China's first. How? Why has no one been able to achieve it before? And the key question is, is it necessary for RISC-V to get on board?
What did the Great Wall do?
A few days ago, Wei Jianjun posted this on Weibo:
Great Wall Motors has made a car-grade chip. Strictly speaking, it was initiated and developed by Great Wall Motors and jointly developed with multiple partners. "Incubation and cultivation" is a more appropriate description.
Let’s briefly explain the meaning of “lighting up”: it means that when the chip has the correct signal and power input, it can successfully perform the self-test and initialization process, and then start to perform the scheduled tasks. Lighting up is an important milestone, indicating that the design and manufacturing process has been successful, followed by the testing and mass production phases.
There are many domestic car companies that "make cores", but the route Great Wall chose to enter the semiconductor industry and the answers it submitted are obviously different.
Great Wall does not follow the popular self-developed AI computing chips in the automotive industry. The Bauhinia M100 series products are actually MCU chips for car body applications.
MCU can be seen as a specific chip that controls a certain category of vehicle functions, such as driving instruments, engine, chassis, etc. In essence, it is a single-chip microcomputer that appropriately reduces the main frequency and specifications of the CPU, and integrates multiple functional modules and interfaces such as memory, timers, A/D conversion, clocks, I/O ports, and serial communications on a single chip to realize terminal control functions. It has the advantages of high performance, low power consumption, programmability, and high flexibility.
They don’t sound as cutting-edge and “high-end” as smart driving AI chips and cockpit chips, and their costs and prices are far lower than them. However, MCUs are an indispensable underlying key component of any car – under a classic car architecture, there are at least dozens of MCUs and as many as hundreds.
In the field of automotive MCU, which is already highly mature and monopolized by overseas giant suppliers, Great Wall has made leadership in many different dimensions in the Bauhinia M series.
The most basic are absolute performance parameters. Officially revealed that the Bauhinia M100 CoreMark is as high as 2.42 (CPU performance), which is 38% higher than competing products and responds faster to the entire machine. And it meets the automotive ASIL-B level requirements. HSM supports national security SM2/3/4, as well as ISO21434 network information security standard.
Even more advanced is the scalability of the Bauhinia M series. For example, 1MB Flash storage space can cover current application scenarios and meet the needs of the next generation EE architecture. It also provides a variety of packaging forms such as 64-144 pins, which can provide more flexible design solutions and reduce costs.
Subsequent models that are still under development will cover power, chassis, domain control and other needs with higher security requirements, higher integration, and greater calculation requirements.
And the most important point that Wei Jianjun emphasizes both inside and outside his words is also the real innovative breakthrough of Great Wall Bauhinia M100:
It does not use the ARM architecture commonly used in car-grade chips, but uses the open source instruction set RISC-V architecture. It is the first car-grade chip in China designed based on the open source RISC-V architecture.
His original words were:
Is that true?
The car core uses RISC-V, what does this mean?
Let’s first understand what an instruction set architecture is.
The instruction set is the core of the operation of the computer system. It stipulates the data type and format, instruction format, addressing mode and the size of the accessible address space, the number, digits and number of registers accessible to the program, and the definition of control registers...
Just like the "Operation Guide" of a computer, no matter what tasks are to be performed with the chip, the "standards" of the instruction set must be followed.
Currently, the two mainstream instruction set architectures are X86 and ARM. X86 is jointly developed by Intel and AMD, and ARM’s intellectual property rights are in the hands of ARM.
But X86 and ARM are not perfect. For example, although X86 is mature and stable, it consumes high energy, has a large processor size, and has a complex system... while the ARM architecture is the opposite of X86.
This results in a situation where X86 dominates servers and desktops, and ARM dominates mobile. The automotive-grade chips we are familiar with have to choose the ARM architecture due to constraints such as energy consumption, volume, cost, etc.
However, the characteristics of ARM also determine its relatively low performance, poor compatibility, poor scalability and other regrets.
But the most important thing is that both X86 and ARM are private instruction sets. If you want to use them, you have to "pay taxes" to Intel and ARM, and future maintenance and upgrades will also rely on the technical support of these giants.
However, RISC-V, which appeared in 2010, is considered a "dark horse" that subverts the chip track pattern - completely open source, allowing anyone to design, manufacture and sell RISC-V chips and software.
In addition to the advantages of "bypassing the barriers and royalties of proprietary instruction sets", RISC-V's openness and flexibility attract developers around the world. It is designed to be simpler and can be customized and expanded according to specific application needs. This lightweight and modular design makes RISC-V perform particularly well in fields such as the Internet of Things, edge computing, and AI.
So, although RISC-V is not the first open source instruction set, according to SHD data, RISC-V adoption will grow at a compound annual growth rate of 40% in the next few years. By 2030, chip shipments based on the RISC-V architecture will exceed 16 billion. Especially it occupies an important position in mainstream applications such as MCU, 5G, and industrial automation.
Seeing this, everyone should also react: unrestricted open source, strong customization, and good edge computing performance.
It fully caters to the pain points of the intelligent transformation of car companies: open source means that the cost and risk of the supply chain are controllable, strong customization means good scalability, a set of IP can adapt to the needs of different models, and edge computing is in line with the new wave of AI-based smart cars.
The first two points respectively correspond to the automotive industry's century-old demands for risk resistance, cost reduction, and platformization. The third point brings more choices and paths to the future "one-brain" centralized EE architecture.
Judging from the technical characteristics of RISC-V, smart cars are the perfect "carrier" for implementation.
In fact, the introduction of RISC-V into cars has been actively discussed in academia and technology companies, including Ali Pingtou Ge, etc., who have launched in-car products, but not many car circles have responded.
To truly take the first step towards the goal of "mass production and launch", Great Wall is at the forefront of the industry.
The change began with the "chip shortage" that was a hot topic in the automotive industry two years ago.
Great Wall launches car-spec RSIC-V, what impact will it have?
The chip shortage will occur in 2021-2022. The main shortage is MCU chips, which have never been taken seriously by car companies: demand is large, unit price is low, supply is stable, and technology is not the most cutting-edge, and no one expected problems.
How serious was it then? Bosch, one of the "big seven" overseas companies, had almost all of its Asian factories forced to suspend production and reduce production. Major car companies took turns "bombing" Bosch's senior management in China, so much so that the CEO at the time said he was under so much pressure that he would "jump off the building."
Great Wall revealed that it will be forced to reduce production by 408,000 units in 2021, accounting for 24.5% of the annual planned output, and by 210,000 units in 2022, accounting for 16.5% of the annual planned output.
So today Great Wall has launched its self-developed RISC-V, which of course has positive significance at the user level, because the vertical integration of the supply chain will inevitably lead to cost reduction, technology decentralization and popularization, and an improved car experience.
But more importantly, at the industrial level, the chip design, testing, packaging and other companies connected in series behind Bauhinia M series MCU have initially built China's own automotive-grade semiconductor supply chain, and intellectual property rights are no longer exclusive to a certain company, and the design and maintenance of automotive cores are no longer controlled by others.
Like all car companies, Great Wall is also exploring cutting-edge technologies such as smart driving and cockpits. But at the same time, it is trying to solve the technology, supply, and safety issues of the underlying semiconductor needs-and it is for the Chinese automobile industry, because the incubation results of the Great Wall Forest Ecology are never only for itself, but open to the entire industry.
It can be foreseen later that as Chinese automobiles take the lead in the RISC-V architecture and go overseas, it is likely to have a huge impact on ARM's dominance of car cores.
On a deeper technical level, because each era has its own car, it must also correspond to the most suitable chip architecture. It can also be seen from the RSIC-V follow-up products disclosed by Great Wall, which is obviously not just a simple proposition of MCU independent replacement.
Including the connection of several different domain controls, AI car control technology closely related to the wire-controlled chassis, and centralized electronic and electrical architecture... These are hotly discussed in the industry and have been looked forward to for many years. It is recognized that they will bring about huge changes in technology. Great Wall has taken the lead in exploring the layout of the entire industry.
Similar scenarios have actually been played out repeatedly by Great Wall. For example, it began deploying smart driving 10 years ago and now leads the end-to-end car-boarding competition.
The Blue Mountain Intelligent Driving Edition topped the sales list of six-seat large SUVs, causing Great Wall's stock price to rise continuously.
Great Wall's advance layout and R&D investment are being transformed into a series of progress, and are receiving intuitive results, affirmation and recognition.
The lighting of RISC-V chips is actually just a part of Great Wall's scientific and technological achievements and intelligent achievements. It is a re-verification and latest practice of the only unchanging "iron law" in the era of smart cars:
Only technology is the only truth to match performance and support sales.
When it comes to car building, only technology is the hard power. Everything else is just fancy tricks. Flashiness will eventually ebb. Only true hard-core technology can achieve steady and long-term progress.