NVIDIA has announced a significant upgrade to its Jetson Orin Nano Developer Kit, now rebranded as the Jetson Orin Nano Super Developer Kit. This update is designed to enhance the performance of generative AI applications, providing developers with a robust tool for deploying advanced models. The new kit offers up to 1.7x improved performance and is available at a reduced price of $249, according to NVIDIA.
Key Upgrades and Features
The Jetson Orin Nano Super Developer Kit boasts several enhancements over its predecessor, including an increase in sparse TOPS from 40 to 67 and memory bandwidth from 65 GB/s to 102 GB/s. The CPU clock speed has also been boosted from 1.5 GHz to 1.7 GHz. These improvements are made possible through a new power mode that elevates the performance of the GPU, CPU, and memory.
Developers can upgrade their existing Jetson Orin Nano Developer Kits to the new Super version with a simple software update. This upgrade enables the kit to handle a wide range of large language models (LLMs), visual language models (VLMs), and vision transformers (ViTs), including those with up to 8 billion parameters.
Enhanced AI Model Performance
The Jetson Orin Nano Super Developer Kit supports an array of machine learning frameworks such as HuggingFace Transformers and NVIDIA TensorRT-LLM, making it an ideal choice for deploying scalable solutions across cloud, edge, and PC environments. The kit’s new capabilities allow it to run emerging Transformer-based models, offering up to 70% higher generative AI performance.
Benchmark tests demonstrate significant performance gains across popular LLMs and VLMs, with models like Llama 3.1 8B and Qwen2.5 7B showing improvements of 1.37x and 1.53x, respectively. ViTs also benefit from the upgrade, with models such as clip-vit-base-patch32 achieving a 1.60x performance boost.
Applications and Use Cases
The enhancements in the Jetson Orin Nano Super Developer Kit make it well-suited for a variety of applications, from robotics to multimodal agents. NVIDIA’s Jetson AI Lab provides a platform for developers to explore these possibilities, offering tutorials and pre-built containers for generative AI projects.
In collaboration with Hugging Face, NVIDIA has accelerated robotic research on the LeRobot platform, which runs on the Jetson Orin Nano Super Developer Kit. Developers can also create generative AI chatbots using the kit, leveraging open-source interfaces like Open WebUI.
Future Prospects
In addition to the Jetson Orin Nano series, NVIDIA is also enhancing the performance of its Jetson Orin NX series. Both series now offer increased compute capabilities, essential for running generative AI models on edge devices. NVIDIA has extended the lifecycle of the Jetson Orin through 2032, ensuring long-term support for these powerful platforms.
With these advancements, NVIDIA continues to pave the way for innovation in generative AI and edge computing, providing developers with the tools needed to tackle real-world challenges.
Image source: Shutterstock
Credit: Source link