Cars are becoming increasingly “connected devices.” Self-driving functions, driving range optimization, and maintenance sensors are just the beginning. Faster ARM processors aim to provide cars with performance previously achievable only in data centre servers. On top of that, the development of new chips will be accelerated thanks to new tools provided by ARM. What will drivers gain from this immense computing power?
ARM emphasizes that cars are now a “computer on wheels” and are now usually the most complex piece of technology someone owns. An entire ecosystem of partners contributes to the continuous development of SDVs, or Software-Defined Vehicles. ARM shares its chip blueprints with NXP, Nvidia, MediaTek and Marvell Technology, among others.
For the first time, ARM is bringing its Neoverse CPUs to the automotive industry with the added moniker of Automotive Enhanced (EA). Previously, Neoverse has already been deployed for a variety of applications, from cloud services to 5G infrastructure and edge computing. These chips scale up to quite large numbers of cores, threads and instructions per second. The bottom line is that Neoverse provides cars with a significantly more powerful computing engine.
Smarter driving, less consumption
Car chips already provide many benefits to users. Integrated control systems function as the beating heart of vehicles today, especially when it comes to electric cars (EVs). However, the most useful application of the extra ARM computing power is quite straightforward. EVs can drive farther than ever without requiring a charging station, but any range extension offered by better hardware is desirable.
AI makes an appearance here, as it almost always does these days. An EV can achieve optimal consumption using artificial intelligence by harnessing as many sensors as possible, ranging from wind conditions to tyre pressure and suspension. Extra computing power simply produces better results faster.
Add to that the fact that machine learning can predict when maintenance is needed. Again, stronger hardware can predict this more accurately and efficiently. An increase in price when buying a car is, therefore, ideally offset by less drastic maintenance costs, with predictive maintenance of issues that are cropping up before their symptoms are noticeable. Smarter cars certainly have their benefits.
Connectivity and better AI driving
ARM emphasizes that cars are increasingly connected to the rest of the world. With that also comes a real threat. More in-car options can be controlled via software than ever, which poses security risks. A threat actor could infiltrate a car’s digital systems and tamper with it, either through malware or simply through remote access. Additionally, the data collected can be significant, even to the point that certain cars are “privacy nightmares“.
The new ARM AE chips are more locked-down than ever to prevent such security issues. Existing features from server and edge chips will now stop memory errors and buffer overflow attacks. At the same time, this underscores that current connected cars lack these safeguards.
Better hardware can also lead to better driving. Advanced driver assistance systems (ADAS) range from more advanced adaptive cruise control to full-blown self-driving functions, with the latter still greatly benefiting from improved computing power. For now, though, this seems to be a problem of both software and hardware, with additional sensors contributing to safer AI driving behaviour. The extra CPU cores and threads within the new ARM AE chips make harnessing all this data much more feasible.
Faster chips
Traditionally, chip development in the automotive industry has used a far slower release cadence than elsewhere. Where state-of-the-art chip processes serve smartphones, desktops and servers, the older, mature nodes serve as the driving force for the automotive sector. This strategy prevents automakers from competing directly for the most advanced manufacturing technologies with parties such as Apple, which is only too happy to snatch virtually all capacity from chip giant TSMC.
That’s about to change. “Next-gen” cars will now run on brand-new production processes designed specifically for the automotive world. This will carry a price tag for both carmakers and consumers, but should bring significantly better features. In any case, it offers more efficient Systems-on-a-Chips (SoCs) than before, which in itself is a positive development.
Thanks to new design techniques, ARM allows partners to create custom designs before a physical processor exists. These are called “virtual prototypes,” made possible through services from AWS, Cadence, Corellium, Siemens and more. It means we can expect much faster development of automotive hardware. But is that something we should really want?
The downside: data at the manufacturer
Better hardware offers, but there are two sides to this story. On the one hand, it offers such objective benefits as higher efficiency and a pre-emptive insight into possible maintenance. On the other hand, the multifaceted nature of ADAS makes it a bit more debatable. How long will it take before we fully trust self-driving features? The answer to that question will vary greatly from person to person.
In addition, the considerable amount of data collection perhaps poses a bigger problem. Although ARM protects drivers from malicious actors, nothing stands in the way of car manufacturers themselves. This could have major consequences for users. In America, General Motors appears to have sent a massive amount of data to insurance companies, The New York Times revealed. The massive pile of data per user means that risk assessment bureau LexisNexis was able to create a full risk profile. It means drivers can be branded by insurers as a potential high-risk case without their knowledge.
Crucially, users have to be aware of what will be done with their data. After all, those who do not use advanced features but do connect their car to the internet are involuntarily exposed to the negative effects of data collection. Newer chips would then effectively function primarily as data aggregators on behalf of the manufacturer and third parties. As in the IoT world, this needs to be standardized to make data collection transparent and opt-in for consumers and business customers alike.
This is on top of the previously mentioned privacy issues surrounding connected cars that were already known. With all the data collected, manufacturers possess an instrument they can turn against users. Faster chips exacerbate the potential consequences, which were already worrisome. If cars henceforth benefit directly from cutting-edge technology, this becomes an even more pernicious issue. What if AI systems, which ARM also likes to talk about, are not as safe as they seem? In that case, we are not just talking about self-driving functions, but also integrations with generative AI. LLMs have already found their way into cars, such as ChatGPT in Mercedes-Benz or a TomTom assistant. Whether this ever leads to danger is an unanswered question, but it makes the current advance of car chips far from carefree.
20 hours ago
