Computing power is "surging", but it can't reach automatic assisted driving

When talking about the evolution of intelligent driving in the past two years, we can use the name of the recently popular TV series "Crazy" to describe it. After the automotive field enters the intelligent track, it not only competes in horsepower, but also in chip computing power. Chips act as the "brain" of new energy vehicles. The higher the computing power level, the stronger the ability of the car system to process data and the better the service ability for software. In order to create data with a higher level of intelligence, major car companies are involuntarily in chip computing power.

Data shows that the scale of China's autonomous driving industry will grow by 24% in 2022, and the shipment growth of smart camera products will exceed 15%, giving rise to a large demand for smart chips. But do our vehicles really need such a high level of computing power, and do consumers have to pay for those "high intelligence"?

Car manufacturers are in a state of panic

Since the second half of 2020, global chip production capacity has continued to be tight, bringing a series of negative impacts such as price increases, hoarding, speculation, and snatching. Among them, the automotive industry, as one of the most sensitive industries in the downstream of the chip supply chain, has been affected by the chip shortage. It is difficult to ignore. With the popularization of intelligence in the automotive industry, functions such as smart cockpits, lidar, assisted driving, and entertainment assistance systems are used as selling points for cars, and car companies' demand for computing power has naturally risen. According to online data, the car with the highest comprehensive computing power on the Chinese market in 2022 is the ET7 from Weilai, which uses 4 NVIDIA OrinX chips, up to 1016TOPS. What does this mean? Simply put, 1000TOPS is equivalent to a chip performing 100 billion calculations per second. This level of system processes the external environment much faster than the driver's brain.

*Data source: Internet

More computing power = better performance?

It can be seen that the computing power has increased, but the autonomous driving function has not become more intelligent and advanced. This means that computing power cannot produce autonomous driving. The mainstream intelligent assisted driving systems on the market are basically between L2 and L3, and L4-level cars have not yet been mass-produced. Data shows that the computing power required for L2-level autonomous driving is only 2~2.5TOPS, the computing power required for L3-level autonomous driving is 20~30TOPS, and the L4 level requires more than 200TOPS.

At the same time, due to the lack of technological maturity and the influence of macro-environmental policies, the intelligent driving functions of major car companies have remained at the L2 stage. Therefore, it is quite a waste of resources to equip smart cars with such high computing power, and it will bring high losses to the chips, and the redundant computing power will still "run in vain" in the chips. In this way, the computing power of 100+TOPS in smart cars is indeed a bit excessively wasteful.

The computing power can be "soaring", but the direction is very important

While car companies are promoting their own models as being equipped with higher computing power and more intelligent, they have invisibly caused "computing power anxiety". In fact, an excellent intelligent assisted driving system does not rely entirely on the level of computing power, but also needs to consider the software algorithm and the degree of adaptation to the chip. For example, Tesla's self-developed FSD chip has a single computing power of 72TOPS, and each car is equipped with two, with a total computing power of 144TOPS. However, thanks to Tesla's software algorithm, its effect on the actual perception layer of the neural network is much better than Nvidia's 200-300TOPS. If the computing power is only used for entertainment functions and the chip computing power cannot be fully utilized, it would be a waste of resources.

Therefore, the greatest significance of "surge" computing power is to fully tap the computing power of chips and transform it into real autonomous driving functional resources. Both chip companies and car companies should calm down from this "surge" and delve deeper into software cooperation.