| Category | Specification |
|---|
| Product Name | NVIDIA T5000 Jetson Thor Module (900-13834-0080-000) |
| AI Performance | Up to 2070 TFLOPS (FP4—Sparse) |
| GPU | NVIDIA Blackwell architecture with 2560 CUDA cores & 96 fifth-gen Tensor Cores |
| Multi-Instance GPU | Supported (MIG with 10 TPCs) |
| CPU | 14-core Arm® Neoverse-V3AE 64-bit CPU |
| CPU Frequency | Up to ~2.6 GHz |
| Memory | 128 GB 256-bit LPDDR5X (~273 GB/s) |
| Video Encode | Multi-stream support: up to 6×4K60 or 24×1080p60 and more |
| Video Decode | Multi-stream up to 4×8K30, 10×4K60, 46×1080p60, etc. |
| Networking | 4× 25 GbE interfaces on extension/carrier |
| USB | Up to 3× USB 3.2 + 4× USB 2.0 |
| Display Output | Up to 4× HDMI 2.1 / DisplayPort 1.4a shared outputs |
| I/O Interfaces | UART, CAN, I2C, SPI, I2S/AUDIO HUB, DMIS, PWM, GPIO |
| PCIe Support | PCIe Gen5 lanes for high-speed ports or NVMe |
| Power Consumption | Configurable 40 W–130 W |
| Module Form Factor | 100 × 87 mm, 699-pin B2B connector |
| Typical Use | Physical AI, robotics, multi-sensor fusion, edge AI |
Product Description
The NVIDIA T5000 Jetson Thor Module (900-13834-0080-000) is NVIDIA’s highest-end embedded AI compute module designed for breakthrough performance in physical AI, robotics, and multimodal edge intelligence. Built on the Blackwell GPU architecture, this module delivers up to 2070 TFLOPS of AI throughput using 2560 CUDA cores and 96 fifth-generation Tensor Cores, combined with 14 Arm® Neoverse-V3AE 64-bit CPU cores for compute-intensive workloads. With 128 GB of LPDDR5X memory and extremely high memory bandwidth (~273 GB/s), it can handle the largest neural networks, real-time sensor fusion, and multimodal reasoning tasks directly at the edge.
This module supports advanced video encode/decode engines capable of handling multiple high-resolution streams simultaneously and offers extensive I/O options including high-speed USB, multi-lane PCIe Gen5 support for NVMe SSDs, and rich industrial interfaces like CAN, UART, SPI, and I2C. Networking capabilities include multiple 25 GbE links on a carrier board, facilitating ultra-fast data movement for sensor aggregation or distributed AI workloads.
Configured for flexible power profiles from 40 W to 130 W, the Jetson T5000 is suited for applications such as humanoid robotics, autonomous machines, edge servers, and industrial AI appliances where both compute density and real-time responsiveness are required. Its compact form factor and broad software support via NVIDIA’s Jetson Linux and AI stacks make it ideal for developers and OEMs building next-generation embedded AI systems.
-750x250.png "Datacom Technologies (BD)"){kind=logo}
