Will autos carry heavy loads of memory?
MADISON, Wis. – For a highly automated or outright self-driving car in, say, 2020, how much memory capacity, to capture, process and store data, will be enough?
A consensus among experts in automotive and memory industries is this connected vehicle will require data storage of at least hundreds of gigabytes (GB) and probably a terabyte (TB).
With that prospect in mind, Western Digital rolled out earlier this week a new embedded flash drive, called iNAND 7250A, specifically designed as an automotive-grade storage device able to perform reliably in extreme conditions in and around the vehicle -- including ambient temperatures down to -40oC and up to 105oC.
Western Digital is predicting “a dramatic growth” in memory demand on the automotive market over the next five years. Martin Booth, senior director of automotive devices product marketing for Western Digital, told us the need for more data storage space per vehicle is no longer just driven by infotainment systems, but by ADAS and autonomous cars needing intelligent data storage subsystems to achieve greater safety.
But a TB of memory for a car of the future? Seriously?
We asked Western Digital and industry analysts to break it down. We began with current semiconductor storage demands.
Mike Demler, a senior analyst at The Linley Group, told us, “I don’t have a solid figure, but on average I expect it’s currently relatively small. There are about 60 processors per vehicle, but they’re mostly 8-bit MCUs with a small amount of flash per chip.”
However, Demler added, “In a luxury car like a Tesla, with the big flat screen display, the storage requirements start to look like a standard tablet (10s – 100s of GB).”
Thomas Coughlin, the founder of the data storage consultancy Coughlin Associates, agreed that the infotainment system is already using hundreds of GB for storage and memory. This will only increase with ADAS and higher-resolution entertainment. “In general what bulks up the storage needs are video--both for the car vision system and entertainment.”
Processing sensory data
The big change in the automotive memory landscape, though, is growth in the sensor data ADAS and automated cars must process, constantly and immediately. Equally important is how much of that captured data must be retained.
Typically, a highly automated car relies on a number of sensors – HD cameras and lidars – to read its surroundings. As many as eight HD cameras are already present in advanced cars, with lidar coming soon.
A Google car (now Waymo), for example, processes a gigabyte of sensor data per second, noted Coughlin.
Western Digital’s Booth explained, “Sensor data itself would be compressed first and then stored in a buffer.”
For autonomous vehicles, most of the data is processed and only a fraction is stored, Coughlin explained. Although some of this will go to the cloud, “real time data processing must occur in the car,” he added.
Continue reading the next page on Embedded's sister site, EE Times: "Future cars face terabyte memory load."