The Path Toward Future Automotive EE Architectures

From a semiconductor market perspective, all eyes are on the automotive domain. According to Gartner, as of 2023, the automotive market is now its second-largest segment, with about 14% of the demand. Only smartphones consume more. As I mused last month in “Automotive Semiconductor March Madness 2024,” those who made a bet on automotive a decade or longer ago are pretty happy these days. Still, many new entrants are hurrying toward automotive, and especially EE architectures are worth a look to understand the critical trends. Whether we are facing evolution or revolution is sometimes hotly debated. And there are some twists in the story here, as some OEMs simply decided to skip ahead and fast forward to a fully integrated vehicle computing solution. Say what? Read on!

Perspectives on automotive EE architectures

Inspired by a colleague and friend who collected every new incarnation of the automotive V-Diagram he encountered (you know who you are), I started collecting diagrams outlining EE architecture evolutions a while back. There are plenty, and I combined a representative subset in the collage below, illustrating some of the different public perspectives from various players in the automotive design chain.

Tier 1, semiconductor, analysts and software perspectives on EE architectures. (Sources: NXP, Renesas, Infineon, Continental, TTTech Auto, and GSA/McKinsey)

There are two commonalities in all perspectives. Clearly, the flat, fully distributed architectures are a thing of the past. And all perspectives eventually arrive at some form of fully centralized computing. Most perspectives partition the car into “Domain Architectures” and “Zonal Architectures” before fully consolidated computing.

When and how remains to be seen, but NXP, for instance, identifies model year 2030 as the target for “Zonal EE.” Critical considerations in assessing the path to get there include, on the one hand, the aspects of the physical transition – how to “zonalize” the networks – and, on the other hand, the elements of the logical transitions – the clustering of the various domains. All of them come with unique choices and challenges for design teams around what features to integrate and how to potentially isolate them, how to communicate between functions, and how to connect sensors. On top of that, considerations around legacy software and – in a connected world – its distribution play a role.

The twist – the hare and the tortoise?

While most automotive players come with individual legacies, a fresh start may create a “Hare and Tortoise” situation. Living in Silicon Valley has its advantages, as from time to time, you witness an incarnation of William Gibson’s quote, “The future is already here, it’s just not evenly distributed yet.”

Last October, I got an invite to the GAAC/ESDA event “Semiconductor Meets Automotive,” featuring a presentation “Software and Hardware System Architecture for Next Generation Vehicles” by Shiv Sikand, Executive Vice President, Drako Motors. My unassuming 1997 Miata drove yours truly up to Milpitas and parked only to find these cars in front, recreating a classic and immediate “Mia/Tia-Lightning McQueen” love story.

The Drakos – Dragon and GTE. My Miata is hiding in the background because she is shy. 

It took me not long to connect the dots here to Dean Drako of IC Manage.

Why did I call this section “The Hare and the Tortoise?”

Well, Shiv introduced DRAKO Drive OS and showed how the two cars on display outside the auditorium already stand at the finish line toward centralized computing. His talk was enlightening as to what is possible starting from a blank sheet. A version of this talk given by Drako Motors’ Chief Software Architect, Richard West, is linked below.

DRAKO Drive OS transforms traditional hardware functions into software running on multi-core x86 processors, providing simultaneous hard real-time and general-purpose computing. Its origin goes back to 2005 with many improvements over the years. For car networking, they developed “one cable to rule them all,” ahem, “… do it all” for data, power delivery, audio & display.

It’s not I2C, CAN, LIN, or FlexRay, as they are bandwidth-limited. It’s not Ethernet, either. The team ruled it out due to its real-time challenges, jitter, absence of hardware clock synchronization, and time-triggered Ethernet not yet commonplace.

It’s USB.

USB?

They developed a custom USB cable offering high bandwidth, low latency, real time, low cost, and low mass. The resulting car architecture is illustrated below.

Boom. Right to central compute. No intermediate steps. Drako Motors also licenses the OS to partners and has a reference platform.

Source: Drako Motors, “A Software and Hardware Architecture for Next Generation Automotive Systems”, https://bit.ly/3u7L0rM

EE architecture evolution and data transport architectures

Not every development team will jump as far and fast as Drako Motors did.

For everyone else, there is a path with many intermediate steps (and silicon) toward fully centralized computing. This evolution has a lot of impact on data transport architectures, as developers need to balance on-chip and off-chip networking, consider safety and security, and storage and caching for shared data is critical everywhere. For that purpose, the world we at Arteris play in – safe coherent and non-coherent networks-on-chips (NoCs), chiplets, last-level caches, and system-on-chip (SoC) integration automation are critical components of the development ecosystem.

Now, all that is left for me is to figure out how to get my hands on a Drako Motor Dragon. The lease for my current Kia Niro is almost up. I will also look for Dean Drako at DAC in June to get a GTE test drive. Not that I would be qualified for a race car, but one can hope, and perhaps I can get a passenger seat.