■ ADSA introduction, a key element in vehicle design

The introduction of Advanced Driver Assistance Systems (ADAS) has now become a key element in vehicle design to improve vehicle safety and usability.

Automakers are working to build increasingly autonomous vehicles and ultimately finding ways to achieve fully autonomous driving (AD).

As user expectations for ADAS and AD, as well as infotainment and personalization, increase, automobiles are evolving into mobile data centers.

Therefore, communication between core hardware elements such as integrated circuits (ICs), circuit boards, and modules required for software-defined vehicles (SDVs) is essential for successful autonomous driving operation.

In fact, some cars already have over 100 million lines of code, according to market research firm Straits Research.The size of the dragon software market is expected to reach approximately USD 58 billion by 2030, with a compound annual growth rate (CAGR) of 14.8%.

The complexity of autonomous driving software and the task of processing massive amounts of data in real time from various vision system sensors such as cameras, radar (Radio Detection and Ranging), LiDAR (Light Detection and Ranging), and ultrasonic waves pose significant challenges.

For example, Figure 1 shows that the traditional communications infrastructure and standards currently used in the automotive industry have reached their limits.

Ethernet and CAN (Controller Area Network) buses will still play a critical role in future automotive architectures, but they will need to be supplemented to meet the high-performance computing platform (HPC) requirements needed to integrate artificial intelligence (AI) and machine learning (ML) into ADAS and AD.


■ Automotive PCIe® technology providing backward compatibility

PCIe (Peripheral Component Interconnect Express) technology was introduced in 2003. It was developed to meet the needs of the computing industry.

PCIe is currently also used in the aerospace and automotive industries, particularly in safety-critical applications implemented in firmware compliant with a standard called DO-254.

PCIe is a point-to-point bidirectional bus in which two devices are directly connected to exchange data. It is a serial bus method, but it is a hybrid method that can implement greater bandwidth by using a single lane or parallel lanes (2, 4, 8, 16).

Additionally, PCIe performance improves with each generation. Figure 2 shows the technological evolution of PCIe.


PCIe has been in service since around generation 4.0 and is already being used in some automotive applications.

But now is a better time to adopt PCIe than ever before, with performance improvements such as data transfer rates of up to 64 GT/s for Gen 6.0 and a total bandwidth of 128 GB/s when using 16 lanes.

What's notable here is that PCIe provides backward compatibility with previous versions.

Based on the premise that automobiles are evolving into ‘data centers on wheels’, let’s look at why PCIe technology is being used in general data centers.

■ PCIe High-performance, low-power ideal bus

A data center consists of one or more servers, along with storage devices, network components, and peripherals including input/output (I/O) to support high-performance computing (HPC) in the cloud.

PCIe is included in today's high-performance processors and is an ideal bus for building low-latency, high-speed connections between servers and peripherals.

For example, Non-Volatile Memory Express (NVMe) is specifically designed to work with flash memory and the PCIe interface.

PCIe-based NVMe SSD (Solid State The read/write speeds of the drives are much faster than those of SSDs using the SATA (Serial Advanced Technology Attachment) interface.

In reality, not all storage systems, SSDs, or hard disk drives (HDDs), can deliver the level of performance that complex AI and ML applications require.

Low latency over PCIe has a direct impact on performance between applications running on servers.

This means that PCIe is included in components other than processors and NVMe SSDs.

PCIe is also included in several components that serve as gateways between the cloud and the systems that access it.

And let's not forget that cars will become self-driving data centers, acting as nodes between 'smart cities'.

NVMe's use in data centers is also gaining popularity in terms of power consumption.

For example, the U.S. Department of Energy estimates that a single massive data center with tens of thousands of devices consumes more than 100 MW of electricity, enough to power 80,000 homes.

NVMe SSDs consume less than a third of the power of similarly sized SATA SSDs.

Power consumption is also an important issue in the automotive industry. In particular, in electric vehicles (EVs), power consumption has a direct impact on driving distance.

Automotive engineers, especially electric vehicle developers, are increasingly focusing on size, weight, power (SWaP) issues.

This is not surprising, considering that future ADAS implementations may see power consumption rise up to 1 kW and may require liquid cooling systems for thermal management.

But there are implications for other industries.

The aerospace industry has been designing for decades to meet SWaP (size, weight, power) and cost (SwaP-C) requirements, and line replaceable units (LRUs) with liquid cooling systems, such as power supplies, have been in use on some military platforms for more than a decade.

■ Optimized ADAS/AD implementation requires PCIe

Data centers have long utilized PCIe hardware and have been actively leveraging PCIe to optimize systems for a variety of workloads.

Data centers are also adept at developing interconnect systems that use a variety of protocols.

For example, while PCle is used for tasks that require immediate responses, it can be used alongside less speed-critical communication methods, such as Ethernet, to support connectivity between geographically dispersed systems.

In automotive environments, such 'less speed-critical' communications include telemetry technologies such as data transfer between sensors or vehicle lighting control functions.

These features do not require PCIe, but the distance is only a few centimeters. PCIe is required for short-distance, high-capacity data communication between ICs that perform real-time processing.

Therefore, to implement optimized ADAS (Advanced Driver Assistance Systems) and AD (Autonomous Driving Systems), PCIe is required along with Ethernet, CAN, and SerDes (Serializer/Deserializer).

Unlike Ethernet, there is no separate PCIe standard for automotive applications, but PCIe has been used in automotive applications in recent years.

Likewise, although there is no specific standard for PCIe for the aerospace industry, large aerospace and defense companies that are constantly pursuing SWaP-C (size, weight, power, cost) optimization are actively utilizing the PCIe protocol in safety-critical applications.

PCIe is emerging as the preferred computer interconnect solution in the automotive industry as solutions must be optimized for interoperability and scalability.

PCIe provides ultra-low latency and low-power bandwidth scalability, enabling efficient data transfer between CPUs and specialized accelerator devices.

Although there is no automotive-specific PCIe standard yet, semiconductor manufacturers are advancing the technology to make PCIe more widely available in harsh automotive environments.


For example, Microchip launched the industry's first automotive-qualified Gen 4 PCIe switch in 2022.

These switches, called Switchtec™ PFX, PSX and PAX, provide the high-speed interconnect required for distributed, real-time safety-critical data processing in ADAS architectures.

In addition to these switches, Microchip offers a wide range of PCIe-based hardware, including flash-based FPGAs and SoCs, as well as NVMe controllers, NVRAM drives, retimers, redrivers and timing solutions.

Finally, another thing the automotive industry needs to consider is that data centers treat CapEx (capital expenditures, or capital expenditures) as a long-term investment, much like a future pension.

Until now, most automotive OEMs have viewed CapEx as a one-time revenue stream that occurs at the point of purchase.

This view is not problematic when it comes to hardware. Of course, most OEMs charge for software updates from time to time, but SDVs require a complete rethink of their business model.

It is no longer appropriate to simply estimate costs based on hardware alone.

■ Automotive solutions, effective application of existing HPC architecture solutions instead of initial development

To improve the level of autonomous driving, cars must become “data centers on wheels,” equipped with high-performance computing capabilities and capable of processing massive amounts of data generated by various sensors.

Fortunately, high-performance computing (HPC) has already been established and is a core technology for high-frequency trading (HFT) systems and cloud-based AI/ML applications.

Proven hardware architectures and communication protocols such as PCIe already exist.

This means that automakers can learn a lot from how HPC is implemented in data centers.

Several cloud service providers, such as AWS and Google, have similarly been developing and optimizing HPC platforms for years, and there is already hardware and software available for immediate use.

Rather than developing a new solution from scratch, automakers would be better served by adapting these existing HPC architectures.


※ Contribution
Daniel Leih, Product Marketing Manager, USB and Networking Business Unit, Microchip Technology