Nvidia Launches Jetson Orin Nano Super: More Power and Reduced Cost for Generative AI | Cpu-z | Gpu vs cpu for gaming | Gpu hardware for gaming | Turtles AI

Nvidia introduces the Jetson Orin Nano Super, a compact development kit designed for developers and hobbyists who want to explore the world of generative AI. With 70% more performance than its predecessor and half the price, it promises to transform home AI projects.

Key Points:

• Increased performance: 67 TOPS and 102 GB/s of bandwidth to handle advanced models.
• Competitive price: now available at $249, versus $499 for its predecessor.
• Updated hardware: Ampere GPU, 6 Arm Cortex-A78AE cores and 8 GB LPDDR5 memory.
• Software versatility: compatibility with Nvidia Isaac, Metropolis and advanced AI tools.

Nvidia has unveiled the Jetson Orin Nano Super, a new iteration of its popular AI development kit. Called "the most accessible generative AI supercomputer," the device promises to be a powerful and affordable platform for experimenting with AI at home or in the office. The heart of the system is a compact module installed on a carrier board, capable of offering advanced I/O connectivity and power. With a design reminiscent of the Raspberry Pi Compute Module, this product is aimed at developers and enthusiasts eager to explore generative AI.

From a technical perspective, the Orin Nano Super stands out for its revamped hardware architecture. It features six Arm Cortex-A78AE cores and an Nvidia GPU based on Ampere technology, equipped with 1024 CUDA cores and 32 tensor cores. Its specifications speak for themselves: 67 TOPS in INT8 calculations, a 70% increase in performance compared to the previous model, and a memory bandwidth of 102 GB/s, supported by 8 GB of LPDDR5 memory. These characteristics make it ideal for those working with large language models, such as Llama 3.1, estimating a generation of about 18-20 tokens per second using a 4-bit quantized version of the model.

From a cost perspective, Nvidia makes this new kit particularly attractive, offering it at $ 249, a figure significantly lower than the $ 499 required for the original model. In terms of connectivity, the Jetson Orin Nano Super does not disappoint either, offering options such as Gigabit Ethernet, DisplayPort, four USB 3.2 Gen 2 Type-A ports, USB-C and two M.2 slots for expansion, useful for integrating cameras or other peripherals. The ability to connect up to four cameras, combined with support for high-end software such as Nvidia Isaac, Metropolis and Holoscan, makes it a valuable tool for robotics and vision processing applications.

Another interesting element is software compatibility. Despite the Arm-based CPU, Nvidia guarantees full support for its development suites. Thanks to the experience accumulated with advanced solutions such as GH200 and GB200, based on Grace CPUs, the kit offers perfect integration with Nvidia libraries and tools for AI. For those already using previous versions of Jetson Orin, the company will release a software update that, according to what is stated, will improve generative AI performance by 70%, aligning it with the new model.

With the Jetson Orin Nano Super, Nvidia once again demonstrates its commitment to offering powerful, accessible and oriented tools for the future of AI.