NVIDIA Unveils Jetson-Thor Computers for Robotics

<p>General-purpose robots and autonomous machines are moving from research labs to real-world mass-market deployment&comma; creating demand for compact&comma; power-efficient AI supercomputers capable of running foundation models at the edge&period;<&sol;p>&NewLine;<p>To meet that need&comma; NVIDIA has introduced the Blackwell-powered T3000 and T2000&comma; new modules based on the <a href&equals;"https&colon;&sol;&sol;www&period;nvidia&period;com&sol;en-us&sol;autonomous-machines&sol;embedded-systems&sol;jetson-thor&sol;" target&equals;"&lowbar;blank" rel&equals;"noopener">NVIDIA Thor<&sol;a> architecture that enable mass-market robotics and edge AI applications at scale&period;<&sol;p>&NewLine;<p><a href&equals;"https&colon;&sol;&sol;www&period;nvidia&period;com&sol;en-us&sol;autonomous-machines&sol;embedded-systems&sol;jetson-thor&sol;" target&equals;"&lowbar;blank" rel&equals;"noopener">Jetson AGX Thor<&sol;a> is powering this next generation of humanoid and robotic systems&comma; with growing adoption across industries&period; Leading companies — including 1X&comma; Agile Robots&comma; Amazon Robotics&comma; Boston Dynamics&comma; FANUC&comma; Hitachi and Techman Robot — are building on the platform&period;<&sol;p>&NewLine;<p>The hardware underpinning those capabilities starts with the Jetson and IGX T3000 modules&comma; which delivers 865 FP4 teraflops of AI compute in a compact form factor roughly half the size and power of the T5000&period; Jetson T3000 combines an NVIDIA Blackwell GPU&comma; an eight-core Neoverse Arm CPU&comma; 32GB of LPDDR5X memory and 273GB&sol;s of memory bandwidth&comma; along with 25 GbE connectivity&period; IGX T3000 delivers the same performance with integrated functional safety while seamlessly running the <a href&equals;"https&colon;&sol;&sol;www&period;nvidia&period;com&sol;en-us&sol;ai-trust-center&sol;halos&sol;robotics&sol;" target&equals;"&lowbar;blank" rel&equals;"noopener">NVIDIA Halos for Robotics<&sol;a> full-stack safety system for robots operating alongside humans&period;<&sol;p>&NewLine;<p>Despite its smaller footprint&comma; the T3000 achieves similar inference performance of the T5000 for multimodal workloads&comma; including large language models&comma; vision language models&comma; vision language action models and world foundation models&period; Migrating to T3000 helps reduce costs amid high memory prices&period;<&sol;p>&NewLine;