Jiwei Network reported that the "era of software-defined automobiles" is setting off an unprecedented tide of change for the century-old automobile industry. Analyst roland Berger pointed out that 2021 will become an inflection point milestone in the development of the intelligent electric vehicle industry; in 2022, the trend of industrial disruption will be further broken, and the innovation of the whole industry chain will also blossom.
"The smart car industry is a parent ecosystem that is driving innovation in materials, energy, transportation infrastructure, chip and semiconductor technology, software technology, intelligent driving, artificial intelligence technology and so on." Dr. Yu Kai, founder and CEO of Horizon, expressed a similar view at the just-concluded 2022 China Electric Vehicle 100 Cloud Forum.
"Changing lanes and overtaking" on the new energy vehicle track is an important step in the rise of China's automobile industry. Since 2015, China's new energy vehicles have achieved the world's first production and sales for 7 consecutive years. According to data from the China Automobile Association, last year, the passenger cars of Chinese brands reached a market share of 44.4%, of which the growth mainly came from new energy vehicle brands, and China has become an important driving force for the transformation of pure electric drives in the global automotive industry. In the first half of the competition of the automobile revolution, the start of China's auto industry has achieved good results.
However, in the second half of the tiebreaker, the competition around intelligent network technology is obviously more critical, and more extensive technological change and ecological reshaping will also deepen. At the same time, with the continuous landing of intelligent applications and the continuous improvement of the penetration rate of vehicle intelligent network functions, the transformation of the automobile industry to technology-driven has become an inevitable trend. In fact, the automotive industry has begun to take action, major vehicle companies, parts suppliers, chip manufacturers, etc. have increased investment in technology research and development, improved the depth of the supply chain, and created a new ecological cooperation model, which will be the next stage of the industry theme.
What to fight in the second half? "Chip + OS" is the digital base of the smart car era
At the 2021 World New Energy Vehicle Conference, the industry has reached a consensus that the market penetration rate of new energy vehicles has continued to increase, and it is currently in a comprehensive market-oriented expansion period oriented to user demand, and has ushered in an inflection point of comprehensive marketization.
The power system has shifted from an internal combustion engine drive to a pure electric drive, the control system is evolving from distributed to centralized control, and the automotive products are moving from closed systems to open systems – this once-in-a-century change in the automotive industry has reached the second half.
"A truly competitive technology, it's not just developed in the lab, it's actually polished in the user experience and feedback." This is an important recognition for us. Yu Kai pointed out. Jiwei Network noted that Yu Kai attended the forum with a new identity - the chief scientist of Horizon, and the focus on "technology" further implied the overall trend of the automotive industry to turn to technology-driven.
So what is the core of the competition in the second half - automotive intelligent technology?
In this regard, Yu Kai pointed out that intelligent computing chips and intelligent operating systems will be the innovative digital base for the future development of the automotive industry. This is also the part of the domestic industry that needs to be broken through urgently.
As the "brain" of smart cars, smart chips are undoubtedly the key to the industry competition. In the era of big data and artificial intelligence, according to IDC statistics, the demand for global computing power will double every 3.5 months, far exceeding the current growth rate of hashing power. With the improvement of the level of autonomous driving, the amount of data generated by cars will increase exponentially, and the competition around large computing power has already begun. Since last year, among the models released by many car companies, the demand for the autonomous driving computing power platform equipped with them has reached the 1000 TOPS level.
But TOPS should not be the only indicator. Horizon believes that in the era of artificial intelligence (AI), a new optimization indicator is needed - "the new Moore's Law in the ERA of AI", proposing that "the computing power is large, it is better to calculate faster" - this is the unique way of playing that Horizon gives to the computing power anxiety of the global industry.
Yu Kai further explained the three factors contained in this optimization indicator at this forum. The first factor, the peak computing power, is actually the peak TOPS that the industry is constantly mentioning in the old Moore's Law era. "But what percentage of so many hashing resources are being used effectively?" Yu Kai pointed out that this actually depends on the innovative architectural design of the chip hardware, which is also the second key factor. The third factor is the software algorithm of artificial intelligence. Yu Kai pointed out that the architecture of artificial intelligence software algorithms has been evolving continuously over the years, and it is foreseeable that the theory of artificial intelligence software algorithms will continue to change in the next 20 years, which means that chip architectures need to be able to continuously adapt to the evolution of software algorithms.
It is worth mentioning that a similar view was mentioned in a MIT scholar paper published in the journal Science in June 2020, arguing that the improvement of computing performance in the post-Moore era lies in the joint optimization of software, algorithms, and hardware architectures.
Yu Kai revealed that the Horizon Journey 5 chip is designed based on such a new Moore's Law idea. Jiwei Network learned that in the Horizon high-performance autonomous driving processor architecture BPU (Brain Processing Unit), a large number of innovative technologies are used, including the use of two independent or collaborative cores to achieve high performance, low latency, data concurrency processing technology, pulse array to improve parallel speed, the use of near-memory computing to make storage and computing closely integrated and sparse acceleration technology, etc., condensing the in-depth insight into AI, deep learning and autonomous driving scenarios. Enable efficient AI computing.
From the result point of view, the Journey 5 chip is compared with the NVIDIA Orin chip, although the Journey 5 only uses half of the chip area and half of the computing resources, but when using advanced neural networks for computing, it can get considerable computing performance. Yu Kai said that the more advanced the neural network, the higher the acceleration performance of the horizon chip.
However, innovative chip design architectures are just one, and there is a simultaneous focus on intelligent operating systems (OS). "Chips and OS are never separated." Yu Kai stressed that there have been many close alliances between chip architecture and operating systems in history, such as Windows and Intel, Android and ARM, etc., because software is difficult to adapt to multiple chip architectures, and only a high degree of software and hardware coordination can maximize computing performance.
From this point of view, in the decisive game of automotive intelligence in the second half, chips and OS will be the innovative digital base of the intelligent car era. It is based on this industrial technology logic that Horizon launched the open source OS combined with the Journey 5 chip - TogetherOS, and worked with partners to build an in-vehicle OS base to help the industry build an in-vehicle OS ecosystem.
Yu Kai stressed that the OS for smart cars should be a public resource, it is an "open, open source, widely participated in the TogetherOS without commercial purposes."
Open BPU IP licensing supports some automakers' self-developed chips
But Horizon doesn't seem to be satisfied with that. In Yu Kai's words, can we continue to promote "making opening up more open"?
"Can we also support some automakers to independently develop differentiated autonomous driving chips with core capabilities?" Yu Kai said that Horizon opened its high-performance autonomous driving processor architecture BPU IP licensing model was also born.
According to reports, Horizon will open up the software support package and chip reference design including the entire BPU to support some OEMs' self-developed chips, "Although we do not think that all OEMs have to develop their own chips, but the benefits of making their own chips are also obvious, which can enhance differentiated competitiveness and accelerate the speed of R&D innovation." Yu Kai explained that through the authorization mode of BPU IP, the development of the whole vehicle can achieve a high degree of coordination from the chip, operating system to the entire automatic driving software, which will bring significant improvement to the iteration speed of the whole vehicle.
The speed of iteration is particularly important in the era of smart cars.
The 2021 Ideal ONE model, which went on sale last May, replaced its predecessor, Mobileye, with two Horizon Journey 3 chips. The Two Journey 3 delivers efficient AI performance for the 2021 Ideal ONE to achieve NOA navigation assisted driving.
It should be known that based on the consideration of supplier stability and reliability, car companies will not easily replace important chip suppliers, but in the era of automotive intelligence, the iteration cycle of automotive functions has been greatly shortened, and the development efficiency has directly affected the market competition performance of car companies. However, Mobileye's non-open strategy makes the ideal ONE unable to cope with the fierce market competition through the rapid upgrade function of OTA, which is also the core reason for the ideal replacement of chip suppliers.
Yu Kai revealed that from the beginning of cooperation between the two companies to the mass production and listing of new models, it took only seven or eight months before and after, and such an ultra-short cycle is a record in the global automotive industry. Based on the design concept of combining soft and hardware, Horizon has made a special optimization design for the ideal automatic driving AI algorithm of this model based on the Journey 3 chip, and the overall computing efficiency is higher, while the open business model of Horizon is more efficient and supports the ideal for software full-stack self-development.
Because the experience logic of users in the era of smart cars is completely different from the era of traditional fuel vehicles, "it used to be a process-oriented form for supply chain optimization, and now it will become more and more like the smartphone market, which means that if you can be like Apple for even six months ahead of competitors, and continue to lead, you win." Yu Kai stressed that in the era of intelligent cars, creating a more open innovation ecology is always the most important, "the chip is in the most upstream link in the entire intelligent information industry ecology, it is naturally an ecological participant, we want to cooperate extensively with sensor manufacturers, Tier1, software manufacturers, intelligent driving technology providers in the ecosystem." ”