NVIDIA Corporation – Jetson AGX Thor

Core Computing Performance: AI Performance: 2000 TOPS (INT8), supports sparse computation acceleration, delivering an 8x performance boost over the previous-generation Jetson AGX Orin. GPU: Based on the next-generation Blackwell architecture, integrating new-generation CUDA cores and RT ray tracing units. CPU: NVIDIA Grace CPU (ARM v9 architecture) with multi-core performance optimised for real-time task processing. Key Specifications Power Efficiency: Optimised power consumption range of 60W–100W (configurable), suitable for automotive and edge devices. Memory: LPDDR5X with 256GB/s bandwidth, enabling low-latency inference for large models. Interfaces: 16x PCIe Gen5 lanes supporting synchronous input from multiple sensors (lidar/cameras). 12x 10Gbps Ethernet ports meeting high-throughput communication demands.