The dawn of the new Jetson… Nvidia Jetson TX2 has come

Well, it’s been a while since I published the latest news about MyzharBot project, what’s better than starting again from a really big news?
Nvidia has presented the new Jetson module, Nvidia Jetson TX2 and it’s ready to get on board of MyzharBot

What’s about the new “embedded monster” launched by Nvidia? While the step from the Jetson TK1 to the Jetson TK2 has been very big, this time Nvidia put the focus on the continuity: the new Jetson TX2 is a module really similar to the Jetson TX1, with the same dimensions and, more important, pin to pin compatible with the former one. We can continue to use the carrier boards that we are using with Jetson TX1, but with more and more computational power.

Nvidia Jetson TX2

Nvidia Jetson TX2

So what’s new respect to the Jetson TX1… sit down and read carefully, it’s impressive:

  • GPU: from Maxwell to Pascal up to 1.3Ghz
  • CPU: added dual core 64bit Denver (up to 2.0Ghz) to the existing four cores 64 bit A57
  • Memory: from 4 GB 64bit LPDDR4 to 8 GB 128bit LPDDR4
  • Storage: doubled the eMMC to 32 GB
  • Video Encode: from 2160p @ 30 FPS to 2160p @ 60 FPS 

Hence what can we do with the new Jetson TX2? We can run 6 DNNs, while making 2x Object Tracking and decoding two 4K 30FPS streams, composing the result and encoding it to 2 streams H265… quite impressive!!!

NVidia has not only worked  on the hardware, new software is coming too. The new Jetpack v3.0 will be released soon and it is highly focused on Artificial Intelligence:

  • Deep Learning: TensorRT 1.0 GA, cuDNN 5.1, DIGITS Workflow
  • Computer Vision: VisionWorks 1.6, OpenCV4Tegra 2.4.13 (OpenCV 3 is available compiling from source Nvidia submitted a lot of improvement for Tegra to the main stream)
  • GPU Compute: CUDA 8.0.64, CUDA Libs
  • Multimedia: ISP Support, Camera Imaging, Video CODEC
  • More: ROS compatibility, OpenGL 4.5, OpenGL ES 3.2, OpenGL EGL 1.4, Vulkan 1.0.3, GStreamer, advanced developer tools… and much more

Also connectivity has been improved:

  • 3x USB 2
  • 3x USB 3
  • PCIe 1×4 + 1×1 or 2×1 + 1×2
  • 6x CSI cameras
  • Support for DSI display
  • Support for HDMI display
  • Digital MIC
  • Digital Speakers
  • Gigabit Ethernet
  • Wifi
  • 5x I2C
  • 3x SPI
  • 2x CAN

Nvidia is highly focusing on Artificial Intelligence for Robotics, Automotive and Security, the new Jetson TX2 will really speed up the embedded applications that require high computational power.

Enjoy the unboxing video and a few photos of the development kit… new videos will follow in the next days exploring CUDA and Computer Vision capabilities:

3 Comments:

  1. Pingback: First boot of the Nvidia Jetson TX2 and CUDA test - MyzharBot v3.1.1

  2. Pingback: First boot of the Nvidia Jetson TX2 and CUDA test – .: Robot-Home.it :.

  3. Pingback: Embedded Computer Vision Real Time? Nvidia Jetson TX2 + VisionWorks toolkit – .: Robot-Home.it :.

Leave a Comment