In conversation with Jamie Broome, Codasip VP of Automotive
Jamie Broome recently joined Codasip as VP Automotive with more than 20 years of semiconductor and automotive experience, after being responsible for Imagination Technologies’ automotive strategy. In this interview, he shares the exciting opportunities ahead for RISC-V, the changing landscape of compute, and how the automotive industry is adapting to the explosion of user technology and the oncoming challenges from connected and autonomous vehicles (CAVs).
How do you see the future of automotive?
The semiconductor industry is constantly changing. It typically adopts an industry as it goes through a transformation, be that consumer electronics in the home, mobile devices or now the automotive segment. Global chip shortages have, at the same time, highlighted how dependent we are on silicon to keep cars on the roads. These shortages are also keeping wait times for new vehicles at an all-time high. But the most potentially disruptive change is an influx of non-traditional, tech players, big and small, seeing opportunities in the market. It’s easy to see why the automotive sector is arguably the hottest in the tech world right now.
A new marketplace in automotive innovation and technology is taking shape, driven by user demand and expectation, this is creating a battleground between those existing pillars, tech giants and new business models. It is widely recognized and promoted by the executive leaders of the automotive OEMs, in conferences all over the world, that 80% of the cars value will be in the electronics, and the ability to differentiate in this field is the key to success. The solutions and ownership of them is subtly different to the other markets that the semiconductor industry adopted in the past.
What do you think is the potential of RISC-V in this evolving automotive industry?
The potential of RISC-V in automotive is real and is transitioning well beyond simply the early adopters. Two things are fundamentally changing: the concept of software-defined vehicles and the need for democratizing innovation at the design level. This transition from hardware being separated to software to design synergy and coherence between them unlocks new services and solutions for car manufacturers, enabling the delivery of updates and upgrades over-the-air (OTA) for a better driver experience. And here, compute becomes a focus for the next generations of vehicles. RISC-V processor IP, expanding and readily customizable, combined with Codasip Studio technology, delivers access to rapid innovation, ownership and cost reduction right into the hands of automotive players. These are companies of all sizes that are experiencing the pains today and only a few have the desire or the pockets deep enough to build an entire vertical semiconductor IP capability.
Where do you see the biggest opportunities?
Throughout the functionality of the car there are many areas of opportunity, old and new areas of electronics that will merge or innovate. However, one area where there is a clear overlap is dedicated acceleration or coprocessors. Whether that’s a dedicated compute mix or specific hardware tailored for its own task, complementing the main processor, there will be a need and a demand for domain-specific accelerators, and opportunity for processors to manage and optimize that acceleration. At the smaller scale, applications such as radar or LiDAR will benefit from this and are both key to tomorrow’s automotive context. Certainly at the larger scale in the architectures we see in AD and L3+ ADAS features.
AD and L3+ ADAS is ML/AI centric and this requires aligned dedicated resources as well as complimentary non-ML processing to make the ML/AI efficient.
Vector instructions are a good example of dedicated ML/AI resources. This is a relatively new addition to RISC-V, but something that has been used in the industry for a while. Supporting this adoption of fast architecture exploration to determine the right set of Vector instructions required for a specific task is essential. That is something Codasip Studio, and its Instruction Accurate modeling, already supports and makes simpler and faster, to understand the gains and required balance of performance needed. We could imagine a sensor or automotive SoC perfectly functioning for the needs of Level 2 autonomous vehicles that now need to adapt the compute mix in a distributed way to achieve the demands of Level 4.
Can you tell us more about the need for ownership in automotive?
Cost is a known coefficient in the automotive supply chain, so first and foremost the correct ownership reduces cost. But driven by the change of value in the vehicle to the electronics and the alignment and pace of the wave of software and system development in automotive, ownership of the innovation is needed to deliver the speed to match the software development as well as the protection of design for differentiation and the advantage one player has over another. BMW did not buy off-the-shelf internal combustion engines, they created the ultimate driving machine centered around their innovation in the complex world of mechanics in the engine. That complexity simply is not available in electric drive trains, but is very much ripe for the taking in electronics.