The supply chain it needs simply doesn't exist, but that could work to GM's advantage:
Cruise never planned to make its own silicon. But in the quest to commercialize robotaxis — and make money doing it — those never planned pursuits can suddenly seem a lot more appealing.
Cruise realized that the price of chips from suppliers was too high, the parts were too big and the reliability of the third-party technology just wasn't there, Carl Jenkins, Cruise's vice president of hardware, told TechCrunch during a tour of the company's hardware lab last month.
Amid a hiring spree that began in 2019 and continued into 2020, Cruise doubled down on its own hardware, including its own board and sensors. The investment has helped the company develop smaller, lower cost hardware for its vehicles. It has also resulted in its first production board the C5, which is powering the current generation of autonomous Chevy Bolts.
When the company's purpose-built Origin robotaxi starts hitting the streets in 2023, it will be outfitted with the C6 board. That board will eventually be replaced with the C7 which will have Cruise's Dune chip. Dune will process all of the sensor data for the system, according to Cruise.
[...] Automakers (not counting Tesla) have taken a more cautious approach to autonomous vehicles that would be sold to consumers. The technology built and proven out by Cruise could eventually make its way into a GM product sold to a customer.
"When we start researching and looking at personal autonomous vehicles there are choices, like does the car have pedals or does it have pedals that are deployable or does it not have pedals at all," Reuss said. "And so we're looking at what people want and those aren't easy questions to answer."
[...] At the end of the tour, Cruise set us up with an autonomous ride in a Bolt.
[...] It was exciting initially, and then boring, which is exactly what driverless ride-hailing should focus on. Yes, it's slightly weird to be in a car driven by a robot, but after 20 minutes of being carted around by a careful robot, the last 10 minutes are spent wondering if you'll get stuck at an intersection just to add some excitement to the ride.
Related: GM's Cruise is Making its Own Chips for Self-Driving Vehicles to Save on Costs
Related Stories
GM's Cruise is making its own chips for self-driving vehicles to save on costs:
GM's Cruise division doesn't want to rely on third-party manufacturers for the chips powering its autonomous vehicles — so, it's making its own. Based on what Carl Jenkins, the company's VP for Hardware Engineering, told Reuters, the main motivator for the switch is the lofty costs associated with paying for other companies' chips.
"Two years ago, we were paying a lot of money for a GPU from a famous vendor," Jenkins told the news organization, referring to NVIDIA. He explained that Cruise couldn't negotiate because it wasn't mass manufacturing autonomous vehicles just yet. [...]
Jenkins has revealed that Cruise had already developed four chips at this point, starting with Horta, which was designed to become the main brains of the vehicle. Dune will process data from sensors, while another chip will process information from the radar. Yet another one will be announced at a later date. These components will power the Cruise Origin, the self-driving electric shuttle the company first announced back in 2020. The Cruise Origin will have no steering wheel or pedals and will instead have four seats inside facing each other. It's intended to be used as a shareable vehicle that's on the road at all times, shuttling passengers to their destinations.
Company executives didn't say how much they spent on the chips' development, but they believe they could recoup their investment once Cruise starts scaling up production. [...] GM chief Mary Barra announced at CES this year that the automaker wants to sell personal autonomous vehicles by the middle of the decade.
(Score: 2, Insightful) by Anonymous Coward on Wednesday October 19 2022, @07:34PM (3 children)
If not they are just part of the problem.
Maybe the auto makers need to rethink throwing dozens of processors into a vehicle in the first place
or that the world needs a well-built SIMPLE vehicle.
(Score: 0) by Anonymous Coward on Wednesday October 19 2022, @08:03PM
We threw it all away [wikimedia.org]
(Score: 3, Informative) by RS3 on Wednesday October 19 2022, @08:43PM (1 child)
There are many factors. Many of the systems in cars, needed or market-driven, have become very complex and one factor is that the design/development teams need to be fairly specialized. The entertainment designers / engineers are very different from the airbag controller people.
Another factor is that even if you made a car with just one main computer, you have a single point of major failure. I suppose you could keep a spare computer but that would be pretty expensive for most people to add to the price of a car.
Another factor is wiring. You'd need probably thousands of wires branching out from just one, or only a few computers.
Small CPUs are super cheap, pennies, much cheaper than all that wiring and the labor to install it all (not to mention the nightmare of repairs!)
The subsystems all need to communicate on a vehicle-wide bus, usually CAN bus [wikipedia.org], so you need intelligent (CPU) controllers in each subsystem.
I agree that it's gone a bit too far with a CPU in the stupidest small thing like a toaster, but again, they're so cheap, and give designers abilities to do things they could never do before (within reasonable cost range).
And of course there are libraries and example code and code re-usability. Plus it's much easier and cheaper to fix some code rather than fix millions of produced circuits that had mistakes, or just didn't work in the production environment of the normal variation in electronic component parameters.
Another big factor: totality of the situation. Nobody that I know of is managing the totality of all of the things using CPUs, and the totality of CPU production. So we are where we are.
I found it very interesting to learn that IBM used to make _all_ of their parts and components, mainly for these kinds of reasons. They didn't want to be at the mercy of another company, their whims, competitors hogging up available parts, etc. In fact, IIRC the PC was a radical departure using "off-the-shelf" components, standard things like TTL, ASCII, etc.
(Score: 2, Touché) by khallow on Thursday October 20 2022, @05:12AM
Beats several single points of major failure.
(Score: 1, Interesting) by Anonymous Coward on Wednesday October 19 2022, @10:14PM (1 child)
"Many of the systems in cars, needed or market-driven, have become very complex"
And no one has ever said "what of all this complexity is actually NECESSARY" ??
Maybe a well-made product that isn't filled with crap might have a market!
(Score: 4, Informative) by RS3 on Thursday October 20 2022, @03:56AM
Much of the need is driven by govt. safety and emissions requirements. Even things that are optional like ABS, 4WD / AWD, stability / traction control, it's more of a market need for safety, not just whimsical gadgets.
(Score: 2) by richtopia on Thursday October 20 2022, @02:42PM
As a reminder we are talking about Cruze, GM's self-driving division. Companies dealing with large volumes of data or very specialized sensors need to design their own chips. In the server space we see Google, Amazon, and Tesla with massive compute improvements for their specific workloads. Tesla has also been working on their own boards and even chips for in car applications.
Probably everyone on Soylent News is familiar with the tradeoff of moving to more specialized hardware (CPU/GPU/ASIC). When we are talking about self-driving vehicles the volume of data and required response time justifies dedicated hardware. With so many competitors in the space, this type of hardware will enable a lower-power and more manufacturable solution.