Ten years ago, I used to say that cars were becoming “smartphones on wheels.” Today’s cars generate more data than ever before, so “servers on wheels” would be a better term.It tells us something about the enormous amount of real-time processing power required to run
An increasingly connected and automated car.
But unlike smartphones, most of this data must be used and processed in the car. To achieve this, automakers are turning to edge computing architectures to bring data capture, control, and storage inside the vehicle.
With this in mind, you can start imagining your car as an “edge data center on wheels.”
Why are automakers rejecting the cloud in favor of edge data centers?
Certain driving-related factors make data transfer to the cloud impractical and even dangerous. Edge data centers enable time-sensitive data to be processed at its source and quickly delivered to the end devices that need it.
The biggest issue is latency. Considering that a self-driving car produces about 1 GB of data per second, this gives you an idea of how much information it needs to process and return to get you to the grocery store quickly.
Of course, the time has not yet come for self-driving cars to become the norm. But even in the average connected he car, when he needs to make an important decision that can mean the difference between getting stuck in traffic or getting to his destination on time, he can transfer his data elsewhere. Forwarding doesn’t work. Also, in the case of EV battery information, power outage on the highway.
When it comes to self-driving cars, waiting on the road is a matter of life and death.
Also, sending all that data to the cloud is very expensive. An edge data center in a car can work cheaply because the cost of entry is low and the infrastructure is already housed inside.
Still, the cloud has a role to play. Data that is not time-critical can be fed to the cloud for later processing and analysis. Edge data centers thus provide an efficient hybrid solution to the high latency and cost challenges posed by connected and ultimately autonomous vehicles.
Increasingly intelligent vehicles require increasingly powerful processors. We’re talking magnitude tera operations per second (TOPS) of about 2 TOPS at level 2 autonomy and 4,000+TOPS at level 5.
Mixing in the demands of an EV system that needs to optimize every milliwatt of power to save energy further inflates the data processing requirements.
There are also various applications that need to process data differently. Immersive in-vehicle infotainment systems require lightning-fast performance. Advanced driver assistance systems (ADAS) require sensors that can identify different road conditions. As 5G access spreads, new solutions are needed to take full advantage of the improved connectivity across the board.
Technologies such as radar, which enables the average parking sensor, and lidar, which can be used to predict whether pedestrians will cross the road, also have different processing requirements.
All of this needs to be reduced to a single SoC (system on chip) to process locally.
The chip should be able to handle all the challenges of being in a moving car. There is no room for error. Reliability, safety, and defect-proof design are therefore paramount in developing chip design and manufacturing for automotive applications.
All these considerations must be taken into account at various stages of chip manufacturing, from process technology and memory design to final test, to create the right solution portfolio to power the connected cars of today and tomorrow. . .