Nvidia has introduced the Jetson Orin Nano Super single-board computer for developers of various artificial intelligence systems, including drones, robots, etc. The novelty provides 70% more performance compared to the Jetson Orin Nano and costs half the price. The performance of the Jetson Orin Nano Super — 67 TOPS for $249, a major improvement over the 40 TOPS of the previous product priced at $499.
The Jetson Orin Nano Super is powered by a hexa-core Arm Cortex-A78AE processor clocked at 1.7GHz and comes with 8GB of RAM. The 1020MHz GPU has 1024 CUDA cores and 32 tensor cores, delivering a peak performance of 67 TOPS.
The system can be powered either via the USB Type-C port or via a proprietary connector, in any case, for maximum performance, the power supply must be at least 25 W. The board is equipped with four USB 3.2 Type-A ports operating at 10 Gbps, two MIPI CSI ports compatible with the Raspberry Pi camera, and M.2 2280 and M.2 2230 connectors for connecting SSDs.
The Jetson Orin Nano Super is also equipped with a 40-pin GPIO connector compatible with Raspberry Pi, which allows the use of the Raspberry Pi Hats expansion board. The Jetson Orin Nano Super runs a specialized version of Linux — L4T, created specifically for use on Nvidia hardware. The boot system can be implemented with both SSD and microSD memory card.
According to Nvidia, Jetson Orin Nano Super demonstrates a performance increase over the original Jetson Orin Nano:
- in learning larger language models by 37–63%;
- in machine vision tasks — by 36–104%;
- in image recognition — by 43–69 %.
Nvidia noted that its other current-generation single-board computers, including the «regular» Jetson Orin Nano, will also receive a performance boost of up to 70 percent after the software update. For example, the performance of the Orin NX 16 GB will increase from 100 to 157 TOPS, and the Orin NX 8 GB – from 70 to 117 TOPS.
The Nvidia Jetson Orin Nano Super single-board computer will be available later this month.