During earnings call day, Elon Musk has announced the company's intentions of building its own AI chipset for its self-driving cars. The product in question is the 'Tesla Computer' also known as 'Hardware 3' which is a Tesla built hardware product. This will be used on the Model S, X and 3 for the requirements that are needed for self-driving capabilities. Tesla thus far relied on the on Nvidia’s Drive platform and with this swap, Tesla says it’s able to focus on its own needs for the sake of efficiency. Pete Bannon, director of the Hardware 3 project said: "We had the benefit of knowing what our neural networks look like, and what they’ll look like in the future." Elon Musk says that the key benefit of this is that it will bring the ability to run the neural network at a fundamental level. Unlike how GPU and CPU would operate, the calculations required for self-driving cars can be done in the circuit level, instead of emulation mode. He even went to post the comparison number; the Tesla’s computer vision software running on Nvidia’s hardware was handling about 200 frames per second, on the other hand, the specialized chip is able to crunch out 2,000 frames per second. More importantly, ...
Monday, 6 August 2018
Tesla planning on building own AI chipset for its self-driving cars
During earnings call day, Elon Musk has announced the company's intentions of building its own AI chipset for its self-driving cars. The product in question is the 'Tesla Computer' also known as 'Hardware 3' which is a Tesla built hardware product. This will be used on the Model S, X and 3 for the requirements that are needed for self-driving capabilities. Tesla thus far relied on the on Nvidia’s Drive platform and with this swap, Tesla says it’s able to focus on its own needs for the sake of efficiency. Pete Bannon, director of the Hardware 3 project said: "We had the benefit of knowing what our neural networks look like, and what they’ll look like in the future." Elon Musk says that the key benefit of this is that it will bring the ability to run the neural network at a fundamental level. Unlike how GPU and CPU would operate, the calculations required for self-driving cars can be done in the circuit level, instead of emulation mode. He even went to post the comparison number; the Tesla’s computer vision software running on Nvidia’s hardware was handling about 200 frames per second, on the other hand, the specialized chip is able to crunch out 2,000 frames per second. More importantly, ...
