Below are our top picks:
1. NVIDIA Jetson Xavier NX Developer Kit
The Jetson Xavier NX developer kit is an enthusiast-level device with a consumer-level price. It takes the TX2 performance and ups it a notch. According to NVidia, the NX performance matrices outperform TX2 by approximately ten times in as little as 10W. That is sure to please a regular tinkerer. Its capability to develop and test energy-efficient, small form-factor projects with highly precise, multi-modal AI inference opens the gate for new breakthroughs.
The module’s computer has a 6-core NVIDIA Carmel ARM v8.2 CPU, 6 MB L2 + 4 MB L3 cache, 8GB computer memory size, and 16GB hardware disk size. Moreover, Its GPU is based on NVIDIA’s latest Volta architecture with 384 CUDA and 48 Tensor Cores. These are quite beast of specs for a consumer-level.
The only problem with this option is that L4T has a very small support community, which means not much software support. If you need software, you will probably have to build it yourself.
Overall, the NVIDIA Jetson Xavier NX Developer Kit has a power-efficient, compact Jetson Xavier NX module for AI edge devices. It’s a perfect portable solution for tinkerers looking into AI or robotics applications. And not just that, it also works great for entertainment and productivity.
Buy Here: Amazon
2. NVIDIA Jetson Nano 4GB Developer Kit
The second best Nvidia Jeston developer kit on our list is perhaps the most underrated SBC on the market. It delivers excellent performance to run modern AI workloads at an extraordinary size, power, and price point. That makes it a great little computer, especially for machine learning and teaching.
The Jetson Nano is also excellent as a general-purpose Ubuntu 18.04 LTS desktop. While the image is based on the preceding LTS, it’s still one of Nvidia’s more polished image. Even with just 4GBs of memory, it runs exceptionally well. The Nano has a very snappy feel while running a REAL full desktop Linux distribution. Yes, even the 8GB RaspberryPi 4 can’t beat the performance.
And then there’s the main draw: the GPU, programming, and its machine learning toolset. Everything comes pre-installed and pre-configured. You can also add other tools quickly via container images. The only downside of this developer kit is that the Maxwell-based 128 Cuda cores are somewhat outdated. But, hey, as long as they get the job done as a teaching tool, it’s all good.
The key takeaway here is that it’s quite a self-contained setup. If you’re a fan of pie, it’s as easy as pie (pun absolutely intended). Everything takes just 10 minutes to get up and to run. For the price, nothing beats it, especially as an independent learning tool.
Buy Here: Amazon
3. NVIDIA Jetson AGX Xavier Developer Kit (32GB)
While Nano is great, it can be slow for serious developers. The Xavier is Linux ARM64 at its finest. Sure, the AGX Xavier is noticeably costly, but it packs a punch when it comes to performance. And that too on just a 30W power level.
Let’s talk a little about the specs. The board is a nice ARMv8 developer box complete with CUDA, TensorRT, & NVIDIA’s libraries. On the other hand, the module has eight ARM v8.2 “Carmel” processor cores, 512-cores Volta GPU (with tensor cores),16GBs of LPDDR4x memory, 32GBs of eMMC5.1 storage, 2 NVDLA deep learning accelerators, and a seven-way VLIW vision processor. That’s some impressive firepower.
However, we love this kit because it comes with a “quiet” mode on. Because of this, it passively cools down with negligible throttling.
We have one minor gripe, though. in case of an electrical event, this unit does not automatically have power. You can jumper in some pins to get it to auto power on, but we didn’t try this method during our trial run. Overall, If you are training networks or doing some video AI, testing robotics, and other autonomous machines, AGX Xavier is the Jetson for you.
Buy Here: Amazon
4. NVIDIA Jetson TX2 Development Kit
The Jetson TX2 is another developer’s kit for the experts that comes nicely optimized for various AI forms. It’s rather hard for beginners to get started with this kit. But even if you’ve never trained a deep learning net, there’s plenty to appreciate here.
As for specs, the TX2 has a dual-Core NVIDIA Denver 2 CPU and Quad-Core ARM Cortex-A57 MPCore processor, 4 GB 128-bit LPDDR4 memory, 256-core NVIDIA’s Pascal GPU, and a 16 GB eMMC 5.1 storage. That translates to a performance three times faster than Raspberry 3. (The Jetson TX2 Development Kit came out in 2017).
To test its performance, we ran deep nets for image recognition using Tensorflow. Initially, the nets were trained using Amazon AWS. The nets transferred flawlessly to the TX2. But, of course, with some effort. This is not a toy. This is a pro engineering tool. It’s is a module that powers a self-driving car or a video-capturing quadcopter. These tasks demand fast processing capability with a low power budget.
That’s why there is no other tool like this. If you need a fast CPU that only draws 15 Watts, NVIDIA Jetson TX2 Development Kit seems like a logical choice.
Buy Here: Amazon
5. NVIDIA Jetson TK1 Development Kit
Finally, we have one of the oldest NVIDIA Jetson developer Kit. Of course, it’s still worth looking into in 2021. If you test the waters with Nvidia developer kits, the TK1 is still a great entry point and an inexpensive GPU platform for development.
The TK1 is built around NVIDIA’s Tegra K1 SOC. It uses an NVIDIA Kepler computing core that feels a little outdated today. However, it’s still a full NVIDIA CUDA platform that lets you develop and deploy compute-intensive systems for computer vision, robotics, agriculture, medicine, and more.
The footprint of this model is rather big and tall. Even though the system runs cool, the fan itself is placed quite high on the kit. As this is an older model, the RAM is also shared between the GPU and CPU, limiting its performance.
Like the options mentioned earlier, NVIDIA offers the entire BSP and software stack for this model. This includes CUDA, OpenGL 4.4, and NVIDIA’s Vision Works kit. With a complete development suite, plus out-of-the-box compatibility and support for cameras and other peripherals, NVIDIA gives you a nice introductory solution to get started with embedded systems.
Buy Here: Amazon
Buyer’s Guide for the Best NVIDIA Jetson Developer Kit
NVIDIA has no dearth of Jetson Developer Kits. So keep these crucial factors in mind when looking into the market for a purchase:
Footprint
The first thing to notice when you unpack the best NVIDIA Jetson Developer Kit should be your first consideration: the footprint. How much space does the kit need in your workspace? Is it heavy? Is the fan placed too high? Kits with a larger footprint aren’t portable. If your kid isn’t portable, then what’s the point of getting one in the first place?
Ease of Use
The developer kit should be ready to use out of the box. It shouldn’t put any limitation on your curiosity to explore AI with various sensors and peripherals.
Support
The next feature you should look into is the support and compatibility. First and foremost is the support for modern AI frameworks like TensorFlow, PyTorch, and MXNet. It should also support as many popular sensors in the AI community as possible. Having a large and vibrant developer community also comes in handy. You can then troubleshoot problems, share open-source projects as well as real-world applications.
How to use (or even use?)
After you receive your product, load the OS, and connect to the internet. Then open a browser text editor, and let it sit there for approx 6 hours or more. Letting it rest overnight is usually better. Afterwards, if there is no sign of rebooting, you should be good to go. However, if you notice rebooting, see if there is any kernel crash file under the “/var/log”? Open it and search for “kernel oops”. If it does show up, don’t waste your energies or time. Just return the product!
Final Thoughts
AI at the edge can unlock incredible potential in everything. Whether it’s healthcare, manufacturing, or agriculture, using the best NVIDIA Jetson developer Kit can make your task-at-hand incredibly rewarding. These kits reduce your software developmental costs and provide a scalable AI strategy for your autonomous machines. We hope this article helped you make up your mind. That’s all for now. Thank you for reading.
from Linux Hint https://ift.tt/2LABHL0
0 Comments