[Memory Tech Day 2023] Provisioning a Software-Centric Automotive Future

Beyond the usual demand for lower price for higher value, what factors are driving increases in memory capacity and operating speed? In the automotive sector, it is the compute infrastructure of software-defined vehicles that has become the motive force for must-have advanced technologies – ones that might only be considered as “nice-to-have” in other sectors. Consider just the software itself, which in today’s vehicles requires hundreds of millions of lines of code. This count quickly reaches into the billions as self-driving features, upgradeable infotainment, and expanded OTA (Over-the-Air) update mechanisms are introduced. Now factor in all the storage required by the new applications themselves, and the urgency for novel memory capacity and bandwidth improvements becomes evident. The movement to a centralized computing architecture, alongside industry momentum towards a more vertically integrated model where manufacturers target in-house control and IP ownership of the entire software stack, means that fewer but more powerful processors will be accessing the shared memory and storage. Implemented in hardware, these trends must inevitably result in:

• • Wider and higher bandwidth memory buses
  • Secure, virtualized spaces for use by multiple applications
  • Flexible chip packaging models to allow product differentiation without “reinventing the wheel” for each customer
  • Readily attachable storage for expansion and customization

Perhaps even more important is the quality control aspect. With fewer components and more centralized operations, the impact of a single point of failure can be devastating. Novel design, verification, and testing methodologies ensure that cost-saving redesigns bring with them life-saving quality controls. Our presenters provide insights into the journey towards autonomy and the consolidation of central computing architecture in software-defined vehicles, showing how Samsung’s memory solutions are enabling the revolution behind the steering wheel. In the beginning of the session, Donovan Hwang provides an overview of Next Generation Intelligent Mobility Architectures and trends. His presentation brings focus to key drivers.

• • Automotive is one of the fastest-growing markets in semiconductors with the ~1000 chips in a typical vehicle today expected to triple in the near future.
  • Vertical Integration means manufacturers are taking their supply chains in-house for critical elements.
  • An architectural transition, from today’s distributed ECUs (Electronic Control Unit) to clustered domain- and zone-centric models, brings with it a radical change to how software is handled: a software-centric model.

As part of the vertical integration trend, software-centric development shows up as one of the critical elements moving in-house with good reasons.

• • Cars run on code, and the ~150 million lines of code needed today will exceed 5 billion lines for a self-driving vehicle.
  • Over-the-Air (OTA) updates are already difficult with 150 software modules in a hundred different languages.
  • For those modules whose IP is licensed from external vendors, enabling OTA updates become an even more onerous task.

Transformation to a centralized software model becomes key to managing future expansion, but will be incredibly difficult to achieve given the distributed, heavily outsourced, and largely haphazard software approach followed today. However, succeeding in this challenge reaps a huge reward: a universal OTA software download and installation infrastructure that can be used for software upgrades and applications. Just as with a cell phone, automotive after-sale revenue coming from app store and feature-upgrade purchases is expected to ramp significantly.

Rohit Bhola delves into the solution details of the next generation intelligent mobility architectures, contrasting today’s typical memory technology choices with the higher capacity solutions that will be needed as autonomous driving features become more mainstream. Keeping up with increasing DRAM demands in automotive applications:

• • Today’s X32 LPDDR4X 4.266Gbps will be moving to x64 LPDDR5X 8.5Gbps.
  • As more autonomous-driving features are introduced, LPDDR5X 8.5Gbps, GDDR6 12Gbps, and GDDR7 32Gbps (per Pin) will be fundamental technologies that provide the essential memory bandwidth.
  • In the fully self-driving future, moving to High Bandwidth Memory (HBM, a 3D-stacked DRAM) could become justified.

Evolving NAND to meet automotive needs of tomorrow:

• • Today’s solution, eMMC, provides a solid foundation.
  • For the ultimate in configurability, Samsung introduces the automotive-grade Detachable SSD.
  • To increase storage efficiency within a given capacity, Samsung offers the CMM-HC (CXL Memory Module - Hybrid Compute) data compression solution.

Novel chip packaging technologies like the Universal Chiplet Interconnect (UCIe chip-to-chip interface) will be enablers for even greater memory speed/capacity/configurability solutions of the future. In the following session, Yuchul (Mike) Hwang addresses the importance of test and quality control to meet the harsh demands of the automotive world, as well as the heightened scrutiny given customer service response. Failure rates in this sector target an incredible 1ppm over the full product lifetime vs. 100pm for normal commercial-grade products. To achieve this, Samsung needed to meet automotive-grade qualification demands and perform burn-in over automotive-range voltage and temperature extremes. But Samsung innovation goes far beyond these basic requirements with programs that:

• • Improve verification by using a “shift left” strategy, introducing detection methodologies ever-earlier during the development cycle for each new generation
  • Go beyond the usual AEC-Q100 testing by doubling test durations and adding reliability testing for aspects including indicative current flow levels, corrosion resistance, and shock and vibration tolerance
  • For Failure Analysis of returned parts, assemble a seasoned group of senior engineers to quickly determine the failure root cause and provide FA reports within two weeks
  • Select product from only the best fab output, re-evaluating fab performance on a daily basis and ensuring that anything less than “superior” quality is removed from the Automotive product flow

These methodologies are key differentiators for Samsung, and the in-depth presentation demonstrates why Samsung is the leader in this area.