NVIDIA's TX1 SoC now comes on a credit-card sized module

NVIDIA's press release states that "Jetson TX1 is the first embedded computer designed to process deep neural networks -- computer software that can learn to recognize objects or interpret information." The 3.4x2inch module includes a Tegra X1 ARM Cortex-A57 processor with 256-core NVIDIA Maxwell graphics, 4GB of LPDDR4 memory, 16GB of eMMC storage, 802.11ac WiFi and Bluetooth, and Gigabit Ethernet support.

AnandTech Article: http://www.anandtech.com/show/9779/nvidia-announces-jetson-tx1-tegr...

The Jetson TX1 Development Kit will be available for preorder starting Nov. 12 for $599 in the United States. The kit includes the Jetson TX1 module, a carrier board (pictured below), and a 5MP camera. The stand-alone module will be available in early 2016 (for $299 in bulk).

The Jetson TK1 (not TX1) was released in 2014 to encourage the development of products based on the Tegra K1 processor. However, according to AnandTech, developers were using the Jetson TK1 outright as a production board, choosing to focus on peripheral and software development instead of system hardware development. With the new TX1, all of the I/O connectivity is provided on a carrier board, enabling rapid development on the credit-card sized TX1 module. After development is finished, the TX1 module can be directly deployed in products, such as drones. 

NVIDIA used a drone application to promote the Jetson TX1

https://twitter.com/NVIDIATegra/status/664238535096926208

Views: 43735

Comment by JB on November 11, 2015 at 5:44am

Nice hardware but at a price.


MR60
Comment by Hugues on November 11, 2015 at 6:38am

does not seem competitive versus existing smaller boards that cost 5 to 10 times less and have all of the connectors onboard without a huge carrier board, like ODroid XU4, Raspberry 2, etc...

who's  gonna be interested to pay more for less performance and bigger/heavier boards ?

Comment by Thomas Stone on November 11, 2015 at 8:00am

@Hugues: I like the XU4. :)

Qualcomm/Intel/NVIDIA have definitely revealed their desire to get their chips into the next-gen drones. Qualcomm announced the Snapdragon Flight as a reference platform targeted at drone development. Qualcomm has invested in 3DR. And Yuneec has announced they are releasing a drone based on the Snapdragon Flight in 2016.

The interesting thing about NVIDIA is that they have already released development platforms to the public. This is very helpful for peripheral developers (such as me :). Also, NVIDIA has a flavor of their Jetson TK1 strapped onto the DJI developer quad (Manifold). 

Comment by Fnoop on November 11, 2015 at 8:46am

@Hughes o_O I think you're reading the wrong spec sheet :)

Comment by Patrick Poirier on November 11, 2015 at 9:23am

The race for the flying super computers is on !!

It will be interesting to watch the energy efficiency ratio for these High performance GPUs and the method deployed to  provide better performance per watt.

I am still betting on  FPGA SoCs, like Xilinx Zynq 7000 family on this category 

Comment by Patrick Poirier on November 11, 2015 at 10:03am

Got my answer = Nice !!

The TX1 uses less than 10W of power, suggesting it should be able to stay aloft for quite a while. Nvidia also has some computer vision libraries that will aid the developer in getting started such as OpenVX1.1 implementation with custom extensions.


MR60
Comment by Hugues on November 11, 2015 at 11:12am

@Patrick, Yes it is true that at 10Watts consumptions, this is much less than an Odroid XU4. However relatively to a quadcopter consumption, this is not the main factor for autonomy. Weight is the main factor. And an Odroid (or Raspberry) is much much lighter and smaller to embark on a UAV than a TX1+carrier board.

Also, Odroid types of boards run standard Ubuntu Linux (15.X) on which you can run a standard OpenCV image processing program.

@Fnoop, Odroid XU4 boards have eight cores versus 4 cores (same ARM family). Ok; TX1 is a 64 bits V8 architecture while XU4 is stilla 32 Bits V7 architecure (but have you really seen programs runs faster in 64 bits versus 32 bits ?). The graphic chip is a Mali-T628 MP6(OpenGL ES 3.0/2.0/1.1 and OpenCL 1.1 Full profile) , and as far as memory up to 64GB of eMMC memory...(4x more than the TX1)..all of that for a smaller price (78$ versus 599$)...So ok, the NVIDIA graphic should be more performant (to be verified with only 10W) but the price/perf ratio makes you wonder: You could use 8 XU4 boards in parallel for the same price of a singleTX1.

Comment by c j g on November 11, 2015 at 11:30am

As Patrick and Hugues has mentioned the GPU and the libraries are the parts we have to look closer. After seeing what the TX could achieve in the field of image recognition (Tesla Model X, some deep learning projects), i see some funny projects coming in ... maybe at first in robots/rovers ... 

Comment by Patrick Poirier on November 11, 2015 at 11:33am

Hello Hugues

My concern was more of the HOW you manage the power on the chip. Nvidia has a reputation of being energy hungry ans therefore you need to take this energy out of the CPU. Odroid XU4-wich is an embedded system that I really like - can manage it with a simple PWM fan unit wich is nice but if you have to setup your NVIDIA with  a Gamer Liquid Cooler on the Quad ... Now you are in trouble !!!!... LOL

Comment by Gary McCray on November 11, 2015 at 11:40am

@Hugues,

I thjink the reality is that the multi GPU boards like the TK1 or TX1 are really good for a few tasks because of the matrix style parallel computing.

In particular for us, 3D vision and obstacle avoidance are the main targets.

The more powerful multiCPU boards like the Odroid or RPI2 are more general purpose and more easily programmed for a wide variety of tasks.

Although a group of 8 XU4s is certainly an interesting configuration, I suspect our multicopters will be using a lot more Tegras than that configuration (at least they will be able to have enough battery to compute and get off the ground.

The fact that the Tegras do come with a decent software development system and that there is quite a bit of existing and available vision oriented software available for it make it a really attractive alternative for us.

I just ordered a TK1 for $99.00 (backordered) from the other Blog post on DIYDrones on the Tegra.

If it works out, I am thrilled to have the smaller, more powerful TX1 to be able to upgrade too.

And for real, intelligent 3D vision, $300.00 to $600.00 is not expensive.

Best Regards,

Gary

Comment

You need to be a member of DIY Drones to add comments!

Join DIY Drones

© 2019   Created by Chris Anderson.   Powered by

Badges  |  Report an Issue  |  Terms of Service