Xilinx has made steady progress in the automotive market with penetration at 29 vehicle makes and 111 models in 2018. To-date, the company has shipped over 150 million FPGAs and SoCs into various automotive sockets, and over 50 million units specifically into ADAS applications. Xilinx believes its “adaptive silicon” will play a key role in meeting the evolving needs of the automotive sector, in particular for ADAS and autonomous driving. We look at some of the arguments that Xilinx presented during an exclusive briefing provided to Strategy Analytics.
- The move towards autonomous driving is changing the solution set requirements from an ADAS perspective. Adaptability and scalability is needed to address the different stages of autonomous driving that are being offered on a concurrent basis.
- LiDAR is a good example where the adaptability of FPGA solutions helps to meet the requirements across the different approaches being explored, ranging from mechanical scanning, flash LiDAR, optically phased arrays and MEMs.
- Infineon’s HSSL extension bus is a Xilinx FPGA optimised implementation used to exchange high speed serial data between various devices.
- Changing regulatory requirements are also a driver. The Euro NCAP roadmap across 2020-2025 is a good example of the changing requirements that will face automobile manufacturers as ADAS safety features, e.g. AEB (automated emergency braking) evolve to cover a range of scenarios (pedestrian, cyclist, junctions etc.), while additional capabilities such as driver monitoring and child presence detection also become a prerequisite.
- AI-enabled platforms will also evolve to cover a broader range of skill sets that will expand over time to include capabilities such as speech recognition, recommendation engines and anomaly detection atop the “basic” requirements to be able to classify, detect and segment.
- To achieve AI capabilities, the processing capabilities need to be able to combine both high throughput and low latency and Xilinx sees its FPGA solutions offering these characteristics with latency down to 3ms.
- Daimler AG leveraged the low latency and power efficiency of Xilinx FPGAs for an AI-based application.
- Xilinx is also powering ZF’s AI-based automotive control unit.
- Being able to reuse existing processing capabilities across different applications and end-use scenarios is an example of real-time “silicon re-use” allowing what Xilinx terms as DFX (Dynamic Function eXchange) where one SoC can rotate through different functions as needed.
- Xilinx also sees the flexibility of FPGA solutions enabling providers to offer differentiation while maintaining IP ownership.
Presently, the company is leveraging its XA Zynq® Ultrascale+™ MPSoC roadmap to serve the automotive industry with the ZU2 and ZU5 products already automotive qualified and the ZU7 and ZU11 products in the process of becoming qualified for the automotive sector.
Looking beyond Zynq, on June 18, 2019, Xilinx announced that it has shipped Versal™ AI Core series and Versal Prime series devices to multiple tier one customers through the company’s early access program. Xilinx touts the Versal as the industry’s first adaptive compute acceleration platform (ACAP), designed to be a highly integrated, multicore, heterogeneous compute platform that can be changed at both the hardware and software levels to dynamically adapt to the needs of a wide range of applications and workloads across the data center, automotive, 5G wireless, wired and defense markets.
Developed on 7nm TSMC process technology, the Versal architecture aims to combine a range of features to enable fast compute, low latency and greater flexibility than CPU and GPU implementations.
Key features of the ACAP architecture include:
- Multiple devices, each with dual-core Arm® Cortex®-A72 application processors, dual-core Arm Cortex-R5F real-time processors to underpin scalar engines for embedded compute applications.
- Adaptable engines for FPGA silicon programmability featuring over 2 million logic cells of adaptable hardware, and over 3,000 DSP engines optimized for high-precision floating point and low latency.
- Up to 400 AI Engines optimized for AI inference and advanced signal processing workloads.
- A flexible, multi-terabit per-second network-on-chip (NoC) designed to integrate all engines and key interfaces and enable software programmability.
Xilinx sees Versal as the next evolutionary step towards providing scalable, programmable and secure solutions for the growing requirements across ADAS and autonomous driving, with the FPGA underpinnings enabling Xilinx to provide solutions that can adapt to the changing automotive goalposts quickly and not be hindered by chip development costs associated with developing a standard product.
No announcement has been made yet on when the Versal architecture will be qualified for automotive. However, with its 16nm automotive-qualifed Zynq® UltraScale+™ MPSoC in mass production and its foundry partner TSMC working on Grade 1 automotive qualification of its 7nm process this year, Xilinx does not see this as a challenge.
Adaptability and scalability is the underlying tenet for these offerings and the Xilinx roadmap emphasises increasing functionality with each generation of device with the eventual aim of combining signal processing, AI, DFX and RF capabilities.
Xilinx believes that it has an advantage over “traditional silicon” providers such as NXP, TI and Renesas. While traditional automotive processor suppliers need to weigh the cost benefits and ROI risk of developing solutions for highly automated driving that may become obsolete in a much shorter timeframe than they are accustomed to, the FPGA-based approach allows Xilinx to develop solutions that can adapt to these changing goalposts as market requirements evolve.
- Which approach will play out as the automotive industry evolves on the road to autonomy, electrification and mobility as a service?
- Is there room for both approaches?
- Are automotive semiconductor systems moving towards a faster lifecycle?
- Does this open the door for other silicon solution providers, more adapted to rapid design turnaround?
Thanks for reading! Feel free to contact me if you want to discuss this post and the underlying questions raised. For more information on Strategy Analytics’ extensive coverage of the automotive industry, take a look at the PBCS (Powertrain, Body, Chassis & Safety), AVS (Autonomous Vehicles Service), AIT (Automotive Infotainment and Telematics) and ACM (Automotive Connected Mobility) services.
See you at the 23rd International Automobil-Elektronik Kongress in Ludwigsburg next week.
