Automotive is one of the fastest-growing segments in the semiconductor industry. the total market for memory (DRAM) and storage (NAND/NOR). in this segment growing from $4 billion in 2021 to $10 billion in 2025. Over 97 million cars are projected to be sold in 2025 with an average of 16 gigabytes (GB.) DRAM and 204GB of NAND in each car. In other words, by 2025, a typical car will have three times more DRAM. and four times more NAND compared to a car sold in 20211. This article discusses the megatrends in automotive. as shown in Figure 2, and how they impact memory and storage growth.
The Society of Automotive Engineers (SAE) defines six levels of autonomy, from Level 0 (L0) to Level 5 (L5), where L0 has no driver assistance features. while an L5 car can drive on its own in all scenarios without a driver, as shown in Figure 3.
Companies such as Waymo and Cruise have put in a great deal. focus on achieving L4 or L5 autonomy for robot axis. which require significant amounts of memory and storage. make critical decisions without driver intervention. Similarly, implementing even more limited L2+ or L3 autonomy requires more memory and storage. for advanced driver assistance and safety (ADAS) features. such as blind spot monitoring, adaptive cruise control,
lane departure warnings, and driver monitoring systems. As shown in Figure 4, by 2030, almost 3 million cars are expected to be fully autonomous and more than 15 million cars are expected to support at least L2+/L3 autonomy.
With increased autonomy, the reliable operation of electronic components. becomes important because malfunction can lead to hazardous situations. Automotive industry standards, such as ISO 26262, with functional safety. aim to ensure a car can navigate or safely come to a stop on its own when a critical system error is detected.
The different levels of functional safety are known as Automotive Safety Integrity Levels (ASIL). where ASIL A is the least stringent (for example, for rear lights) and ASIL D is the most stringent (for example, for braking). Software is also becoming increasingly important in automotive. Today’s high-end cars can have up to 100 million lines of code, and in the future, an L4/5 car could have between 300 million to a billion lines of code2. In addition to the system software, products like eMMC, UFS, and SSDs have embedded software (also known as firmware) to manage the functionality of the storage device. To ensure that software development methodology meets best in-class practices, Automotive Software Process Improvement and Capability Determination (ASPICE) is increasingly being requested by automotive customers. ASPICE has five levels, from L1 to L5. L5 is the most demanding while L3 is widely accepted as the standard for excellence. Memory and Storage Implications to Support Increased Autonomy Memory Requirements • Greater than one terabyte/second (TB/s) memory bandwidth to support the compute performance needed for self-driving cars (L4/L5) • Higher interface bit width for improved throughput • Significant increase in DRAM capacity with greater than 50GB needed for L4/5 autonomy • Highest level of functional safety (ASIL D) for DRAM
Storage Requirements • Increased endurance for L4/L5 event-based data logging applications (black box), potentially requiring 150 petabytes3 (PB) of total bytes written • Significantly increased densities with greater than 1TB of storage for L3+ autonomy • Functional safety level ASIL B or higher with ASPICE L3