In recent years, NVIDIA has focused part of its business on computing for autonomous cars and artificial intelligence. The company just presented the twenty-second most powerful supercomputer in the world, the so-called DGX SuperPOD.
The DGX SuperPOD is a supercomputer born to provide an infrastructure for artificial intelligence that allows the massive demand for the power needed by NVIDIA itself for its autonomous car development program. The system has been built in just three weeks, with 96 NVIDIA DGX-2H, and Mellanox interconnects technology. This set achieves a processing power of 9.4 petaflops and is capable of training a vast number of deep neural networks necessary for autonomous driving.
This system can be acquired by customers who are interested in having this processing power. Training this type of neural network requires hardware that few machines in the world can have. Tesla is already on the game with its Autopilot.
The DGX SuperPOD consists of 1,536 NVIDIA V100 Tensor GPUs with NVIDIA NVSwitch and the Mellanox network. The system is working continuously, optimizing autonomous driving software and training neural networks at a much faster speed than with previous technologies.