FSD V11 Release and Upgrades

Tesla's Full Self Driving program is hitting a lot of big milestones lately, and with the news of FSD Beta Version 11's release notes being shared with the public, we're also hearing about the company's plans for the software's rollout - and a look at some next generation hardware.

Tesla started the rollout of their FSD Beta V11 on November 11th, with the news that it would be the first iteration of their autonomous driving software that would run on a single stack - meaning Tesla's highway-focused Autopilot and its main FSD software would be combined and operating as one program.

The EV company's autonomous driving programs have been a huge draw to the brand, and allow for Tesla vehicles to do things like change lanes, give driver alerts, even navigate and park the car on its own to some extent.

And while these systems aren't fully autonomous - requiring driver supervision - Tesla's beta testers have been eagerly awaiting each update - especially Version 11, which the company believes should run a lot smoother as a single stack.

For obvious reasons, the biggest part of the V11 release notes has to do with the merging of the two programs - officially activating "FSD Beta on highway". As the entry says, the legacy highway system still relies on several single-camera networks. But the FSD Beta uses a multi-camera video network, and their next-gen planner which the company says will allow for more complex roadway interactions that don't focus on the lanes as much as the old system did - which should make it more flexible.

To go with that, there are several entries related to highway behaviour. Things like improved merging, road debris detection,  improved decisions on lane-changes based on speed, and reduced sensitivity for those lane-change decisions if your car is in "CHILL mode" - which is a customization setting that lets you set how quickly your Tesla accelerates and slows down.

There was definitely a lot of time spent on the lane-changing behaviour. Aside from the earlier two points dealing with speed, there are two more that deal with when your Tesla may decide to make a quick "jerk" manoeuvre, and the steps it takes to minimise the need to make them. 

This is apparently where the extra flexibility comes in, as the system will be reportedly less strict on where it is in the lane, as it anticipates lane changes with a newly improved trajectory optimization, which lowers the latency of the calculations made by 20% using "Numerical tricks for more efficient computations."

We're not exactly sure exactly what tricks they're leveraging there, but this whole update seems to be something Tesla believes is stable enough to commit to a wider release.

In a couple of Twitter conversations, CEO Elon Musk said that the company was looking to widen the Version 11 beta release before Thanksgiving.

What they meant by "widen" was a little unclear, but Twitter users came to the rescue again on the 20th, asking Elon if he meant V11 was going to be going out to people who currently have the beta, or to everyone who purchased FSD, and won't have to do the safety score - which is a system Tesla uses to ensure drivers they invite to the FSD Beta program are safe enough, based on five "Safety Factors" that involve things like hard braking and aggressive turning.

Elon responded with "Latter", making it pretty clear that they want to get Version 11 to every FSD user by the end of November.

That's a pretty clear statement about how confident Tesla is in their new version.

But the shiny new software improvements aren't the only ones coming to the company's FSD system.

We recently got reports that Tesla has placed a gigantic order for new computer chips for their next-generation FSD computer from Taiwan Semiconductor Manufacturing Company, or TSMC - one of the world's largest semiconductor companies.

It's likely this is the chip Musk mentioned at the launch of their previous Samsung-produced chip at the Tesla Autonomy Day back in 2019.

That chip had been designed to completely replace Tesla's older hardware, which had been powered by Nvidia equipment.

But while Elon's announcement on their next-gen chip detailed that it would be coming out in roughly two years after that event, we hadn't heard much about it since TSMC reported receiving an order so large that Tesla would be in TSMC's top seven customers next year - when the order would presumably start to be filled.

We don't know much more about it, other than it will be supplied to Tesla's factories from TSMC's facility in Arizona, and that Tesla confirmed the new chip wouldn't be required to achieve Full Self Driving - but it would definitely increase performance.

Tesla's previous chip from 2019 was a part of their Hardware 3.0 self-driving computer, and improved their processing speeds by a factor of 21. Pete Bannon - then head of Autopilot Software - said he had never been part of a project where the factor of improvement was more than 3.

And Elon said that this new chip is supposedly 3 times better than that one - so it looks like we're about to see a line start to form for retrofits.

Previous
Previous

Tesla's March to Lithium

Next
Next

Starship Static Fires 14 Raptor Engines