Innopolis installed the supercomputer for algorithm elaboration of the movement of pilotless KAMAZ

In March, 2018 information on what in University Innopolis was set supercomputer Nvidia DGX-1 based on graphic appeared processors. Using the equipment specialists of the Russian IT university are going to develop algorithms of the movement pilotless KAMAZ.

University Innopolis

According to the university, in the computer — 8 graphic processors Tesla V100: 5120 cores and 21 billion transistors inside. They help the machine to reach performance in the 960th TFLOPS (960 trillion operations per second) at accomplishment of calculations from a floating comma. Experts noted that in work with algorithms Deep Learning this device is 96 times faster than the server computer based on the central processor. Nvidia DGX-1 is capable to reduce time necessary on calculations and training of neural networks, from days and weeks to hours.

At the same time the size supercomputer is equal to normal server. In the device also two 20-core Intel Xeon E5 v4 processors-2698 with a frequency of 2.2 GHz, operating system Ubuntu Linux and access to Nvidia GPU Cloud — to the software catalog for deep learning and high-performance computing optimized under graphic processors.

The supercomputer will be involved in solving of tasks of machine learning for autonomous driving — the project of university on creation of a pilotless system for KAMAZ with aero investigation of the territory the drone and for the tender "Winter City" for which we create the unmanned vehicle adapted to the movement in the winter without marking at different times day — the head of the center of development of robotics of the University told Innopolis Alexander Klimchik.[1]

Notes