NVIDIA Expands Jetson Thor Family with T3000 and T2000 Modules
NVIDIA announced the expansion of its Jetson Thor computing family with the addition of two new modules, the T3000 and T2000. These modules are aimed at mainstream mid-to-high-end applications, offering enhanced AI computing power.
NVIDIA announced on July 15th that it is expanding its Jetson Thor computing family by introducing the new T3000 and T2000 modules. These additions complement the previously launched T5000 and T4000 models.
The new T3000 and T2000 modules are positioned for the "mainstream mid-to-high-end" segment. The T3000 offers 865 TFLOPS and the T2000 provides 400 TFLOPS of FP4 sparse AI compute performance. They are designed to bridge the gap between the previous generation AGX Orin and the Jetson Thor family's T5000/T4000 modules.
The T3000 module integrates an eight-core Neoverse Arm CPU, 32GB of LPDDR5X memory with 273GB/s bandwidth, and a 25GbE Ethernet connection. The T2000 model is equipped with 16GB of memory. NVIDIA states that the T3000 offers similar inference performance in multimodal workloads as the T5000, and its use can help reduce costs amidst high memory prices.
Additionally, NVIDIA has introduced new memory optimization features and intelligent skills within its Jetson software. These developer tools allow for significant memory savings on Jetson Thor and Jetson Orin systems within days, enhancing performance and efficiency.