Nvidia has recently announced the Jetson AGX Xaiver module. This chip will help make robots more smarter in the near future. Nvidia is aiming to be the brains behind the next-generation autonomous robots therefore the new Jetson AGX Xavier module does exactly that. The Jetson AGX Xavier is supposed to offer 20 times the performance than its predecessor, the Nvidia Jetson TX2. It is also also supposed to be 10 times more energy efficient than the TX2. The Jetson AGX Xavier module is available today from distributors worldwide for $1,100/ unit for quantities of 1,000 units or more. It is said that the AGX Xavier module will be used in all kinds of machinery ranging from delivery robots to inspection drones to medical instruments.
Nvidia states that the Jetson AGX Xavier module has 3 different configurable operating modes, 10-W, 15-W and 30-W. Nvidia claims that the Jetson AGX Xavier module is fit for running in Robots, Computer Vision, medical instruments and autonomous machines. The board of the chip measures 10 x 8.7 cm therefore making it fit in the palm of your hand. The neutral-network output is 32 TOPS this is due to Xavier’s centerpiece a Volta GPU with 512 shader ALU’s and a whopping 64 tensor cores. It contains an eight-core Nvidia Caramel ARM 8.2 processor with 8 MB of L2 cache and 4 MB of L3 cache.This is connected to 16 GB of LPDDR4x on a 256-bit bus. More specifications of the chip are mentioned below.
|GPU||512-core Volta GPU with Tensor Cores|
|CPU||8-core ARM v8.2 64-bit CPU, 8MB L2 + 4MB L3|
|Memory||16GB 256-Bit LPDDR4x | 137GB/s|
|Storage||32GB eMMC 5.1|
|DL Accelerator||(2x) NVDLA Engines*|
|Vision Accelerator||7-way VLIW Vision Processor*|
|Encoder/Decoder||(2x) 4Kp60 | HEVC/(2x) 4Kp60 | 12-Bit Support|
|Size||105 mm x 105 mm|
|Deployment||Module (Jetson AGX Xavier)|
Nvidia’s Jetson AGX Xavier will be supported by Nvidia’s JetPack software. JetPack offers board support package (BSP), an Ubuntu Linux Operating System, Nvidia CUDA and TensorRT software libraries for deep learning. It’s also supported by the NVIDIA DeepStream SDK which is a toolkit for real time situational awareness.
Nvidia has also released a Bundle Kit with the chip. This will of course be bought for development. While customers who are interested in going to commercial production will buy the module.
Nvidia in the past has released TX2 4 GB, TX2 8 GB and the TX2i. These chips are meant for industrial environments. These chips are also considerably cheaper than the Jetson AGX Xavier. With the TX2 4 GB coming in at 299$ per unit.