Tech news in 3 minutes

Noctua NH-U14S DX4677 Quick Look With the Intel Xeon 600 Series

4 d ago

**NVIDIA Expands Jetson Thor Lineup with Lower-Cost T3000 and T2000 Boards for Robotics and Edge AI** NVIDIA this week announced it will expand its Jetson Thor lineup of robotics and edge AI boards with two new mid-grade models—the Jetson Thor T3000 and T2000—set to launch in Q1 2027, offering lower-cost, lower-power alternatives to the existing high-end T4000 and T5000 boards. The move addresses rising component costs, particularly memory prices, that have limited customer adoption of NVIDIA’s Blackwell-based Thor hardware. The T3000 features 8 Arm Neoverse V3AE CPU cores, a 1536-core Blackwell iGPU delivering 865 TFLOPS of sparse FP4 performance (72% of the T4000), 32GB of LPDDR5X-8500 memory with full 237GB/s bandwidth, and 25GbE networking. With a max TDP of ~65W, NVIDIA pitches it as a budget-friendly option for robotics and multimodal AI workloads that are memory bandwidth-bound, offering similar inference performance to the T5000 in such scenarios. An IGX version with functional safety support will also be available. The T2000 is a further cut-down board with 6 CPU cores, 1024 CUDA cores (400 TFLOPS sparse FP4), 16GB of memory on a half-width bus (137GB/s bandwidth), and 10GbE networking. It targets lighter-weight edge AI applications rather than complex robotics. Alongside the hardware, NVIDIA released memory-optimized agent skills to help customers reduce memory consumption, enabling them to move to lower-memory configurations within the same hardware tier. The T3000 and T2000 will join the existing Thor lineup alongside the previous-generation Orin boards (AGX Orin, Orin NX, Orin Nano) for mainstream and entry-level products. Pricing has not been announced.