Until today, Tesla was using NVIDIA technology inside countless self-driving Model S and Model X vehicles but today the company announced its own in-house designed processors have been installed into thousands of Model S and Model X vehicles since March.
Tesla will be putting its new Full Self Driving Chip in the new Model 3 vehicles, something it has been doing for just over a week now. Elon Musk announced the new chip during Tesla's Autonomy Investor Day in Palo Alto, California, where he referred to it as "the best chip in the world". The Full Self-Driving Chip measuring in at 260 square millimeters packing 6 billion transistors, and has 21x the performance of what NVIDIA's chip was providing.
The new chip is designed by Tesla and manufactured on Samsung silicon on the 14nm FinFET node, with the self-driving chip purpose-built to handle everything for self-driving cars including car sensors that can now detect things much further away than current AI processors. Tesla made a point that the new chip isn't over-engineered and that it doesn't need crazy high-end CPU and GPU specs for its specific tasks, so it is built for a sole purpose and excels heavily because of it.
The company will be pumping out the new Full Self Driving Computer boards with two of the chips, which is used for redundancy. You don't want a full self-driving car relying on a single processor and then that processor fails, so the inclusion of a second processor is a pretty crazy thing for Tesla to offer.
Musk explained: "Any part of this could fail, and the car will keep driving. The probability of this computer failing is substantially lower than someone losing consciousness - at least an order of magnitude".
From here, Tesla says that every car it makes will have the hardware it needs for full self-driving, where from now on out Musk says: "All you need to do is improve the software". Full self-driving cars with in-house Tesla-designed chips? Alrighty then.