Tesla's new Autopilot hardware runs current features at just 5% of compute capacity
The latest semi-autonomous features have pushed Autopilot HW2.5 to around 80% capacity.
Tesla CEO Elon Musk has highlighted the significant jump in Autopilot processing capacity from the current hardware to the HW3 tech developed in-house.
The company recently began rolling out its 'seamless' Navigate on Autopilot feature, allowing vehicles to change lanes and maneuver highway interchanges without first receiving driver confirmation.
Musk says the HW3 computer "is at about 5 percent compute load for these tasks or 10 percent with full fail-over redundancy," compared to the existing Nvidia-powered HW2.5 system that has been pushed to around 80 percent of its processing capacity.
"Two, independent system-on-chip architecture, with each SoC having two NN accelerators that can perform simultaneous health-check calculations to protect against a soft error," the executive added, referring to the Autopilot HW3 chip architecture.
As Tesla begins to roll out even more Full Self Driving features that require HW3, existing cars will be retrofitted with the latest hardware to support the more advanced features. To be clear, however, "Full Self Driving" will initially be a bit of a misnomer as drivers will still be required to keep their eyes on the road and hands on the wheel.