Nvidia released the latest member of the Jetson family, the Nano, at the 2019 GTC, with a price tag of just $99. This product has attracted widespread attention. To meet the needs of a broader audience, this article provides a detailed analysis of its various features.
To better help everyone understand the Nano, Nvidia released an introductory video on YouTube:
Nano Basic Parameters:
With this low-cost Jetson tablet, the Nano uses a Tegra chip similar to the Jetson TX1 from a few years ago. The Tegra X1 SoC features a quad-core Cortex-A57 processor and a 128-core NVIDIA Maxwell GPU. One advantage of using the Tegra X1 is that the open-source Linux kernel supports it better than the newly released SoC, and there is even open-source Tegra Maxwell graphics support in the new driver stack.The Jetson Nano also offers 4GB of LPDDR4 memory, Gigabit Ethernet, 12 MIPI channels, and 4 USB ports, capable of driving up to two displays simultaneously. These features, along with the Maxwell GPU, make the performance of the Nano far superior to most Arm developers competing in the sub-$100 market.
Basic Interface Information:
Unlike high-end Jetson boards with eMMC storage, the Jetson Nano relies on a microSD card for storage. The connections on the development kit include four USB 3.0 Type-A ports, HDMI 2.0, DisplayPort 1.2,40-pin header, MIPI CSI camera connector, micro-SD slot, M.2 WiFi slot, and Gigabit Ethernet. The downside of this board is the lack of integrated WiFi, but if you are interested in wireless connectivity, you will indeed need an external card.
The Jetson Nano supports software components from high-end Jetson boards like CUDA and TensorRT; the same JetPack software can also run on the Nano. The “Linux 4 Tegra” on the Jetson Nano targets Ubuntu 18.04 LTS, although we have seen other Linux distributions also increasing support for other Jetson boards.
The Maxell GPU in the NVIDIA Jetson Nano development kit can provide 472 GFLOPs FP16 of computing power, using four Cortex-A57 cores, which is much better than most $100 Arm boards. The 4GB LPDDR4 memory has a bandwidth of up to 25.6 GB/s. Video encoding/decoding supports 4K60hz, with a maximum of 8 1080p30Hz HEVC decoding capabilities, and 4K30Hz 4x 1080p30Hz HEVC encoding capabilities.
Below are the latest test results published by Phoronix:
A series of evaluations were conducted on the following devices under different modes:
– Jetson TX1 Max-P
– Jetson TX2 Max-Q
– Jetson TX2 Max-P
– Jetson AGX Xavier
– Jetson Nano
– Raspberry Pi 3 Model B+
– ASUS Tinker Board
– ODROID XU-4
Due to space limitations, this article only shows part of the content; more detailed information can be found at:
https://www.phoronix.com/scan.php?page=article&item=nvidia-jetson-nano
Undoubtedly, the cost-performance ratio of the Jetson series Nano is very high, and its corresponding inference capability can complete high-quality model inference. The following video showcases Nvidia’s demonstration of the Nano’s applications in various fields.
For further communication, please add the assistant on WeChat (please note your name, organization, and direction).
