NVIDIA Expands Jetson Thor Platform with New T3000 and T2000 Modules
NVIDIA announced on July 15 the expansion of its Jetson Thor computing platform family, adding T3000 and T2000 modules to the existing T5000 and T4000 lineup. Targeting the mainstream mid-to-high-end segment, the T3000 delivers 865 TFLOPS and the T2000 offers 400 TFLOPS of FP4 sparse AI performance. The new modules bridge the gap between the previous-generation AGX Orin and the flagship Thor modules, while new memory optimization software helps developers reduce deployment costs.
Highlights
- NVIDIA announced the Jetson Thor T3000 and T2000 modules on July 15, expanding the platform beyond the existing T5000 and T4000.
- The T3000 delivers 865 TFLOPS of FP4 sparse AI performance with 32 GB LPDDR5X memory and 273 GB/s bandwidth; the T2000 offers 400 TFLOPS with 16 GB memory.
- Both new modules are positioned to bridge the previous-generation AGX Orin and the flagship Jetson T5000/T4000 in the product lineup.
- New Jetson software memory optimization technology combined with Agentic AI skills enables significant memory savings on Jetson Thor and Orin systems within days.
- The Jetson Thor platform is widely deployed in drones, autonomous robots, and industrial automation edge AI applications.
NVIDIA announced on July 15 the expansion of its Jetson Thor computing platform family, introducing the T3000 and T2000 modules alongside the previously launched T5000 and T4000, delivering a more complete product lineup for the edge AI market.
New Modules Target Mainstream Mid-to-High-End Segment
The T3000 and T2000 are positioned in the mainstream mid-to-high-end segment, offering 865 TFLOPS and 400 TFLOPS of FP4 sparse AI compute performance, respectively. Within the Jetson product family, they serve as a bridge between the previous-generation AGX Orin and the flagship Jetson T5000/T4000 modules.
T3000 Hardware Specifications
The Jetson T3000 integrates the following core specifications:
- 8-core Neoverse Arm CPU
- 32 GB LPDDR5X memory with up to 273 GB/s memory bandwidth
- 25GbE wired networking connectivity
By comparison, the T2000 features 16 GB of memory, making it better suited for cost-sensitive applications.
Multimodal Inference Performance and Cost Advantages
NVIDIA states that the T3000 delivers inference performance comparable to the T5000 in multimodal workloads. For users operating in high-memory-cost environments, migrating to the T3000 can effectively reduce overall deployment costs, striking a better balance between performance and budget.
Memory Optimization Software and Agentic AI Skills
NVIDIA also announced the integration of new Jetson software memory optimization technology with Agentic AI skills. According to the company, developers can achieve significant memory savings on both Jetson Thor and Jetson Orin systems within just a few days, further enhancing deployment flexibility at the edge.
The Jetson Thor platform is widely used in drones, autonomous robots, and industrial automation edge AI applications. This product line expansion is expected to give developers greater hardware flexibility when building next-generation edge AI systems.
原文來源： 查看原文
FAQ
Newsletter
Subscribe to our Low-Altitude Industry Newsletter
Daily curated news on low-altitude economy and drone industry, delivered to your inbox.
Reviewed and published by the LAETimes editorial desk ·
