Literally two hours ago ended the event by NVIDIA at Computex 2018. Journalists were told about the current success of the company and has demonstrated a new productive system for problems of artificial intelligence and a platform for robots.
Immediately before the press conference, NVIDIA CEO Jen-Hsun Huang (Jensen Huang) has pleased journalists biscuits and sandwiches, and then moved on to his presentation. The first part of the report was dedicated to PC gaming, and head of the company noted that over the past 8 months the market of gaming devices increased significantly, largely due to a significant increase of players in the “battle Royal” (+10 million players in a week).
Then Jen-Hsun Huang focused on the constant increase of computing power, including for tasks of artificial intelligence. After some time, the software will create the new software. Therefore, NVIDIA should be ready to provide the market with powerful systems.
Immediately the head of the company demonstrated DGX-2 – the most productive system in the world for complex problems of artificial intelligence. This is the first dvuhterabaytny system, equipped with 16 connected GPU, which provide five times the performance task deep learning. The system is equipped with the NVIDIA DGX and is based on a scalable architecture with NVIDIA NVSwitch that allows to solve the most complex tasks in the field of artificial intelligence.
After that, Huang also showed HGX-2 – cloud-based server to work with AI. The platform is designed for high-performance computing and algorithms deep learning. In addition, it is able to perform calculations with high precision FP64 and FP32 used in 3D-modeling and scientific research. Also available to work with Int8 and FP16 modes for training neural networks. This configuration provides the necessary computing power to interact with the AI. Cloud server works with NVSwitch interface that allows you to combine up to 16 graphics cards Tesla V100 and turn them into one workstation.
It was followed by a block devoted to the importance of artificial intelligence and how it changes in various spheres of activity. Today, billions of intelligent machines working in industries such as logistics, SELSKOE agriculture, construction, manufacturing, and others. And for these self-driving cars NVIDIA has created a single-chip system Xavier. It contains OCTA-core CPU of its own design NVIDIA Volta GPU with 512 CUDA cores and 20 of tensor kernels designed for calculations occurring in the framework of machine learning and deep learning.
And finally, was presented roboblather NVIDIA Isaac computer-based Jetson Xavier.
Xavier Jetson, judging by the description, based on SoC Xavier, which contains about 9 billion transistors and makes more than 30 TOPS (trillion operations per second). Also, the configuration includes a Volta Tensor Core GPU OCTA core ARM64 CPU, dual accelerator deep learning NVDLA, image processor, processor, vision and video processor. All this allows to make dozens of algorithms simultaneously in real time to process sensors data, odometry, localization and mapping, machine vision and perception, and path planning. Due to this level of computing power of AI, Autonomous machines can perceive the world with superhuman capabilities, to detect and recognize their environment, using various sensors.
Module for developers will be available for $1299 in August and will include software Isaac robotics.