Difference between revisions of "Jetson"

From eLinux.org
Jump to: navigation, search
(Corrected the GPU core count in Jetson Nano)
m
(20 intermediate revisions by 2 users not shown)
Line 1: Line 1:
The Jetson line of embedded Linux AI and computer vision compute modules and devkits from NVIDIA:
+
The NVIDIA Jetson line of embedded Linux AI and computer vision compute modules and devkits:
* [[Jetson TK1]]: single-board 5" x 5" computer featuring Tegra K1 SOC (quad-core 32-bit Cortex-A15 + 192-core Kepler GPU), 2GB DDR3, and 8GB eMMC.
+
 
* [[Jetson TX1]]: carrier-board + compute module featuring Tegra X1 SOC (quad-core 64-bit Cortex-A57 + 256-core Maxwell GPU), 4GB 64-bit LPDDR4, and 16GB eMMC.
+
* [[Jetson AGX Xavier]] Developer Kit and series of production modules
* [[Jetson TX2]]: carrier-board + compute module featuring Tegra X2 SOC (quad-core 64-bit Cortex-A57 + dual-core NVIDIA Denver2 CPU + 256-core Pascal GPU), 8GB 128-bit LPPDR4, 32GB eMMC.
+
* [[Jetson Xavier NX]] Developer Kit and production module
* [[Jetson Nano]]:  carrier-board + compute module featuring Tegra X1 SOC (quad-core 64-bit Cortex-A57 + 128-core Maxwell GPU), 4GB 64-bit LPDDR4, 4K video encoder/decoder.
+
* [[Jetson Nano]] Developer Kit, Jetson Nano 2GB Developer Kit, and production module
* [[Jetson AGX Xavier]]: carrier-board + compute module featuring Xavier SOC (octal-core 64-bit ARMv8.2 + 512-core Volta GPU with Tensor Cores + dual [http://nvdla.org/ DLAs]), 16GB 256-bit LPDDR4x, 32GB eMMC.
+
* [[Jetson TX2]] Developer Kit and series of production modules
 +
* [[Jetson TX1]] Developer Kit and production module
 +
* [[Jetson TK1]] Developer Kit for the Tegra K1 SOC
 +
 
 +
Here are some quick links and references to get started:
 +
* Jetson Developer Site - [http://developer.nvidia.com/embedded developer.nvidia.com/embedded]
 +
* Jetson Zoo - [https://eLinux.org/Jetson_Zoo eLinux.org/Jetson_Zoo]
 +
* JetPack Downloads - [http://developer.nvidia.com/jetpack developer.nvidia.com/jetpack]
 +
* Community Forums - [https://forums.developer.nvidia.com/c/agx-autonomous-machines/jetson-embedded-systems forums.developer.nvidia.com]
 +
* Partner Products - [https://developer.nvidia.com/embedded/jetson-partner-supported-cameras Supported Cameras] | [https://developer.nvidia.com/embedded/community/jetson-partner-products Production Systems]
 +
<br />
  
 
== NVIDIA Jetson Modules ==
 
== NVIDIA Jetson Modules ==
Line 10: Line 20:
 
{| class="wikitable" style="text-align: center;"
 
{| class="wikitable" style="text-align: center;"
 
|-
 
|-
! Features !! [[Jetson Nano]] !! [[Jetson TX1]] !! [[Jetson TX2|Jetson TX2 / TX2i]] !! [[Jetson AGX Xavier]]
+
! Features !! [[Jetson Nano]] !! [[Jetson TX1]] !! [[Jetson TX2|Jetson TX2 series]] !! [[Jetson Xavier NX]] !! [[Jetson AGX Xavier|Jetson AGX Xavier series]]
 
|-
 
|-
 
| || [[File:Jetson-Nano-Compute-Module-400px.png|260px|center]]
 
| || [[File:Jetson-Nano-Compute-Module-400px.png|260px|center]]
 
  || [[File:NVIDIA Jetson TX1 module.jpg|300px|center]]
 
  || [[File:NVIDIA Jetson TX1 module.jpg|300px|center]]
 
  || [[File:NVIDIA_JTX2_Module_400px.png|300px|center]]
 
  || [[File:NVIDIA_JTX2_Module_400px.png|300px|center]]
 +
|| [[File:JetsonXavierNX-Module_TopDown_400px.png|260px|center]]
 
  || [[File:Xavier-module-topdown-alpha-300px.png|275px|center]]
 
  || [[File:Xavier-module-topdown-alpha-300px.png|275px|center]]
 
|-
 
|-
 
| CPU || ARM Cortex-A57 (quad-core) @ 1.43GHz || ARM Cortex-A57 (quad-core) @ 1.73GHz || ARM Cortex-A57 (quad-core) @ 2GHz +
 
| CPU || ARM Cortex-A57 (quad-core) @ 1.43GHz || ARM Cortex-A57 (quad-core) @ 1.73GHz || ARM Cortex-A57 (quad-core) @ 2GHz +
 
NVIDIA Denver2 (dual-core) @ 2GHz
 
NVIDIA Denver2 (dual-core) @ 2GHz
|| NVIDIA Carmel ARMv8.2 (octal-core) @ 2.26GHz   
+
|| NVIDIA Carmel ARMv8.2 (6-core) @ 1.4GHz
 +
(6MB L2 + 4MB L3)
 +
|| NVIDIA Carmel ARMv8.2 (8-core) @ 2.26GHz   
 
(4x2MB L2 + 4MB L3)
 
(4x2MB L2 + 4MB L3)
 
|-
 
|-
| GPU || 128-core NVIDIA Maxwell @ 921MHz || 256-core NVIDIA Maxwell @ 998MHz || 256-core NVIDIA Pascal @ 1300MHz || 512-core Volta @ 1377 MHz + 64 Tensor Cores
+
| GPU || 128-core NVIDIA Maxwell @ 921MHz || 256-core NVIDIA Maxwell @ 998MHz || 256-core NVIDIA Pascal @ 1300MHz || 384-core Volta @ 1100MHz + 48 Tensor Cores || 512-core Volta @ 1377 MHz + 64 Tensor Cores
 
|-
 
|-
| DL || colspan="3" style="text-align: center;" | NVIDIA GPU support (CUDA, cuDNN, TensorRT) || dual NVIDIA [http://nvdla.org/ Deep Learning Accelerators]
+
| DL || colspan="3" style="text-align: center;" | NVIDIA GPU support (CUDA, cuDNN, TensorRT) || colspan="2" style="text-align: center;" | dual NVIDIA [http://nvdla.org/ Deep Learning Accelerators]
 
|-
 
|-
| Memory || colspan="2" style="text-align: center;" | 4GB 64-bit LPDDR4 @ 1600MHz &#124; 25.6 GB/s || 8GB 128-bit LPDDR4 @ 1866Mhz &#124; 58.3 GB/s || 16GB 256-bit LPDDR4x @ 2133MHz &#124; 137GB/s
+
| Memory || colspan="2" style="text-align: center;" | 4GB 64-bit LPDDR4 @ 1600MHz &#124; 25.6 GB/s || 8GB 128-bit LPDDR4 @ 1866Mhz &#124; 58.3 GB/s || 8GB 128-bit LPDDR4x @ 1600MHz &#124; 51.2GB/s || 32GB 256-bit LPDDR4x @ 2133MHz &#124; 137GB/s
 
|-
 
|-
| Storage || MicroSD card || 16GB eMMC 5.1 || colspan="2" style="text-align: center;" | 32GB eMMC 5.1
+
| Storage || MicroSD card || 16GB eMMC 5.1 || 32GB eMMC 5.1 || 16GB eMMC 5.1 || 32GB eMMC 5.1
 
|-
 
|-
| Vision || colspan="3" style="text-align: center;" | NVIDIA GPU support (CUDA, VisionWorks, OpenCV) || 7-way VLIW Vision Accelerator
+
| Vision || colspan="3" style="text-align: center;" | NVIDIA GPU support (CUDA, VisionWorks, OpenCV) || colspan="2" style="text-align: center;" | 7-way VLIW Vision Accelerator
 
|-
 
|-
| Encoder || colspan="2" style="text-align: center;" | 4Kp30, (2x) 1080p60, (4x) 1080p30 || 4Kp60, (3x) 4Kp30, (4x) 1080p60, (8x) 1080p30 || (4x) 4Kp60, (8x) 4Kp30, (32x) 1080p30
+
| Encoder || colspan="2" style="text-align: center;" | 4Kp30, (2x) 1080p60, (4x) 1080p30 || 4Kp60, (3x) 4Kp30, (4x) 1080p60, (8x) 1080p30 || (2x) 4Kp30, (6x) 1080p60, (12x) 1080p30 || (4x) 4Kp60, (8x) 4Kp30, (32x) 1080p30
 
|-
 
|-
| Decoder || colspan="2" style="text-align: center;" | 4Kp60, (2x) 4Kp30, (4x) 1080p60, (8x) 1080p30 || (2x) 4Kp60, (4x) 4Kp30, (7x) 1080p60 || (2x) 8Kp30, (6x) 4Kp60, (12x) 4Kp30
+
| Decoder || colspan="2" style="text-align: center;" | 4Kp60, (2x) 4Kp30, (4x) 1080p60, (8x) 1080p30 || (2x) 4Kp60, (4x) 4Kp30, (7x) 1080p60 || (2x) 4Kp60, (4x) 4Kp30, (12x) 1080p60 || (2x) 8Kp30, (6x) 4Kp60, (12x) 4Kp30
 
|-
 
|-
| Camera || colspan="2" style="text-align: center;" | 12 lanes MIPI CSI-2 &#124; 1.5 Gbps per lane || 12 lanes MIPI CSI-2 &#124; 2.5 Gbps per lane || 16 lanes MIPI CSI-2 &#124; 6.8125Gbps per lane
+
| Camera || colspan="2" style="text-align: center;" | 12 lanes MIPI CSI-2 &#124; 1.5 Gbps per lane || 12 lanes MIPI CSI-2 &#124; 2.5 Gbps per lane || 14 lanes MIPI CSI-2 &#124; 2.5 Gbps per lane || 16 lanes MIPI CSI-2 &#124; 6.8125Gbps per lane
 
|-
 
|-
 
|| Display
 
|| Display
| colspan="3" style="text-align: center;" | 2x HDMI 2.0 / DP 1.2 / eDP 1.2 &#124; 2x MIPI DSI || (3x) eDP 1.4 / DP 1.2 / HDMI 2.0 @ 4Kp60
+
| colspan="3" style="text-align: center;" | 2x HDMI 2.0 / DP 1.2 / eDP 1.2 &#124; 2x MIPI DSI || (2x) DP 1.4 / eDP 1.4 / HDMI 2.0 @ 4Kp60 || (3x) eDP 1.4 / DP 1.2 / HDMI 2.0 @ 4Kp60
 
|-
 
|-
| Wireless || M.2 Key-E site on carrier || 802.11a/b/g/n/ac 2×2 867Mbps &#124; Bluetooth 4.0 || 802.11a/b/g/n/ac 2×2 867Mbps &#124; Bluetooth 4.1 || M.2 Key-E site on carrier
+
| Wireless || M.2 Key-E site on carrier || 802.11a/b/g/n/ac 2×2 867Mbps &#124; Bluetooth 4.0 || 802.11a/b/g/n/ac 2×2 867Mbps &#124; Bluetooth 4.1 || colspan="2" style="text-align: center;" | M.2 Key-E site on carrier
 
|-
 
|-
 
|| Ethernet
 
|| Ethernet
| colspan="4" style="text-align: center;" | 10/100/1000 BASE-T Ethernet
+
| colspan="5" style="text-align: center;" | 10/100/1000 BASE-T Ethernet
 
|-
 
|-
 
|| USB || (4x) USB 3.0 + Micro-USB 2.0
 
|| USB || (4x) USB 3.0 + Micro-USB 2.0
| colspan="2" style="text-align: center;" | USB 3.0 + USB 2.0 || (3x) USB 3.1 + (4x) USB 2.0
+
| colspan="2" style="text-align: center;" | USB 3.0 + USB 2.0 || USB 3.1 + (3x) USB 2.0 || (3x) USB 3.1 + (4x) USB 2.0
 
|-
 
|-
| PCIe || PCIe Gen 2 x1/x2/x4 || PCIe Gen 2 x5 &#124; 1×4 + 1x1 || PCIe Gen 2 x5 &#124; 1×4 + 1×1 or 2×1 + 1×2 || PCIe Gen 4 x16 &#124; 1x8 + 1x4 + 1x2 + 2x1
+
| PCIe || PCIe Gen 2 x1/x2/x4 || PCIe Gen 2 x5 &#124; 1×4 + 1x1 || PCIe Gen 2 x5 &#124; 1×4 + 1×1 or 2×1 + 1×2 || PCIe x5 &#124; 1x4 (Gen 4) + 1x1 (Gen 3) || PCIe Gen 4 x16 &#124; 1x8 + 1x4 + 1x2 + 2x1
 
|-
 
|-
 
| CAN  
 
| CAN  
| colspan="2" style="text-align: center;" | Not Supported || colspan="2" style="text-align: center;" | Dual CAN bus controller
+
| colspan="2" style="text-align: center;" | Not Supported || Dual CAN bus controller || Single CAN bus controller || Dual CAN bus controller
 
|-
 
|-
 
|| Misc IO
 
|| Misc IO
| colspan="4" style="text-align: center;" | UART, SPI, I2C, I2S, GPIOs
+
| colspan="5" style="text-align: center;" | UART, SPI, I2C, I2S, GPIOs
 
|-
 
|-
 
|| Socket || 260-pin edge connector, 45x70mm  
 
|| Socket || 260-pin edge connector, 45x70mm  
| colspan="2" style="text-align: center;" | 400-pin board-to-board connector, 50x87mm || 699-pin board-to-board connector, 100x87mm
+
| colspan="2" style="text-align: center;" | 400-pin board-to-board connector, 50x87mm || 260-pin edge connector, 45x70mm || 699-pin board-to-board connector, 100x87mm
 
|-
 
|-
 
|| Thermals
 
|| Thermals
| colspan="4" style="text-align: center;" | -25°C to 80°C
+
| colspan="5" style="text-align: center;" | -25°C to 80°C
 
|-
 
|-
| Power || 5/10W || 10W || 7.5W || 10/15/30W
+
| Power || 5/10W || 10W || 7.5W || 10/15W || 10/15/30W
 
|-
 
|-
| Perf || 472 GFLOPS || 1 TFLOPS || 1.3 TFLOPS || 32 TeraOPS
+
| Perf || 472 GFLOPS || 1 TFLOPS || 1.3 TFLOPS || 21 TeraOPS || 32 TeraOPS
 
|}
 
|}
 +
 +
== Software Support ==
 +
NVIDIA Jetson production modules and developer kits are all supported by the same '''[https://developer.nvidia.com/embedded/develop/software NVIDIA software stack]''', enabling you to develop once and deploy everywhere. '''[https://developer.nvidia.com/embedded/jetpack JetPack SDK]''' includes the latest [[Jetson/L4T|Jetson Linux Driver Package (L4T)]] with Linux operating system and CUDA-X accelerated libraries and APIs for AI Edge application development. It also includes samples, documentation, and developer tools for both host computer and developer kit, and supports higher level SDKs such as DeepStream for streaming video analytics and Isaac for robotics.
 +
 +
[[File:Jetson Software.png|800px|Jetson Software]]
 +
 +
=== JetPack Components ===
 +
<div style="width:45%;column-count:2;-moz-column-count:2;-webkit-column-count:2">
 +
* [[Jetson/L4T|NVIDIA Jetson Linux (L4T)]]
 +
* [https://developer.nvidia.com/cuda-toolkit CUDA Toolkit]
 +
* [https://developer.nvidia.com/cudnn cuDNN]
 +
* [https://developer.nvidia.com/tensorrt TensorRT]
 +
* [https://developer.nvidia.com/embedded/visionworks VisionWorks]
 +
* [https://developer.nvidia.com/deepstream-sdk DeepStream]
 +
* OpenCV
 +
* OpenGL
 +
* Vulkan
 +
* V4L2 extensions
 +
* GStreamer extensions
 +
* [https://docs.nvidia.com/jetson/l4t-multimedia/index.html L4T Multimedia API]
 +
* [https://developer.nvidia.com/nsight-systems NVIDIA Nsight Systems]
 +
* [https://developer.nvidia.com/nsight-graphics NVIDIA Nsight Graphics]
 +
* [https://developer.nvidia.com/nsight-compute NVIDIA Nsight Compute]
 +
</div>
 +
 +
See [https://docs.nvidia.com/jetson/ docs.nvidia.com/jetson] for online documentation about JetPack. <br />
 +
See [https://developer.nvidia.com/jetpack developer.nvidia.com/jetpack] to download the latest JetPack.
 +
See these
 +
 +
=== Jetson Zoo ===
 +
 +
The '''[[Jetson Zoo]]''' is a repository of open-source frameworks and packages that can be installed on Jetson, in addition pre-trained DNN models.  It provides instructions and pre-built binary installers for popular Machine Learning frameworks such as TensorFlow and PyTorch.
 +
 +
== Ecosystem Products & Cameras ==
 +
 +
The '''[https://developer.nvidia.com/embedded/community/ecosystem Jetson Ecosystem]''' includes a diverse set of companies producing add-ons, accessories, sensors, and software for Jetson such as carrier boards, enclosures, cameras, and custom design services.  For more info, see the directory of [https://developer.nvidia.com/embedded/jetson-partner-supported-cameras Supported Cameras] and [https://developer.nvidia.com/embedded/community/jetson-partner-products Production Systems].  For those interested in real-time Linux support, see [https://www.concurrent-rt.com/redhawk-linux-nvidia-jetson-support/ Concurrent RedHawk].
 +
 +
Also, each Jetson wiki page includes a list of ecosystem products that are compatible with it:
 +
 +
* [https://elinux.org/Jetson_TX1#Ecosystem_Products Jetson TX1 Ecosystem Products]
 +
* [https://elinux.org/Jetson_TX2#Ecosystem_Products Jetson TX2 Ecosystem Products]
 +
* [https://elinux.org/Jetson_Nano#Ecosystem_Products_and_Sensors Jetson Nano Ecosystem Products]
 +
* [https://elinux.org/Jetson_Xavier_NX#Ecosystem_Products_.26_Cameras Jetson Xavier NX Ecosystem Products]
 +
* [https://elinux.org/Jetson_AGX_Xavier#Ecosystem_Products_.26_Cameras Jetson AGX Xavier Ecosystem Products]

Revision as of 12:30, 20 October 2020

The NVIDIA Jetson line of embedded Linux AI and computer vision compute modules and devkits:

Here are some quick links and references to get started:


NVIDIA Jetson Modules

Features Jetson Nano Jetson TX1 Jetson TX2 series Jetson Xavier NX Jetson AGX Xavier series
Jetson-Nano-Compute-Module-400px.png
NVIDIA Jetson TX1 module.jpg
NVIDIA JTX2 Module 400px.png
JetsonXavierNX-Module TopDown 400px.png
Xavier-module-topdown-alpha-300px.png
CPU ARM Cortex-A57 (quad-core) @ 1.43GHz ARM Cortex-A57 (quad-core) @ 1.73GHz ARM Cortex-A57 (quad-core) @ 2GHz +

NVIDIA Denver2 (dual-core) @ 2GHz

NVIDIA Carmel ARMv8.2 (6-core) @ 1.4GHz

(6MB L2 + 4MB L3)

NVIDIA Carmel ARMv8.2 (8-core) @ 2.26GHz

(4x2MB L2 + 4MB L3)

GPU 128-core NVIDIA Maxwell @ 921MHz 256-core NVIDIA Maxwell @ 998MHz 256-core NVIDIA Pascal @ 1300MHz 384-core Volta @ 1100MHz + 48 Tensor Cores 512-core Volta @ 1377 MHz + 64 Tensor Cores
DL NVIDIA GPU support (CUDA, cuDNN, TensorRT) dual NVIDIA Deep Learning Accelerators
Memory 4GB 64-bit LPDDR4 @ 1600MHz | 25.6 GB/s 8GB 128-bit LPDDR4 @ 1866Mhz | 58.3 GB/s 8GB 128-bit LPDDR4x @ 1600MHz | 51.2GB/s 32GB 256-bit LPDDR4x @ 2133MHz | 137GB/s
Storage MicroSD card 16GB eMMC 5.1 32GB eMMC 5.1 16GB eMMC 5.1 32GB eMMC 5.1
Vision NVIDIA GPU support (CUDA, VisionWorks, OpenCV) 7-way VLIW Vision Accelerator
Encoder 4Kp30, (2x) 1080p60, (4x) 1080p30 4Kp60, (3x) 4Kp30, (4x) 1080p60, (8x) 1080p30 (2x) 4Kp30, (6x) 1080p60, (12x) 1080p30 (4x) 4Kp60, (8x) 4Kp30, (32x) 1080p30
Decoder 4Kp60, (2x) 4Kp30, (4x) 1080p60, (8x) 1080p30 (2x) 4Kp60, (4x) 4Kp30, (7x) 1080p60 (2x) 4Kp60, (4x) 4Kp30, (12x) 1080p60 (2x) 8Kp30, (6x) 4Kp60, (12x) 4Kp30
Camera 12 lanes MIPI CSI-2 | 1.5 Gbps per lane 12 lanes MIPI CSI-2 | 2.5 Gbps per lane 14 lanes MIPI CSI-2 | 2.5 Gbps per lane 16 lanes MIPI CSI-2 | 6.8125Gbps per lane
Display 2x HDMI 2.0 / DP 1.2 / eDP 1.2 | 2x MIPI DSI (2x) DP 1.4 / eDP 1.4 / HDMI 2.0 @ 4Kp60 (3x) eDP 1.4 / DP 1.2 / HDMI 2.0 @ 4Kp60
Wireless M.2 Key-E site on carrier 802.11a/b/g/n/ac 2×2 867Mbps | Bluetooth 4.0 802.11a/b/g/n/ac 2×2 867Mbps | Bluetooth 4.1 M.2 Key-E site on carrier
Ethernet 10/100/1000 BASE-T Ethernet
USB (4x) USB 3.0 + Micro-USB 2.0 USB 3.0 + USB 2.0 USB 3.1 + (3x) USB 2.0 (3x) USB 3.1 + (4x) USB 2.0
PCIe PCIe Gen 2 x1/x2/x4 PCIe Gen 2 x5 | 1×4 + 1x1 PCIe Gen 2 x5 | 1×4 + 1×1 or 2×1 + 1×2 PCIe x5 | 1x4 (Gen 4) + 1x1 (Gen 3) PCIe Gen 4 x16 | 1x8 + 1x4 + 1x2 + 2x1
CAN Not Supported Dual CAN bus controller Single CAN bus controller Dual CAN bus controller
Misc IO UART, SPI, I2C, I2S, GPIOs
Socket 260-pin edge connector, 45x70mm 400-pin board-to-board connector, 50x87mm 260-pin edge connector, 45x70mm 699-pin board-to-board connector, 100x87mm
Thermals -25°C to 80°C
Power 5/10W 10W 7.5W 10/15W 10/15/30W
Perf 472 GFLOPS 1 TFLOPS 1.3 TFLOPS 21 TeraOPS 32 TeraOPS

Software Support

NVIDIA Jetson production modules and developer kits are all supported by the same NVIDIA software stack, enabling you to develop once and deploy everywhere. JetPack SDK includes the latest Jetson Linux Driver Package (L4T) with Linux operating system and CUDA-X accelerated libraries and APIs for AI Edge application development. It also includes samples, documentation, and developer tools for both host computer and developer kit, and supports higher level SDKs such as DeepStream for streaming video analytics and Isaac for robotics.

Jetson Software

JetPack Components

See docs.nvidia.com/jetson for online documentation about JetPack.
See developer.nvidia.com/jetpack to download the latest JetPack. See these

Jetson Zoo

The Jetson Zoo is a repository of open-source frameworks and packages that can be installed on Jetson, in addition pre-trained DNN models. It provides instructions and pre-built binary installers for popular Machine Learning frameworks such as TensorFlow and PyTorch.

Ecosystem Products & Cameras

The Jetson Ecosystem includes a diverse set of companies producing add-ons, accessories, sensors, and software for Jetson such as carrier boards, enclosures, cameras, and custom design services. For more info, see the directory of Supported Cameras and Production Systems. For those interested in real-time Linux support, see Concurrent RedHawk.

Also, each Jetson wiki page includes a list of ecosystem products that are compatible with it: