Artificial Intelligence is much needed in the present era to bring out the best of the best technologies. And in a company like Tesla, AI has an important meaning in the development of autonomous and electric vehicles. Therefore, to speed up the AI workloads the company came up with a new in-house Dojo Supercomputer, and also unveiled D1 Dojo custom application-specific integrated circuit (ASIC) for AI training at its AI Day Event, where Elon Musk also unveiled the “Tesla Bot.“
The Dojo SuperComputer system, which automaker claims to be the fastest AI training machine in the world, has been in development for the last two years and has a D1 chip that uses a 7nm manufacturing process along with 362 teraflops of processing power. 25 such chips are placed on a single training tile, and 120 such tiles come together to make a server cabinet, amounting to over an exaflop of power, said Venkataramanan, senior Director of Autopilot hardware.
These D1 chips will help train automakers’ vehicles from the videos obtained from the inside cameras. The model requires extensive computation but as Elon Musk says, “Dojo operational will be out next year.”
The work would come to the forefront after producing the AI chips for the last two years. These chips are capable of making reflex decisions about what happens on the road.
“This was entirely designed by Tesla team internally. All the way from the architecture to the package. This chip is like GPU-level compute with a CPU level flexibility and twice the network chip level IO bandwight,” said Venkataramanan.
Currently, Tesla offers ‘Full self Driving Capability’ for $10,000. The vehicle is capable of changing lanes, moving in and out of the parking spot, etc. Later, this year the new advantage of automatic steering on city streets will also be available, which was to come in 2019.
The company’s Dojo Supercomputer will train the neural networks making self-driving, even more, easier and it will also help develop future AI-required technologies.
“Dojo” was Tesla’s first shot in developing supercomputers and was impressive as well, so it is expected that the future will see more bewildering improvements in Technology.