Memory in automotive designs: An industry in transition

Memory chips serving automotive markets have come a long way over the past decades, and if there is one area that has driven the shift in their use inside vehicles, it’s the adoption of advanced driver assistance systems (ADAS) and autonomous driving technologies. New automotive technologies—radar, lidar, hi-resolution imaging, and object recognition—have mandated high-density memory for internal and external perception applications.

So, Micron Technology Inc., celebrating 30 years in the automotive memory business, is now reinvigorating its focus on memory devices serving ADAS-enabled and autonomous driving vehicles. Here, Garima Mathur, director of automotive strategic marketing at Micron, is quick to point out that the Boise, Idaho-based memory supplier has been in the memory market for automotive when it wasn’t that big or exciting.

She also notes that memory sockets in automotive have changed a lot in the past decade, citing two significant automotive trends: autonomy and rich infotainment. Regarding autonomy, Mathur acknowledged that a few years ago we thought that by now every car would be driving by itself in certain parts of the world. “That process is now happening in steps, and the current goal is to make our cars safer.”

Regarding infotainment, she pointed to screens and heads-up displays, which seem to emulate the smartphone experience inside the car. “We are bringing our digital lifestyle in the cars.” Mathur added that these things are very exciting from memory and storage perspective.

Figure 1 Everything except electrification is driving the need for higher densities in memory and storage. Source: Micron

“Everything except electrification is driving the need for higher densities in storage,” she said. “As we increase the level of autonomy, there are more sensors in the car, which inevitably leads to more sensor fusion and compute requirements.” In other words, when the car is acting on its own, it needs more compute, and to support that compute, more DRAM is required to store all the data points.

Next, Mathur added that memory and storage inside the cars must be automotive grade, which encompasses certain certifications and qualifications because quality and reliability are a critical part of the value proposition. “Even at the design stage, we incorporate all the automotive requirements into the wafer design process, which goes all the way to end products.”

Functional safety-compliant DDR5 DRAM

Micron claims to have launched the industry’s first automotive low-power DDR5 DRAM (LPDDR5) memory that is hardware-evaluated to meet the most stringent Automotive Safety Integrity Level (ASIL) requirement for functional safety: ASIL D. These functional safety-evaluated DRAM chips are compatible with ADAS applications like adaptive cruise control, automatic emergency braking systems, lane departure warning, and blind spot detection systems.

“Autonomous vehicles need powerful, trusted memory that can enable real-time decision-making in extreme environments,” said Kris Baxter, corporate VP and GM of Micron’s Embedded Business Unit. He also pointed out that Micron’s hardware evaluation of DRAMs has been independently assessed and verified by exida, an automotive safety expert.

Alexander Griessing, COO and principal safety expert at exida, added that while functional safety is essential to developing advanced automotive systems, memory has had a somewhat neglected commercial off-the-shelf existence. “Micron’s automotive LPDDR5 with a laser focus on ISO-26262 functional safety is setting a new standard for the rest of the memory industry.”

Apparently, data-intensive automotive technologies are on the rise; the ADAS-enabled vehicles now run over 100 million lines of code. That, in turn, requires hundreds of tera operations per second, and here, LPDDR5 can address these requirements with a 50% increase in data access speeds and more than 20% improvement in power efficiency.

As a result, these DRAMs enable modern vehicles with near-instantaneous decision-making from the fusion of multiple sensors and inputs. Micron’s automotive LPDDR5 is also ruggedized to support extreme temperature ranges and qualified for automotive reliability standards such as AEC-Q100 and International Automotive Task Force 16949.

Figure 2 LPDDR5 enables high-performance compute for cars while minimizing power consumption for both electric and conventional vehicles. Source: Micron

Market research firm Gartner projects that the automotive memory market will grow to $6.3 billion in 2024, more than doubling from $2.4 billion in 2020. It’s worth noting here that, according to Yole Intelligence estimates, DRAM and NAND flash dominate the automotive memory market with a combined share of 80%. Here, DRAMs capture 41%, NAND flash chips acquire 39% and the rest is shared among memory technologies like NOR flash and EEPROMs.

No more focus on legacy memory

So, what’s new in the memory space for automotive applications? Mathur told EDN in an exclusive interview that Micron is working closely with automotive design engineers as their use cases are changing. “Automotive in the past used to be on the tail end as OEMs would take something robust, well tested, and fail-proof,” she added. “Consequently, more legacy memory technologies would go inside the car.”

That landscape has changed entirely over the past decade, and there is little focus on legacy memory technologies. “For instance, on the DRAM side, we see requirements for high bandwidth both in infotainment and ADAS designs,” Mathur said. Especially on the ADAS side, memory technology requirements are changing as we move toward higher automation levels.

On the storage side, she noted that eMMC used to be mainstream. “But we are seeing a demand for higher performance, so USF is now picking up pretty strong.” That shows how densities will grow in the future, whether it’s UFS or other solution, Mathur added.

Figure 3 The architectural shift in automotive compute designs is expected to drive more demand for memory devices. Source: Micron

In the final analysis, she noted that while these are end consumer trends, what’s happening under the hood is architectural changes. “There used to be hundreds of ECUs in a vehicle, but they are being consolidated now,” Mathur said. “Automotive compute designs are moving from discrete ECUs to domain-based architecture to zonal/central architecture.”

According to her, that will have a fundamental impact on the compute content, and subsequently, on memory and storage that go in the car. The memory company, which has been in the automotive market for 30 years, is confident that this shift will benefit it. Especially when it’s a one-stop shop for memory devices going into modern vehicles.

Related Content

Memory use in automotive
Busy Road Ahead for Automotive Memory
LPDDR5 DRAM ready for Level 5 autonomy
Automotive memory: Many types and applications
Another NAND Flash for Automotive, OTA, AI and More

googletag.cmd.push(function() { googletag.display(‘div-gpt-ad-native’); });

The post Memory in automotive designs: An industry in transition appeared first on EDN.

Source link

Related Articles

Leave a Reply

Your email address will not be published. Required fields are marked *

Back to top button
Translate »