Source: Content by Semiconductor Industry Watch (ID: icbank) original, author: Li Chenguang, thank you.
Cars have never been AMD's forte.
Nowadays, after the consumer electronics market represented by mobile phones tends to be saturated, smart cars are regarded as the next most promising smart terminal market. With the acceleration of the trend of "new four modernizations" of automobiles, the industry has set off an unstoppable wave of automotive electronics, becoming one of the hottest markets for semiconductor companies with high-performance computing platforms today.
According to Statista data, the global automotive semiconductor market will be about $46 billion in 2021 and is expected to exceed about $55.5 billion in 2023.
Global Automotive Semiconductor Market Size Forecast 2019-2023;
Semiconductor industry observation mapping
Intel, which once led the PC era, regarded the automotive industry as the next outlet after missing the mobile era; Nvidia, which was pushed to the altar by artificial intelligence, bet on the new spring of autonomous driving; Qualcomm, which enjoyed the mobile dividend, entered the automotive market with great interest to find a new way out...
AMD, which has made many decision-making mistakes, missed the era of mobile chips, and lost the server market, is also seeking a future in the automotive field.
How to prop up AMD's "car dream"
Last year, AMD's Ryzen RDNA 2 processor was used in Tesla's new Model 3 and Model Y's in-vehicle infotainment systems. With the AMD Ryzen Ryzen processor to defeat Intel, it replaced the Intel A3950 processor previously equipped with domestic Tesla, becoming the first batch of chips equipped with the high-performance version of the domestic Tesla Model Y Performance.
The move could be seen as AMD's "first shot" in entering the automotive market. AMD is looking for opportunities to apply GPU technology to the automotive sector, in addition to mergers and acquisitions to enhance the automotive sector or AMD's other way out.
Obviously, the acquisition of Xilinx is AMD's primary plan and purpose to attack the data center market. However, the automotive sector may be another market in which it can focus on development.
So, can Xilinx's FPGA technology provide AMD with enough ammunition to lead the automotive semiconductor market?
As the demand for high-performance SoCs in smart cars continues to grow steadily, Xilinx has also been aggressively expanding into the automotive sector over the past few years. Today, Xilinx is AMD's FPGA division, which continues to focus on Xilinx's current FPGA, adaptive SoC, and software roadmap work.
Xilinx's FPGAs, which have been designed into many slots in ADAS-equipped vehicles to process data from cameras, imaging radar and lidar, are already leading designers and suppliers of advanced vehicle sensor processors
In intel's Mobileye-dominated front camera market, Xilinx is second. In the assisted driving stage, Mobileye and Xilinx occupy the leading position in the industry, and after the rise of L1-L2 level autonomous driving technology, the autopilot chip market has long been controlled by Mobileye and Xilinx two players. Xilinx is also a major supplier of data processing chips for surround view cameras, already in production in Bosch and Magna products, and some other large suppliers are also in the verification and production phases.
In addition, Xilinx dominates the processors of emerging 4D imaging radars, accounting for about 85%-90% of the emerging market share. In the field of lidar, there are also many industry customers who design and use Xilinx-based chips.
In the field of autonomous driving, highly automated driving and fully autonomous driving are the general trend in the future travel. Xilinx Automotive (XA) platforms play a key role in powering autonomous driving modules for high-speed data aggregation, pre-processing and distribution (DAPD) and computational acceleration. The adaptive XA SoC platform not only optimizes the processing of a growing number of complex, security-critical applications, but also meets the computational latency, performance, power efficiency, and functional safety requirements of OEMs and Tier 1 vendors between sensors and domain controllers.
At the same time, the Xilinx Automotive (XA) platform transforms the in-vehicle experience for drivers and passengers across a range of systems, including infotainment, driver monitoring systems (DMS) and in-vehicle surveillance systems (ICMS). Provide safety and reliability for automotive applications through the flexibility of platforms such as the XA Zynq-7000 and Zynq UltraScale+ MPSoC.
In addition, in the area of vehicle electrification and security gateways, Xilinx's products and platforms build customers with secure, high-bandwidth, customized networks that meet the architectural requirements of their system characteristics, so that automotive systems can meet future needs.
It can be seen that Xilinx products have formed a wide range of application space in the automotive field. Xilinx's automotive business has been growing steadily for nearly 15 years. According to statistics, Xilinx's cumulative shipments in the field of automotive electronics have exceeded 205 million, of which shipments in the field of ADAS have reached 80 million pieces, and in 2020 alone, Xilinx shipped nearly 20 million devices to the automotive industry.
FPGAs, on the other hand, are criticized for being much more expensive than SoCs developed to perform the same task. But that complaint is also fading, with Willard Tu, senior director of AMD's automotive business and Adaptive Computing SoC, arguing that the flexibility of the FPGA platform is compensating for its higher development costs.
The advantage of FPGAs is the flexibility they exhibit as automotive electronics performance changes rapidly. For example, when a 1 megapixel imaging sensor needs to be upgraded to 4 megapixels or even 8 megapixels, instead of being forced to redesign the entire SoC to accommodate these changes, Xilinx has proven that its existing sensors can support 8-megapixel cameras.
Similar logic applies to 4D imaging radar and lidar, the characteristics of FPGAs are to adapt and change, to meet the needs of multi-generation products and technologies, and changing artificial intelligence algorithms also make FPGAs more suitable for use in the automotive field, better to achieve a balance between cost and performance.
Overall, xilinx currently focuses on several major subdivisions such as ADAS, autonomous driving, in-vehicle systems, electrification and networking, providing vehicle-grade product solutions from 28nm, 16nm to 7nm, rich development tools, etc., helping major customers to continuously accelerate the pace from innovation to mass production.
The addition of Xilinx provides AMD with a new beachhead in the automotive sector. With Xilinx's FPGA capabilities, AMD can add new revenue from the automotive space while gaining access to automotive OEM and Tier1 customers that Xilinx has been serving for years.
Xilinx's success in the automotive sector depends on the trust and choice of OEMs and Tier1 suppliers in a rapidly changing market. The industry has seen Xilinx's better performance in the previous waves or two, but with the improvement of the level of autonomous driving and the rapid expansion of the scale of automotive semiconductors, Xilinx's early advantage is fading.
In this regard, the editor-in-chief of Ojo-Yoshida Report believes that Xilinx needs to provide a more "hard" semi-customized solution that can surpass FPGAs to enable customers to switch to FPGAs. To succeed in the automotive market, chipmakers need not only chips, but solutions down from the board level.
Similarly, AMD needs to do more, including a product roadmap that combines its embedded experience with Xilinx FPGAs, as well as a long-term strategic planning and commitment to the automotive sector.
We haven't seen that yet.
The dominance of automotive chips
With Xilinx entering the AMD camp, the underlying competition in the era of smart cars will undoubtedly enter the trend of hegemony, and Intel, Nvidia, and Qualcomm are all menacing.
Intel: The "King" of the ADAS Era
Before acquiring Mobileye, Intel was not successful in the automotive market.
In fact, as early as 2015, Intel acquired altera, an FPGA manufacturer, but this operation is not to break into the automotive market. It wasn't until 2017, when mobileye was acquired for $15.3 billion, that it became a leading supplier of automotive semiconductors.
Intel's interest in Mobileye stems from the Israeli company's huge success in the ADAS market. Mobileye's EyeQ family of SoCs and their software stacks are powerful and extremely energy efficient, surpassing other highly programmable solutions such as GPUs and FPGAs.
For a long time, Mobileye was the absolute king of assisted driving.
Looking back on its history, from 1999 to 2001, Mobileye developed prototype products at an annual iteration rate, and in 2001 proposed a plan to solidify the self-developed algorithm onto the chip to integrate into the car, the later EyeQ series of chips.
From the launch of the EyeQ1 in 2007 to the launch of the EyeQ3 in 2014, Mobileye has been invincible in the field of ADAS...
According to data released by Intel, by 2021, Mobileye has more than 30 automakers around the world with orders for 41 models of the new ADAS project. So far, EyeQ's self-driving chips have carried more than 50 million smart cars worldwide, and the shipment of EyeQ system integration chips has exceeded 100 million pieces.
Mobileye EyeQ part of the chip
Earlier this year, Mobileye "sacrificed" a new product line, the most advanced and powerful to date for autonomous driving, the EyeQ Ultra, which has achieved energy efficiency optimization under the computing power of 176TOPS, is a streamlined self-driving car chip, and has established strategic cooperation with Volkswagen, Ford and China's Extreme Krypton Motors.
Also introduced with The EyeQ Ultra are two new EyeQ system integration chips for ADAS solutions, the EyeQ 6L and EyeQ 6H, based on a 7nm process and featuring cost-effective pure camera solutions.
It can be seen that in addition to mobileye continuing to consolidate the traditional market, another goal is high-end intelligent driving.
In 2004, the EyeQ chip came out and changed the ADAS market in one fell swoop. So far, in terms of technology and products, Mobileye has formed a systematic combination. However, with the rapid evolution of autonomous driving technology, whether it is the release of new products, the cooperation and upgrading with many mainstream manufacturers, or the new abacus, it is difficult to cover up Mobileye's frustration in the market, and the "singing" sound continues.
At present, many car companies have begun to give up the use of Moblieye's chips, Weilai, Xiaopeng, Weima and other new car-making forces of the next generation of flagship models are using NVIDIA Orin chips, ideal car in 2021 using horizon journey 3 chips, even the most loyal customer BMW also announced that it broke up with Mobileye and switched to Qualcomm. As one of Mobileye's most solid partners, BMW's "defection" is considered the end of Mobileye's "ADAS era" and the beginning of a new chip war.
Although Mobileye dominates the assisted driving stage, its business model and product performance iteration capabilities may no longer be strongly competitive in the field of autonomous driving and intelligent cockpit. In the territory of L4-L5 level autonomous driving chips, there are more competitors.
According to current sources, Mobileye is seeking an independent listing. This IPO not only has Mobileye's own business, but also includes Intel's projects related to vehicle automatic driving. In the future, in addition to Mobileye's continued focus on system-level products such as autonomous driving platforms, it may also enter the Robotaxi market.
In order to continue to "hold" the market share, Mobileye is constantly working hard.
Nvidia: "Upstart" in autonomous driving
In the field of autonomous driving chips, another major manufacturer is NVIDIA.
From Parker to Xavier, NVIDIA didn't get a big market in the front-loading market. Nvidia Orin is the first large-scale computing platform to receive large-scale mass production orders, including ideal, Weilai, Xiaopeng, Zhiji, Gaohe, Jidu, Mercedes-Benz, Volvo, Jaguar Land Rover and other brands, will begin to mass production on the car.
In early 2021, NVIDIA released the next-generation chip platform DRIVE Atlan, which will be applied to L4 and L5 level autonomous driving, and will start to provide samples as soon as 2023, and will be equipped with market models around 2025. At the same time, a chip that will be equipped with a mass-produced model in 2025 at the earliest has been deployed, and the final force is as high as 1000TOPS.
From Xavier's 30 TOPS, to Orin's 254 TOPS to Atlan's 1000TOPS, Nvidia has also made great strides in chip computing power.
Some of Nvidia's automotive chips
At the just-concluded GTC 2022, NVIDIA announced that the autonomous driving chip Orin is about to be officially put into production. For many car companies that have long announced that they will be equipped with Orin chips, it can be regarded as a reassuring pill.
At the same time, NVIDIA has launched a new generation of autonomous driving platform DRIVE Hyperion 9 based on Atlan chips, with mass production scheduled for 2026. Hyperion 9 can support L3 autonomous driving and L4 parking in parking lots. It will be more than twice the amount of data processing as the 8th generation platform, and twice the performance of the 8th generation platform.
NVIDIA firmly occupies a monopoly position in the GPU market of automotive master control chips, maintaining a market share of 70% all year round. At present, more than 25 car companies and autonomous driving companies have chosen NVIDIA. These partners will contribute more than $11 billion in revenue to NVIDIA over the next six years.
Qualcomm: Take down the smart cockpit chip "Highland"
As a mobile phone communication chip giant, Qualcomm has successively launched 602A, 820A, 8155 and other car machine chips in the past few years, which not only breaks the cockpit chip market pattern monopolized by traditional automotive chip manufacturers, but also seizes the market heights of the Internet of Vehicles and cockpit chips in one fell swoop.
In early 2021, Qualcomm released the fourth-generation Snapdragon automotive digital cockpit platform, as well as the Snapdragon Ride autonomous driving platform. So far, Qualcomm has extended from the Internet of Vehicles and cockpit chips to ADAS, autonomous driving and other fields, and completed the full deployment of the intelligent car market.
With the advantage of intelligent cockpit, after the launch of the Snapdragon Ride autonomous driving platform, Qualcomm also quickly reached cooperation with many OEMs such as GM and Great Wall, and successfully poached BMW from Mobileye to continuously expand its customer base.
In addition, Qualcomm has been enriching its software ecosystem over the past few years, for example, supporting Valeo's latest generation Park4U (cross-level parking solution), partnering with Seeing Machines to provide embedded DMS solutions, and expanding full-stack solutions through the acquisition of Vineier's Arriver software platform.
According to reports, the Snapdragon Ride platform will not only include hardware, but also provide "security middleware, operating system and drivers" for the chip. In addition, Qualcomm will also provide key parts of autonomous driving, such as positioning, perception and behavior prediction software.
At present, Qualcomm ranks first in the three major areas of telematics, vehicle networking and next-generation intelligent cockpit. Globally, more than 25 car companies have chosen Qualcomm Snapdragon Auto's digital cockpit platform.
Write at the end
It is not difficult to foresee that in the next few years, the competition in the automotive chip track will become increasingly fierce.
Judging from the current market situation, Mobileye and traditional automotive chip manufacturers hope to meet the next-generation system needs of car companies through "upward" upgrades, while Nvidia, Qualcomm, and domestic companies including Horizon and Black Sesame are holding high and using large-scale computing platforms to seize the high-end intelligent driving track.
AMD, as a "late entrant", has made a good start in the automotive market by incorporating Xilinx under its umbrella. But FOR AMD to make a meaningful effort in this area with competitors like Nvidia, Qualcomm, and Mobileye, it can't rely solely on Xilinx's current FPGA portfolio or its customer list.
The Ojo-Yoshida report notes that the strength of FPGAs is important, but it is not enough on its own. AMD must build an automotive semiconductor roadmap that tightly integrates Xilinx's FPGA technology with its embedded expertise. The addition of semi-customized ASICs is critical to Xilinx's prospects and consolidating AMD's position in the automotive market.
The war of automotive chips is a long-term power competition, and the temporary lead of any manufacturer can be understood as a victory in a stage competition, but this competition is far from over.
AMD, which has missed out on key markets many times, is well versed in this.