1- What is NVIDIA Jetson?
2- Members of NVIDIA Jetson family
3- NVIDIA Jetson Comparison of family
In this post, we will focus on the general purposes, properties and usage areas of the NVIDIA Jetson Family. Also we include NVIDIA Jetson comparison table and NVIDIA Jetson performance comparison with its introductions.
NVIDIA Jetson is a series of embedded computing boards and AI platforms for every autonomous device. Each system is a complete SoM (System-on-Module) that includes a CPU, GPU, DRAM, flash storage, and PMIC.
Jetson family is designed by NVIDIA to speed up machine learning applications with a low power system design, high-performance capability, and flexible form factors to build software-defined intelligent machines for a wide range of edge applications. These qualifications make the NVIDIA Jetson family ideal for all deep learning, AI, and visualization processes.
The entire product family uses a common software stack that facilitates and broadens deployment. NVIDIA equips Jetson platforms with the same Jetpack SDK, which includes a BSP (board support package), Linux OS, and Nvidia CUDA. It works with third-party platforms.
The members of the Jetson family are
- Jetson Nano
- Jetson AGX Xavier
- Jetson Xavier NX
- Jetson TX2
- Jetson TX2 NX
- Jetson TX2 4GB
- Jetson TX2i
Now, you will see NVIDIA Jetson comparison in the table below.
Jetson Nano is the best suitable tool for people who wants to start learning about AI and robotics. Jetson Nano, supported by NVIDIA JetPack that contains CUDA-X software stack, which is used for AI-based products across all industries, enables us to add unbelievable new and advanced features to millions of small, low-power AI systems with just 5 to 10 watts. When we look at the NVIDIA Jetson Comparison table above, we see that Jetson Nano can provide 472 GFLOPS of AI performance in this low power envelope.
The Jetson Nano Developer Kit allows us to run multiple neural networks for applications in parallel and processes lots of high-resolution sensors synchronically. This small and powerful computer can be used for object detection, segmentation, image classification, and speech processing.
Jetson TX2 is a supercomputer that gives extraordinarily high and fast computing, accuracy, and power efficiency with as little as 7.5 Watt, which is 2.5X more performance than Jetson Nano and it can be seen in the NVIDIA Jetson Comparison table above. It's 69.6 mm x 45 mm or 50 mm x 87 mm small form-factor gives advantages to use it on products like drones with NVIDIA Pascal GPU architecture which contains 256 NVIDIA CUDA, 8GB memory and 59.7GB/s of memory bandwidth.
Jetson TX2 NX
Jetson TX2 NX is ideal for AI solutions for several areas with its compact form factor, power-efficiency and it gives a new AI performance for elementary-level products. If we have to make a Jetson comparison, Jetson TX2 NX performs 2.5 times more than Jetson Nano and has same form factor and pin compatibility with Jetson Nano as well as Jetson Xavier NX. TX2 NX’s compact form factor and energy efficiency makes it suitable for different areas such as smart cities, health sector and more.
Jetson TX2 4GB
Jetson TX2 4GB makes possible to run neural networks with two times more computing performance and power efficiency than Jetson TX1( Jetson TX1 is not included in the Jetson Comparison Table). It can work in the range between 7.5W and 15W and be easily integrated into different products. Also, Jetson TX2 4GB has different standard hardware interfaces from low-speed GPIO ports to high-speed PCIe and USB interfaces.
When we look at the NVIDIA Jetson Comparison Table, Jetson TX2i is designed to be ideal for high-performance edges computing devices like robots, machine vision cameras, and portable medical equipment with its durable design and small form. Jetpack context contains all software tools for developing in embedded platforms. Also, it has ROS compatibility, OpenGL, improved developer kits and more.
Jetson AGX Xavier
Jetson AGX Xavier is designed especially for autonomous machines. AGX’s high performance makes it preferable for specific areas such as obstacle detection, sensor fusion, localization and mapping, visual odometry, and path planning algorithms critical to next-generation robots. With the Jetson AGX Xavier module, users can specify operating modes at 10W, 15W, or 30W for their applications, and 100mm x 87mm size assures the major workstation performance at 1/10 the space of a workstation.
The NVIDIA Jetson AGX Xavier Developer Kit is supported by the NVIDIA JetPack and DeepStream SDKs, with the CUDA, cuDNN, and TensorRT software libraries. The Developer Kit powered by the new NVIDIA Xavier processor. It performs 20X more and 10X energy efficient of its predecessor, the NVIDIA Jetson TX2 which can be shown on the NVIDIA Jetson Comparison Table.
Jetson Xavier NX
Jetson Xavier NX enables us to reach a supercomputer performance efficiency level and fits the power of the NVIDIA Xavier SoC into a 70mm x 45mm module which is the same size as Jetson Nano. When we make the NVIDIA Jetson comparison, in the range of the small size modules, Jetson Xavier NX presents magnificent performance and power advantage with a rich set of IOs. This let us run all modern AI networks and frameworks for deep learning, computer vision, multimedia, and more.
To 21 TOPS performance brings it to the ideal position for high-performance computing and AI. Moreover, Xavier NX has 384 NVIDIA CUDA Cores, 48 Tensor Cores, 6 Carmel ARM CPUs, and two NVIDIA Deep Learning Accelerators engines, preferred to run modern neural networks in parallel and process high-resolution data from multiple sensors simultaneously for its advanced features.
Supporting multiple power modes, such as lower power mode, create a chance to give more of your power for sensors and peripherals, and use the entire NVIDIA software stack at the same time.
Thank you for reading our blog post.