Palit Pandora — Compact AI Computer Powered by NVIDIA Jetson Orin NX Super

Specifications

Overview

Palit Microsystems — one of the world's largest graphics card manufacturers by volume — has entered the edge AI market with Pandora, its first AI computer. Powered by the NVIDIA Jetson Orin NX Super system-on-module, Pandora delivers 117 TOPS (8GB) or 157 TOPS (16GB) of AI inference performance in a remarkably compact 123 × 145 × 66 mm chassis weighing just 470 grams. This marks a significant strategic expansion for Palit, extending its graphics technology expertise from GPUs and workstations into purpose-built embedded AI computing.

Pandora is engineered for space-constrained deployments where performance and efficiency are critical. It pairs an active cooling system with a removable base and support for 3D-printed custom shells, making it unusually flexible for prototyping, integration, and specialized projects. Connectivity is comprehensive: dual 1 Gigabit Ethernet ports, four USB ports with OTG support, an HDMI 2.0 output, and audio I/O. Four M.2 slots allow simultaneous expansion for additional SSDs, Wi-Fi modules, and 5G/LTE cellular connectivity.

For robotics and industrial integration, Pandora adds MIPI camera inputs, a 14-pin UART header, CAN bus, I2C, and GPIO — the sensor and control interfaces essential for autonomous vehicles, AMRs, and factory automation. Pre-loaded with the NVIDIA JetPack SDK and fully compatible with NVIDIA Isaac and ROS2, Pandora provides a production-ready software foundation for vision AI, generative AI, and robotics workloads right out of the box.

Key Benefits

Ultra-Compact Form Factor: At 123 × 145 × 66 mm and 470 g, Pandora is among the smallest active-cooled Jetson Orin NX Super computers available, ideal for embedded and edge deployments. Four M.2 Expansion Slots: Simultaneously support NVMe SSDs, Wi-Fi, and 5G/LTE modules for flexible, field-upgradeable configurations. Robotics-Ready I/O: MIPI CSI camera inputs, CAN bus, 14-pin UART, I2C, and GPIO make Pandora a drop-in brain for AMRs, drones, and industrial control. Removable Base & 3D-Printed Shells: Simplified component integration plus customizable enclosures accelerate prototyping and OEM productization. NVIDIA Full Stack: JetPack, CUDA, TensorRT, NVIDIA Isaac, and ROS2 compatibility streamline development from proof-of-concept to production.