Comparing Embedded AI Modules: NVIDIA's Jetson Nano, TX2, and AGX Xavier

If you need high-performance edge computing ― and you want to do the bulk of your processing on-location ― turn your eyes toward the NVIDIA's line of Jetson modules.

We first met this line in late 2015 when NVIDIA introduced the TX1. The manufacturer is currently phasing out the original model, while more advanced successors are here to take its place in the form of the TX2, AGX Xavier, and Nano. Each model fills a different role in terms of computing power, form factor, and interfaces, and we'll outline the difference between the three in this article.

Jetson TX2 ― Embeddable AI

As the direct successor to the TX1, the Jetson is an embeddable computing module, that NVIDIA claims "brings true AI computing at the edge."

Here are some of the TX2's most impressive features:

- AI Performance: 1.3 TOPs

- GPU: 256-core NVIDIA Pascal

- CPU: Dual-Core NVIDIA Denver 2 64-bit processor and quad-core ARM Cortex-A57 MPCore unit

- Power: 7.5W/15W (10W/20W for 'i' version)

- Physical dimensions: 87 x 50mm

- Cost: $399, 8GB version; $299, 4GB version; $749 for industrial module

In addition to the standard TX2 (8 and 4 GB flavors), NVIDIA also created a hardened industrial version known as the TX2i. Compared to its standard cousin, the TX2i can withstand greater vibration, temperature and humidity ranges, and dust. With a ten-year estimated lifespan, it lasts twice as long as the TX2. All of this means a higher, more "industrial" price tag of $749, and slightly elevated power requirements.


900-83310-0001-000 | Jetson TX2 Module

NVIDIA 模块上的系统 - SOM 查看

Jetson Nano ― Compact and Cost-Effective AI

When your design requires a small computing module with a reasonable price tag, the Jetson Nano presents an excellent option. The device is still impressively capable, though its processing power is significantly lower than other modules. The Nano also boasts lower power consumption as well.

Here are of the features that make the Nano stand out:

- AI Performance: 472 GFLOPs

- GPU: 128-core NVIDIA Maxwell

- CPU: Quad-Core ARM Cortex-A57 MPCore

- Power: 5W/10W

- Physical dimensions: 70 x 45mm

- Cost: $129

The development board for the Nano is also much smaller than the TX2 ― while not quite as small as the Raspberry Pi ― at 100 x 80mm, we could put both the Nano and the Pi into the same general size "class." In fact, as illustrated by this small 'JetBot' project,  NVIDIA seems to have built the Nano with the educational/maker/small robot market in mind. While the Jetson Nano module has a projected price tag of $129, the now-available developer kit is only $99, presumably to encourage early experimentation and adoption.


945-13450-0000-000 | Jetson Nano Developer Kit

NVIDIA 嵌入式系统开发板和套件 查看

Jetson AGX Xavier

According to NVIDIA's product FAQ, "Jetson AGX Xavier is something far beyond previous Jetson modules." Given its staggering 32 TOPs performance — nearly 25x the Jetson TX2 — that seems like an easy claim to justify.

Here are some of the Xavier's features:

- AI Performance: 32 TOPs

- GPU: 512-core NVIDIA Volta GPU

- CPU: 8-Core ARM v8.2 64-bit NVIDIA Carmel CPU

- Power: 10W, 15W, 30W

- Physical dimensions: 105 x 105 mm

- Cost: $1,199 for 100 units or more, or $1,099 for 1000 units or more

With power requirements less than an incandescent light bulb, the Xavier has ten times the energy efficiency of the TX2, all while delivering a staggering 20 times the performance. The Xavier is also physically bigger than the TX2 — 100mm x 87mm — and has a more expensive at a list price of $1,199 per unit if you buy 100 units or more. But for advanced edge-computing applications, this should be a small price to pay for a module that will enable a wide range of embedded possibilities.



Sorry, your filter selection returned no results.