[Autohome?Interpretation of fresh technologies]?Huawei was founded in 1987 and is a Chinese company that started out manufacturing communications equipment. After more than 30 years of accumulation, Huawei has developed into the world's largest 5G equipment supplier. With the rapid growth of the artificial intelligence chip market, Huawei successfully launched the world's first mobile phone processor chip equipped with an artificial intelligence acceleration unit - Kirin 970 - in 2017 with the help of Cambrian's chip IP. The success of this chip has further convinced Huawei of the development potential of artificial intelligence technology and accelerated its pace of self-developed artificial intelligence processors. The domain controller of the autonomous driving system is where artificial intelligence chips come in handy. How powerful is Huawei's artificial intelligence chip? How does it empower self-driving cars? Let's take a look today.
●? Let’s first talk about the familiar mobile phone processor NPU. Huawei’s self-developed artificial intelligence chips were first used in mobile phone chips, and were later applied to servers and automotive autonomous driving systems. Next, let’s first talk about the familiar mobile phone processor equipped with an artificial intelligence acceleration unit. Huawei's artificial intelligence acceleration unit is called "NPU", and its full name in English is "Neural-network Processing Units", which translates to neural network processing units. The NPU can accelerate face recognition, speech recognition, image recognition and other functions on mobile phones, thereby providing mobile phone users with a better experience. In "How big is the gap with foreign countries?" I mentioned in "Talk about China's Autonomous Driving Chips" that the NPU mounted on the Kirin 970 processor actually comes from Cambricon's Cambricon-1A processor IP, which is installed on the Huawei Mate?10 mobile phone. The NPU of the Kirin 980 processor subsequently released by Huawei (mounted on the Huawei Mate?20 mobile phone) uses the Cambricon-1H processor IP of Zhongke. Judging from the parameters currently published online, the NPU computing power of Kirin 810 is 0.83TFLOPS (Note: 1TFLOPS = one trillion floating point operations per second), and the NPU computing power of Kirin 820 is 1.33TFLOPS. The manufacturer has not announced the specific computing power values ??of Kirin 990 and Kirin 985, but according to data from the AI ??Benchmark website of Zurich Institute of Technology, the NPU computing power of Kirin 990 is higher than that of Kirin 820. According to test data, its computing power is estimated to be 2TFLOPS. above. The NPU computing power of Kirin 985 will be slightly higher than that of Kirin 820. Kirin 710A is the car-grade version of Kirin 710 launched in 2018. This chip has completely independent intellectual property rights and will be manufactured by SMIC. It is a pure "Chinese chip". After Huawei's Kirin 710 enters the automotive industry and is successfully launched, it will seize market share with Qualcomm Snapdragon 820A. As these new cars equipped with Snapdragon 820A are launched one after another, they have set off a wave of high-tech cockpit trends. Huawei's Kirin 710A enters the market at this time, just in time to catch up with this trend. I believe that many Chinese brand companies will consider using Huawei's digital cockpit chip with pure Chinese origin in the future. Although there is a certain gap between the performance of Kirin 710A and Huawei's latest Kirin chip, and there is no integrated NPU core, there is no problem in meeting the performance requirements of multi-screen display and interaction in the digital cockpit. Huawei's entry into the automotive field through digital cockpit is just a test of the waters, and further deployment in the field of self-driving car chips is the key.
●?Huawei’s autonomous driving network architecture Huawei’s founder and president Ren Zhengfei once said in an interview with the media that Huawei will not manufacture complete vehicles, but Huawei will manufacture Internet of Vehicles modules, electronic parts in cars, and also Possibly the best in the world. In this case, Huawei will definitely seize the dominance in the two fields of Internet of Vehicles and autonomous driving. If Huawei's ADN target architecture becomes an industry standard, it will become an extremely important profit growth point for Huawei in the next 10 years.
Huawei's goal is very ambitious, that is, to cover various communication and computing equipment for high-end autonomous driving systems and provide complete one-stop solutions. In layman's terms, it is to achieve the effect of "If you want to buy communication and computing equipment for self-driving cars, look for Huawei." Although this article focuses on chips, Huawei's strength lies not only in chips. It is also outstanding in network communications. Before talking about chips, I would like to give a simple example of vehicle-road collaboration to illustrate this point. This simple example can well illustrate how Huawei empowers autonomous driving through network and communication technologies. It can be said that the future self-driving car is by no means an "isolated island", but a mobile tool connected to everything.
●? Now that Huawei’s autonomous driving “core” and soft capabilities are complete, we briefly talked about the application of Huawei’s network architecture in vehicle-road collaboration. Now let’s formally talk about Huawei’s autonomous driving platform and chip. Since the autonomous driving platform is based on Huawei's artificial intelligence chip, we still have to briefly understand Huawei's artificial intelligence chip. The applications of Ascend 910 are geared towards cloud devices that require extremely high computing power, while the applications of Ascend 310 are geared towards terminal devices that have certain requirements on energy consumption ratio, but there will be some overlap between the two. In the table below, I summarize the current information about Huawei’s Atlas series products using the above two chips. For artificial intelligence acceleration chips to actually run, they need the cooperation of the CPU. Some of the Atlas series products mentioned above are equipped with Intel's server-class CPU, while others are equipped with Huawei's own Kunpeng 920? CPU chip. This is very different from the practice of most other chip manufacturers that only do B2B business and only provide development documents after signing a contract. Regardless of the developer's ability, more developers adopting Huawei's equipment development will help build Huawei's artificial intelligence ecosystem faster, which will accelerate Huawei's development in the field of artificial intelligence. Of course, Huawei will also provide targeted software services for different segments. Due to space limitations, we will not go into details here. Interested friends can go to Huawei Cloud to learn more. While other chip companies are still struggling to find breakthroughs in the industry, Huawei has provided comprehensive and complete solutions for various industries. Looking at the dense list of solutions for various industries on the official website, I deeply feel how powerful Huawei is. Perhaps it is Huawei's strength that is an important reason why other companies are struggling to pursue it.
●? MDC autonomous driving platform equipped with Ascend 310 Okay, after talking about chips, let’s talk about Huawei’s autonomous driving platform. Huawei's latest flagship autonomous driving platform product is the MDC600, which can meet the performance requirements of domain controllers for L4 level autonomous driving. For L3 level conditional autonomous driving, Huawei has launched the MDC300. The Ascend 310 single-chip computing power is 16TOPS, the power consumption is 8W, and the energy consumption ratio is 2TOPS/W; the FSD chip on the Tesla Autopilot?3.0 processing unit has a single-chip computing power of 72TOPS, the power consumption is about 36W, and the energy consumption is The ratio is 2TOPS/W; NVIDIA's latest DRIVE AGX Orin platform, equipped with the Orin chip, has a single-chip computing power of 200TOPS, a power consumption of 45W, and a power consumption computing ratio of 4.4TOPS/W. In comparison, the energy consumption ratio of Shengteng 310 has caught up with the international mainstream level. The Ascend 310 is currently manufactured using TSMC's 12nm process. As the production process is upgraded to 7nm or even 5nm in the future, there is room for further improvement in its energy consumption ratio.
The MDC autonomous driving hardware platform and the in-vehicle operating system Hongmeng kernel have successively obtained the ASIL-D level in the ISO26262 international standard for road vehicle functional safety (hereinafter referred to as "ISO26262 ASIL-D level"), and will follow Huawei's Adaptive AUTOSAR architecture If the development tool chain and autonomous driving algorithm can also pass the ISO26262 ASIL-D level, then Huawei's MDC autonomous driving system will become the world's first commercial autonomous driving system that meets the ISO26262 ASIL-D level. What does it mean to comply with ISO26262? ASIL-D level? This means that the MDC autonomous driving system on the vehicle will only experience a random hardware failure every 100 million hours. The safety factor is quite high. The probability of ordinary users encountering random hardware failures when using the MDC autonomous driving system is extremely low. The system architecture of the MDC autonomous driving platform is scalable. By increasing or decreasing the number of CPU cores, the number of artificial intelligence acceleration cores and the number of IO interfaces, it can meet the needs of high-, medium- and low-end passenger cars from driving assistance to high-end intelligent driving. different usage scenarios. To date, Huawei's MDC autonomous driving platform has signed more than 18 customers, including SAIC, Geely, JAC, FAW Hongqi, Dongfeng Motor, Suzhou Jinlong, Neolithic, Shandong Haorui Intelligent and other companies are on the list. From the cloud to the terminal, Huawei has a complete set of autonomous driving solutions. So can we say that Huawei is invincible? No! What Huawei lacks in the field of autonomous driving is massive big data on actual road conditions. Without this data to train the autonomous driving system, the system can only stay in laboratories and closed test drives. To this end, Huawei is also actively cooperating with relevant companies, such as NavInfo, which owns map resources, and a number of Chinese vehicle companies to make up for its own shortcomings in road condition big data and further build up its technical barriers in the field of autonomous driving.
●? Editor’s summary: Huawei’s influence in the industry is unquestionable, and it is at the forefront of the world in terms of technology and products. A big tree attracts trouble, and Huawei has been sanctioned by the United States in recent years. If the status quo continues for a long time, it will have a considerable impact on Huawei's artificial intelligence chips, which this article focuses on. How to solve the problem is a difficult problem facing Huawei. Judging from some current information, Huawei's measures such as building a chip factory in the UK, developing RISC-V architecture processors, and cultivating the HMS cloud service ecosystem are the key to its breakthrough. We will also Continue to pay attention to further developments of the situation. (Photo/Text/Autohome?Chang Qinglin)