The next generation of hands-free driving for General Motors will cover 2 million North American roads, a ten-fold increase over its current Super Cruise system’s 200,000-mile capability. For this new “Ultra Cruise” Level 3 hands-free driving system, the automaker will adopt the Snapdragon Ride platform. Super Cruise works only on scanned and mapped divided-lane highways.
Ultra Cruise will debut on the 2024 Cadillac Celestiq, a premium electric sedan, and should come closer to providing door-to-door autonomy. To match the Tesla Model S or Mercedes-Benz EQS, it would need a range of 300-plus miles; the price could be in the $100,000 range. Cadillac already has some Escalade SUVs with base prices over $100,000.
Going with Qualcomm is a big change in suppliers. Much of the auto industry has used vision company Mobileye as the leading provider of the chips and software that underpin vision-based advanced driver assist systems (ADAS).
While the Intel-owned company continues to grow its business, Mobileye’s competition in the ADAS and automated driving system (ADS) market is heating up rapidly with both Nvidia and now Qualcomm taking away chunks of market share.
Qualcomm is the relative newcomer to this sector with its Snapdragon Ride platform, but General Motors is taking advantage of that as it prepares to launch its next-generation Ultra Cruise hands-free driving system next year.
Qualcomm also announced at CES 2022 that both BMW and Renault will adopt the Snapdragon Ride platform by 2025.
If It Doesn’t Build a Car, Apple Could Sell Chips to Automakers
Another potentially interesting player in the automotive chip sector is Apple. The maker of Macs and iPhones has had an automotive development program going that includes automated driving for at least seven years.
Whether or not Apple actually ever produces a branded vehicle remains to be seen, but the company has proved its mettle in chip design with its mobile devices and more recently Macs. Apple Silicon offers everything that automakers are looking for, extremely high performance and low energy consumption. Even if Apple never builds a car, they could have a very healthy side business as a chip vendor to other automakers.
Soon, Point-to-Point Driving
At GM’s investor day in October 2021, GM revealed the first public details about Ultra Cruise, a system that had been hinted at since at least 2019. Unlike the current Super Cruise that is limited to driving on divided highways, Ultra Cruise will enable point-to-point hands-free driving on all types of paved roads. Eventually, it will work on all roads.
Providing the ability to drive in urban and suburban areas will require more sensors, much more sophisticated software and a lot more computing power than what GM is using for Super Cruise today. That’s where Qualcomm comes in.
Automakers Already Use Qualcomm for Infotainment
The San Diego-based company’s Snapdragon systems-on-chip (SoC) power many of the smartphones and tablets not named by Apple. In recent years several automakers have started using derivatives of those SoCs for infotainment systems and at CES 2020, Qualcomm announced the Snapdragon Ride for ADAS and ADS.
The first iteration of Snapdragon Ride consists of two chips, the ADAS processor and an artificial intelligence (AI) accelerator. They can be used in various combinations to provide whatever level of performance is needed. GM was the first automaker to adopt the Snapdragon Ride platform and will actually be using a couple of different combinations. The Cadillac Lyriq EV launching this spring will use one ADAS processor to run its Super Cruise system. Ultra Cruise will use two ADAS processors and one AI accelerator.
Nvidia has been developing large single chip solutions like the Orin that is launching this year and Atlan that is due in 2024. Nakul Duggal, SVP and GM of Automotive at Qualcomm, explained that “when you have these massive SOCs, remember that there is a yield aspect that you’re going to take care of, there is a thermal aspect that you’re going to take care of.”
A multi-chip solution allows manufacturers to scale different combinations depending on what functionality is needed, while large chips may have lower production yields and thus higher system costs.
Current Self-Driving Computers Can Draw 1,000 Watts
Jason Ditman, chief engineer for Ultra Cruise explained that by getting in early as Qualcomm was still defining the details of the chips, it was able to get the right balance latency, energy consumption and packaging space. Energy consumption is a real challenge for automated vehicles with the computers often consuming well over 1,000 watts, or 1 kilowatt-hour for every 60 minutes of driving. In an EV where every watt counts toward range, minimizing the draw from computers, lights, climate control and other systems is crucial.
Fewer, Bigger Chips Consolidate More Functions
With 360 trillion operations per second, the three-chip Ultra Cruise configuration provides enough computing performance that GM was able to consolidate some of the functions previously run in separate electronic control units like radar adaptive cruise control and lane keeping assist into the Ultra Cruise computer. This reduces cost and complexity by eliminating a lot of extra wiring and the need to find places to mount those other computers.
The whole system fits into a box that is roughly the size of three thin laptops stacked up. Its relatively modest power consumption of 65-70 watts means that the box can get by with forced air cooling. The full-self-driving computer in Tesla vehicles requires liquid cooling.
Qualcomm Hardware, Software from GM
While Qualcomm is also developing its own software stack that can provide everything from basic driver assist functions to a fully automated driving system, GM has opted to just use the hardware. The entire software stack for Ultra including sensor processing, prediction, path planning and control has been developed in-house at GM.
“The sun never sets on the Ultra Cruise software team,” Ditman says. They’re widely dispersed, if not quite literally in daylight sun. Ditman says there are research and tech centers in Limerick, Ireland; in Canada; in Austin, Texas; and at GM’s main research post in Warren, Michigan.
GM and Qualcomm have worked together more closely than usual to bring both Snapdragon Ride to market. Watch for more such collaborations as automakers move to bring more software development in-house and integrate it on higher performance computing platforms.