone computer. Basically, you plug in a HDMI monitor or TV, plug a keyboard into the USB3.0 port, plug a mouse into the included micro-B to female USB adapter and plug that into the micro-B USB2.0 port on the board. Then plug the 12V power supply in, press the small POWER button, then watch it boot up into Linux4Tegra (Ubuntu 14.04 with some drivers pre-configured). When it asks for the password for user "ubuntu", just type "ubuntu" to log in. If you have an Ethernet router then simply plug an Ethernet cable from the board into your router (or plug in a USB Wifi dongle) to have internet access, and you are ready.

Remote access to a Jetson board through the Ethernet port

To access the board remotely through a local network from a PC or laptop, follow the Remote Access instructions so you can control the device from the keyboard & mouse & monitor on your PC or laptop and share your desktop's Wifi or Ethernet internet access to the attached device.

Recommended first steps now that your board has internet access

Open a command-line terminal to perform some initial operations. If you are using the Jetson TK1's graphical environment then click on the top-left icon in Ubuntu Unity and type "terminal" to open a command shell, or if you have remote access to the device through a network then open an SSH command shell into your device from your desktop such as by running "ssh ubuntu@tegra-ubuntu". Now you are ready for initial configuration.

Add the Universe package repositories, since you will often need packages from Universe for code development:

sudo apt-add-repository universe
sudo apt-get update
If you will use the shell command-line a lot

Install "bash-completion" (it allows you to hit the "Tab" key to auto-complete your shell commands) and "command-not-found" (it shows which package you probably need to install if you run an unavailable command). These 2 tools are extremely useful when using the commandline, but were not installed by default in Ubuntu 14.04. Simply run this:

sudo apt-get install bash-completion command-not-found

Note: now you need to log back in for it to start using bash-completion and command-not-found.

You probably should also change the shell prompt (by adjusting "PS1" in the ".bashrc" file in your home directory) to be more useful, such as getting the shell prompt to have a different color than regular commands, and make it obvious if a command returned with an error. There are thousands of custom .bashrc configurations on the web, including Shervin's that provides a different colored shell prompt depending on whether a command was succesful or returned an error.

If you will use the graphical environment (Unity) a lot

You might want to try some suggestions at "http://itsfoss.com/things-to-do-after-installing-ubuntu-14-04/" or similar, such as to turn off the desktop shopping suggestions that are enabled by default in Ubuntu 14.04 (despite the spyware concerns discussed by huge numbers of people) by running this:

gsettings set com.canonical.Unity.Lenses disabled-scopes "['more_suggestions-amazon.scope', \
    'more_suggestions-u1ms.scope', 'more_suggestions-populartracks.scope', 'music-musicstore.scope', \
    'more_suggestions-ebay.scope', 'more_suggestions-ubuntushop.scope', 'more_suggestions-skimlinks.scope']"
If you need more disk space on the eMMC

The eMMC on the Jetson has a capacity of 16GB, however some instructions or boards default to only using 8GB of the drive (or 12GB, see discussion). If you require more disk space in your rootfs such as for installing toolkits or compiling large projects, you can flash the Jetson (from a Linux desktop) to have a larger filesystem (note that this will erase all data on the Jetson TK1, and it takes roughly 1 hour to flash the whole eMMC!):

[user@host Linux_for_Tegra]$ sudo ./flash.sh -S 14580MiB jetson-tk1 mmcblk0p1
sending file: system.img
/ 15032385536/15032385536 bytes sent
system.img sent successfully
Create, format and download  took 2791 Secs
Time taken for flashing 2792 Secs

The maximum value of the flash.sh -S flag that works successfully is 14580MiB. Greater values (like 16GiB) are beyond the capacity of the eMMC when the other system-required partitions are included.

Shutting down Jetson TK1 safely

Just like any Linux computer, the recommended way to shut-down or turn off Jetson TK1 is to click Shutdown in the GUI or run this in a terminal, to ensure the filesystem will not be corrupt:

sudo shutdown -h now

Tutorials for developing with Jetson TK1

The following are tutorial projects for hardware and/or software development. They show the easiest way to do a certain task, while the Reference Information section below has more detailed pages. If you have something useful to contribute about Jetson TK1 or embedded Tegra then please do so.

Tutorial Description Programming Difficulty Electronics Difficulty
Hello World Create a simple program that prints "Hello World!", by compiling code directly on your device #
CUDA Install CUDA then build & run some CUDA sample projects #
Nsight Documentation and tutorials on GPU debugging and profiling with Nsight, which is installed with CUDA #
OpenCV Install OpenCV then grab camera frames or build & run some OpenCV samples #
OpenGL Links to many OpenGL and OpenGL ES sample projects #
Full Body Detection Shows how to perform full body detection (something that even desktop CPUs are too slow for!) from a webcam and display the face, using OpenCV #
GPIO Input & Output Turn on an LED or send a signal to an Arduino microcontroller using GPIO # #
Vision-controlled GPIO Turn on an LED whenever a face is detected in your camera # #
Battery Power Power your Jetson TK1 from a battery pack #
Automatic Pan Tilt Build an autonomous pan-tilt face tracking camera, that tracks faces as they move around # #
Optical Flow Motion Generate the optical flow motion vectors to see how things are moving ##
Video Stabilization Real-time video stabilization such as for a robot's onboard camera ##
Program An Arduino Program an Arduino microcontroller from your Jetson TK1 (instead of from a PC) ##
Communicate To An Arduino Connect an Arduino board to a Jetson TK1 board, with communication between then ## ##
Follower Robot Get a wheeled robot to drive towards the nearest person ## ##
Walking Follower Robot Get a 2-legged robot to walk towards the nearest person ### ##

Jetson TK1 Reference Information (more detailed than the tutorials above)

General Topics

  • JetsonTK1 Support Page: Includes a lot of the open-source hardware documents, such as the mechanical drawings, Bill of Materials (BOM), electrical circuit schematics, CAD files, Jetson TK1 module spec, Tegra K1 SOC Technical Reference Manual (TRM), etc.
  • System Info: Find out the hardware info & what is available on the Jetson TK1 such as disk space, RAM, and devices.
  • Cloning & Backup Save the Jetson's eMMC (system.img) via the flash port and restore it to other Jetson boards.

Networking Topics

  • Wifi & Ethernet Adapters: Discusses which Wifi or Ethernet cards have been tested on Jetson TK1.
  • Remote Access: Control the device from your PC or laptop, and share your desktop's Wifi or Ethernet internet access to the attached device.

Software Topics

Performance and Power Topics

  • Performance: How to maximize or minimize CPU & GPU clocks for different performance & power draw.
  • Power Management: Electrical power related issues, such as powering the board from a battery or seeing how much power is drawn.
  • Thermal: How to replace the fan with a heat-spreader or heatsink, and measure the board's temperatures.
  • Graphics Performance: Discusses graphics performance of the Tegra TK1 on the board, including comparison with competing solutions.
  • Computer Vision Performance & Power Draw: Describes the different Tegra hardware optimizations in OpenCV, and shows power & perf measurements for many computer vision applications in OpenCV, CUDA and VisionWorks.

Hardware Expansion Topics

  • Cameras & Webcams: Discusses USB cameras and CSI cameras, stereo cameras and Time-Of-Flight 3D depth cameras.
  • mini-PCIe Add-on Cards: Shows some possible add-on cards that can be used in the mini-PCIe port.
  • GPIO: How to control digital output pins and read input data pins on Jetson TK1.
  • I2C: How to communicate between microcontrollers & devices using I2C on Jetson TK1.
  • Bluetooth: Use wireless Bluetooth devices such as keyboards, mice & speakers, or communicate to a smartphone or tablet.
  • Enclosures & Cases: Various cases that are available to protect a Jetson TK1.
  • Audio: Issues related to audio input or output, such as microphones or speakers.

Projects using Jetson TK1

The following are descriptions, photos and/or videos of projects featuring Jetson TK1. Feel free to add your own!

Linux 3.15 + Wayland + Weston + Nouveau running on Jetson TK1

Upstream Linux 3.15 + Wayland + Weston + Nouveau running a 100% open-source OS on Jetson TK1, by CodeThink in UK. If you want to try a similar setup on your Jetson TK1, read this and this.


NVIDIA's SCOL "Super-Computer-On-Legs" robot, by Shervin Emami in Australia. Performs Optical Flow video stabilization and HOG person detection to walk towards nearby people. Tutorial is coming soon!

GE Lidar + Optical Flow demo

Jetson TK1 Lidar range finder + camera Optical Flow robotics demo by General Electric Intelligent Platforms in USA. GE IP are developing a ruggedized Tegra K1 based module.

Explorer Robot

Explorer Robot by the Officine Robotiche in Italy

MyzharBot Robot

MyzharBot by Walter Lucetti in Italy is a crawler robot made to study Navigation algorithms based on Computer Vision, Machine Learning and 3D Sensors. The project is open source and open hardware and has its own blog where you can find every information to replicate it.

Project TURBO by GE

Project TURBO (TK1 Unmanned Reconnaissance Bot) is a low-cost mobile research platform developed by GE Intelligent Platforms, exploring CUDA-accelerated autonomy, sensing & perception powered by Tegra K1.

Cluster Computer

Several users are building cluster computers using a handful of Jetson TK1's:


A user is using CUDA accelerated computer vision to find balloons (or other red round things). This article explains the algorithm and has a link to the source.

Jedroid v1.0.jpg

Project Jedroid Android KK on Jetson TK1. Pure AOSP, Pure open source code. Preview Image Ready! Enjoy!


NBA 2K14 Android game using Jedroid on JetsonTK1 Demo video

Transformative Reality

TR4Tegra. Smart-glasses with a depth camera, connected to Jetson TK1 in a backpack, detects object location and notifies a blind user with a ~25x25 neural implant info

Linux distributions running on Tegra

Jetson TK1 comes preloaded with NVIDIA's Linux4Tegra (L4T) distribution. However it is possible to install other distributions on a Tegra device:

  • Linux4Tegra (L4T) distribution: basically Ubuntu 14.04 with pre-configured drivers for bootloader, kernel, OpenGL, X.Org, Multimedia, etc.
  • Gentoo Linux from SD: run Gentoo directly on an SD-card without modifying the contents of the eMMC storage. (Will soon support many other Linux distros such as Debian, Ubuntu, Redhat, etc!)
  • Gentoo Linux: run Gentoo Linux (from eMMC) instead of Ubuntu Linux.
  • Android: The user "lucasdai" is working on a community-supported version of Android on Jetson TK1.
  • others?

Upstream/mainline OS & kernel system software on Tegra

NVIDIA's Tegra SoCs are well supported by mainline OSS such as Linux and U-Boot. This section contains an index of topics related to running mainline software on Tegra.

Other embedded Tegra boards created by NVIDIA

Jetson TK1 was the first embedded board that NVIDIA created for the general public, but there have also been some other Tegra boards, including the automotive-grade Tegra-K1 based Visual Compute Module and the Jetson Pro development platform, both for the automotive industry (requires an NDA and large sales figures, etc).

Other SOCs created by NVIDIA

SOCs: NVIDIA made several previous generations of Tegra SOCs for the mobile, automotive and MP3 player industries.

Other links

Resources is a list of links to Tegra-related documentation and code outside of this wiki, such as TRMs and mainline kernel links.