Fact Checked by 
Nvidia has released a new mini-PC specifically aimed at developers and hobbyists working in the artificial intelligence space. It’s a follow-up to their previous single-board computer (SBC), Jetson Nano released in 2019 and part of the company’s mega push into AI.
Dubbed Jetson Orin Nano Super, it’ll provide improved performance for AI applications, as well as a bump in performance. While there are a lot of SBCs on the market, Nvidia’s is particularly interesting as the company rarely produces its own PC systems.
Nvidia is known for its graphics cards (GPU), and more recently, its AI hardware. The company has exploded in value due to its continued developments in the space, as well as providing a vast majority of companies with the necessary equipment.
This particular board comes with an ARM Cortex A78E processor and a custom Nvidia GPU with 1024 CUDA cores and 32 Tensor Cores. It’s not the most powerful machine in the world, but in the SBC space, it’s reasonably powerful.
It’ll also provide 67 TOPs (trillions of processes per second), which is how AI is measured. For comparison, the SBC Raspberry Pi 5 and its AI kit can provide around 13 TOPs. An M4 Mac Mini can do 38 TOPS.
Nvidia has included a custom version of Linux for projects to run on, complete with Nvidia theming.
Nvidia Jetson Orin Nano performance is impressive – so is its price
In a video breaking down its performance by Dave’s Garage, the Orin was put through its paces with a local large language model (LLM). This is essentially like running your own version of ChatGPT.
Using Meta’s Llama 3.2, it managed to run at 21 tokens a second. A token is the metric used to count how many characters an LLM can produce. OpenAI’s estimates put it at 4 characters per second.
Compared to major PC rigs or workstations that are running LLMs, it’s not much. However, when you consider it only draws 25 watts of energy at its maximum output, it’s very impressive. It manages this by leveraging the CUDA cores, Nvidia’s tech for processing.
The Jetson Orin Nano isn’t going to be someone’s replacement PC. It’s more for development, and eventually to be built into an embedded solution or be the jumping-off point for a larger project.
The Jetson Orin Nano is also interesting thanks to its price point. Nvidia has slashed the price from $499 to $249, making it a cheaper entry point into AI development. Since the pandemic and electronics shortage, as well as cryptocurrency jacking up the price of graphics cards thanks to scarcity, Nvidia has rolled with it.
Its prices on GPUs have inflated, with the upcoming RTX 5090 rumored to be north of $1500 and nearing $2000. Even its lower-end cards, like the RTX 4060, have been criticized for being overtly expensive compared to the competition.
Nvidia has its big CES 2025 event coming up, where it’ll share more about the world of AI and its future for gaming hardware. We’ve yet to see what the Nano can do gaming-wise, however.
