@linuxdev
Many thanks for answering my novice question. There are definitely a number of people in the development community who could handle that design, but I am probably not the right guy.
My most basic carrier board design would require USB3 and USB2. But other users would require many more I/O connectors). I just checked the 6 layers on the TX1 carrier. I could probably get away with designing a board for a few connectors… maybe.
We realize the pre-order site currently only contains the North America menu items right now, sorry for the inconvenience. The kits will start shipping worldwide close to Dec 20, hopefully sooner if the EU EMI certs (for WiFi/BT) come through early.
I should hear more back on this today/tomorrow, and will let you guys know as soon as I know. Thank you for bearing with us as we iron out a few of these finer (albeit important) post-launch details :) BTW, really excited for FIRST season 2016!
If you check here (https://developer.nvidia.com/embedded/downloads), currently we have posted the complete design files and schematics for the Jetson TX1 carrier board. Very shortly we will be uploading the comprehensive “Carrier Design Guide” which contains every detail about signal routing & signal integrity for the different interfaces. Today, the module databook from the link above contains much of the pin-out info.
Ideally an off-the-shelf vendor will design a miniaturized carrier for resale in the community. Two gold stars and lots of potential volume to whoever designs a mini carrier (and possibly an enclosure) first!
The Jetson TX1 Developer Kit and Reference Carrier are open-source hardware. We share the BOM and design files online (see above). We aren’t by any means making huge margin or anything on the devkit. The devkit is lower volume than the module and thus relatively more expensive to manufacture in small batches. Technically you are free to fabricate your own as build to print. You don’t have to purchase our devkit in order to access our reproducible design files.
Starting in early January, anyone will be able to purchase the Jetson TX1 module through distributors worldwide.
USB2 is actually somewhat picky on 4-layer boards at 480Mb/sec. USB3 running at 5000Mb/sec needs very good construction down to the separation and size of data lanes tuned specifically to the board’s material. Not having a lot of other traces simplifies being able to achieve good USB3 routing, but it still won’t be an “at home” project.
@linuxdev
Thanks again. Personally, my experience is limited to simple PCB designs manufactured through SEEED, OSH Park, etc, along with basic reflow soldering. It looks like I would have to tap into some other folks for help… Maybe we will see some miniaturized carrier designs pop up (as suggested by @dusty_nv). That would be great.
@dusty_nv
Many thanks for the reply. Yes, miniature carrier boards will definitely be helpful for UAV integration!
If you are curious about the preliminary conversations among a few developers and potential users, the links are below.
[url]http://bit.ly/1QkKsT4[/url]
[url]http://bit.ly/1SmNB2m[/url]
Best,
Thomas
Are there any plans to support the TX1 SDK and tools (I’m especially interested in Linux for Tegra + CUDA) on the SHIELD Console? If so then the console would be a nice analog to the Jetson TK1 (particularly for hobbyists). If not, please consider it, NVIDIA!
There’s been preliminary talk of releasing this, but it’s more of a question for https://devtalk.nvidia.com/default/board/158/shield-development/
The Inquirer story makes it sound like Linux on the TX1 will only be supported through Linux4Tegra. That’s not exactly surprising, but it is a deal-breaker for me, and probably a fair number of other folks. Requiring an NVidia-provided, turnkey distribution for access to the GPU is too constraining. I’m disappointed they haven’t chosen to follow the x86 model of closed-source blobs which are reasonably compatible with most Linuxes, and easily buildable for different kernel versions.
The one silver lining, I suppose, is that L4T may continue to receive updates for the TK1, unless NV decide to fork it into TK1- and TX1-specific branches. That means we could see a 4.x kernel for the TK1… some day…
(Moving to 4.x probably means backing out a bunch of proprietary cruft in order not to conflict with the [non-GPU-related] Tegra support now in mainline, so I’m not surprised L4T has been slow to catch up. But that’s my point: We would be vastly better served if NV took a modular approach, reassigning L4T people to the mainlining effort and leveraging as much of that code as possible.)
If you’re listening, NVIDIA: Please, please, consider providing a toolkit with your closed-source userspace components and enough documentation that the community can apply it to modern Debian or Ubuntu. And for the non-userspace components, provide us a choice of kernel versions, including at least one from the latest stable branch. You could do all of this, outsource the other “boring” parts of crafting a distribution to the community, and probably save time in the process.
The $299 1000 off price seems high for the module
Is this the real commercial price and are there plans for lower cost SKUs?
Apart from the somewhat high pricing, this is a really exciting development from nVidia
The $299 is 1-off module price. Eventually the module distributors will handle the volume discounts.
Yes. And I have a Jetson TK1 running Debian Jessie right now. But AFAIK there is no way to install GPU/CUDA support on top; certainly not if you are running a different kernel than L4T.
This is the issue I see… the value proposition for Jetson as an embedded platform is mostly about its high-performance-computing features, but those features are not accessible unless we adopt L4T and the compromises that entails (non-standard distro, old kernel, poor support, etc).
We are working towards reconciling downstream L4T with mainline kernel, ideally permitting wider adoption across distros with full GPU-acceleration. Currently Ubuntu Fedora and SUSE have it working with JTK1. It is a much bigger issue than Jetson to consider an open release of the entire stack. All things considered we try to keep Jetson as flexible as possible and with the highest performance.
Ok, sounds good!
;)
The EU sites have started rolling out. See here:
http://www.nvidia.de/object/jetson-tx1-dev-kit-de.html
http://www.nvidia.co.uk/object/jetson-tx1-dev-kit-uk.html
http://www.nvidia.fr/object/jetson-tx1-dev-kit-fr.html
The preorder links included in the UK and FR pages should go live shortly.
Ok, today we posted some more documents to [url]https://developer.nvidia.com/embedded/downloads/[/url]
Included is now reference carrier spec, thermal design guide, and deep learning whitepaper. More to come.
Here’s a post summarizing the piece on inference with deep learning:
[url]http://devblogs.nvidia.com/parallelforall/inference-next-step-gpu-accelerated-deep-learning/[/url]
The gist is that JTX1, while consuming only 6W, classifies more images per second with Caffe AlexNet than Intel’s high-end i7-6700K (@62.5W)
Hey Dusty,
nice collection of information here. One thing got me confused though, the blog entry says the devkit will be available in Europe starting 20th December, while the german retail page tells the 20th January, which one is correct?
You have earlier mentioned that the 299$ Modules will be available in early January. Does that also apply for the Europe market? Can you aready share the pricing of the module in Euro?
Thanks for posting these! I just noticed that there is a super capacitor as a power source on the carrier for an RTC (C72 in the schematics)! You can see it below the bottom right corner of the module on the board in this picture: http://i1.wp.com/jetsonhacks.com/wp-content/uploads/2015/11/IMG_1133.jpg
This is awesome and makes me very happy with this board and it’s potential for FRC!
Awesome!
Any info of preorder in other EU countries not france or uk… Spain please… ;)
