Drive PX 2 from NVIDIA house some amazing components. First of all, it has 12 CPU cores, 8 teraflops worth of processing power which similar to about 6 Titan X video cards. With this design, it is not surprising that it can achieve up to 24 trillion operations per second which should be enough for powering self driving cars.
As the road environment keeps on changing, with pedestrians and cyclists, therefore, deep learning is also an essential component for self driving car on top of all the superb technology. DIGITS is a deep learning platform introduced by NVIDIA so that the self driving cars will be able to learn by itself as well as share it with other self driving cars through a cloud-based network.
DIGITS has its own network named Drivenet which basically consists of nine separate neural networks embedded within each other. Getting information from the network will require 40 billion operations which no doubt needs a lot of computing power to do it and surely store a lot of information within it.
Drivenet able to identify objects based on their colour code. By far, it is programmed to be able to identify five different classes of objects including pedestrians and motorcyclists. All these are particularly important for self driving vehicles.
Volvo will be the first automaker which going to use PX2 for their self driving car. It is a good start after all.