Author: Zen Kuda Nixjoen
WHAT IS IT?
Described by NVIDIA as “the world’s smallest supercomputer” the NVIDIA Jetson Xavier NX is one the latest additions to NVIDIA’s deep learning – accelerator Jetson family. The device is pin-compatible with the Jetson Nano and includes flash storage, DRAM, a GPU, a CPU and PMICs. A Jetson Xavier NX developer kit may resemble an ordinary Single Board Computers (SBC). The only difference is that the Jetson Xavier is much more advanced and packs a huge punch. Let’s get into details.
SPECS AT A GLANCE
CPU: 6-core up-to-1.9GHz NVIDIA Carmel ARMv8.2
6MB L2 and 4MB L3 cache
GPU: NVIDIA Volta with 384 CUDA Cores and 48 Tensor Cores
Accelerators: 2x NVIDIA Deep Learning Accelerators (NVDLA)
RAM: 8GB LPDDR4x
Storage: microSD, M.2 Key M NVMe
USB: 4x USB 3.1, 1x USB 2.0 Micro-B
Connectivity: Gigabit Ethernet, M.2 Key E 802.11ac 2×2 2.4/5GHz, Bluetooth 5.0
Camera Inputs: 2x MIPI CSI-2
GPIO: 40-pin header with UART, SPI, I2C, I2S, and PWM
Video Encode (H.264/H.265): 2x 4k30, 6x 1080p60, 14x 1080p30
Video Decode (H.265): 2x 4k60, 4x 4k30, 12x 1080p60, 32x 1080p30
WHAT’S IN THE BOX?
- NVIDIA Jetson Xavier NX module and reference carrier board
- Quick Start Guide and Support Guide
- Power adapter: AC power brick and power cord
JETSON XAVIER NX vs JETSON NANO
Both the Jetson Xavier NX developer kit and the Jetson Nano are machine learning platforms; however, the Jetson Xavier is different because it is not an entry level device, it is more advanced, powerful and is specifically designed to support applications that require a lot of processing power. Even though the 8GB RAM and M.2 NVMe support make the Jetson Xavier NX a huge upgrade from the Nano, the more notable change is in the form of processing power. CUDA tests show that depending on the application, the Xavier NX is between 2-7 times faster than the Jetson Nano.
Here are some CUDA performance numbers to compare the two:
|Jestson Nano||Xavier NX|
|ConvolutionFFT2D (in secs)||15.1||8.4|
|FastWalshTransform (in secs)||12.2||3.5|
|MatrixMul (in GFlop/s)||30.2||215.25|
The NVIDIA Jetson Xavier NX is based on a system-on-chip which packs six NVIDIA Carmel ARMv8.2 general processing cores. It also contains a Volta GPU delivering 384 CUDA cores and 48 Tensor cores. The Volta GPU was introduced in order to incorporate hardware specifically designed for accelerating deep learning workloads to its more graphics-focused GPU designs. A pair of NVIDIA Deep Learning Accelerator units are also included for increased acceleration. They are designed to run simultaneously with a GPU workload and increase the INT8 performance to 21 TOPS.
Although the hardware is impressive, the fun doesn’t end there. The Xavier NX offers “Cloud Native Computing.” The Cloud Native vision incorporates a technology known as containerization. What this means is that the Xavier NX will be able to run self-contained micro-services in an already defined environment. Containerization allows the development of a self-contained serviced to be carried out without having to upgrade all the other applications. This separation of the OS from the application increasing the options for scaling.
The CPU capabilities of the Jetson Xavier NX are rather disappointing considering its price. In single-threaded benchmarking it’s only slightly faster than the Rockchip RK3399 found on other similar devices at a much cheaper price. However, you can improve performance considerably by changing CPU modes. Switching from the 6-core 15W operation mode to the 2-core 15W operation mode increases the performance in single-thread workload by almost 50%, the only downside is that you lose 4x cores.
In all honesty the NVIDIA Jetson Xavier NX is not for beginners. I do not recommend it. However, for intermediate and advanced users who are seeking good performance and hardware-accelerated machine learning it is a great pick. Although on the pricier side, retailing for $399 on Amazon, it is worth it. The development kit is awesome for product design and development.
As for beginners, before graduating to the level of the Jetson Xavier NX you could learn the basics of machine learning and artificial intelligence on a more beginner friendly device such as a Raspberry Pi. Or instead, you could get yourself a Jetson Nano and enjoy access to the same core platform and a less steep learning curve.
You can get the NVIDIA Jetson Xavier NX here.
We are a participant in the Amazon Services LLC Associates Program, an affiliate advertising program designed to provide a means for us to earn fees by linking to Amazon.com and affiliated sites.