Intel Inside, now involved in smart cars

The most popular one at the moment is the domestic 3A masterpiece "Black Myth: Wukong". Just after it was launched, the number of people online on the Steam platform reached 1.045 million within an hour.

However, players have discovered that the most difficult part of the journey to the West is configuration.

Ordinary PCs simply cannot support such a huge gaming system.

But what is exciting is that the future smart cockpit can. Lynk & Co Z10 has made it a reality. On a test bench, "Black Myth: Wukong" can run smoothly at 2K resolution and full HD quality with a display quality close to 60 frames.

This means that the smart cockpit has become a mobile Internet cafe, and the core driving this change is the cockpit chip.

On August 8, Intel, which announced its entry into the automotive market, threw a high-performance beast into the field of cockpit chips.

A discrete graphics card (ARC A760-A) with a computing power of up to 229TOPS and 16G DDR6 video memory, supporting AAA games and 4K display. It will be commercially available in 2025.

Intel will have to be ruthless in this "Second Battle" in the automotive business.

On the one hand, after missing out on the mobile phone market dividend and suffering a reduction in PC market share, the automotive market is a battle it cannot lose;

On the other hand, competition in the cockpit chip market is fierce. While Qualcomm has maintained its dominant position, powerful players such as Nvidia, AMD, and MediaTek are also eyeing it.

So when the wave of AI+cars hit, Intel, like other chip companies, saw new opportunities and followed two methodologies.

• 
  

• Going with the trend, we leverage the technology trend of generative AI to realize AI-defined smart cockpits, giving the large end-side model a huge scalability space;

• 
  

• The chip computing power has been expanded to 229TOPS, which is more than 7 times that of Qualcomm's flagship product 8295 chip.

To some extent, Intel has chosen a path of "great force brings miracles".

It is betting that the cockpit chip market's demand for scalable AI power consumption performance will make computing power the key to winning.

01. Seize the new trend of AI-defined cockpit

After acquiring Mobieye and entering the intelligent driving business, Intel did not forget another piece of the market.

So, when the new trend of AI+smart cockpit came, it decided to take action personally, take advantage of the opportunity, and fight a prepared battle.

This involves two aspects.

One is timing.

Intel once tried to enter the cockpit chip market with its leading 14nm process chips, but the mobile operating system at the time was not compatible with the X86 structure, so it needed to wait for a wave of smart cars.

Jack Weast, general manager of Intel's automotive business unit, said that the trend of smart cars is consistent with that of smart phones.

Many people believe that the turning point of smartphones was the release of the iPhone in 2007, but the real turning point was when smartphone sales exceeded feature phones in 2013 with the popularization of 4G. Based on this calculation, Intel believes that the real turning point of smart cars will be 2024.

In addition, intelligence will determine the second half, and cockpits are actively embracing AI.

Jack 表示，「2025 年，中国将有 80% 的智能座舱渗透率，AI 将成为智能座舱最显著的特征。」

The cockpit chip launched by Intel will also begin commercial use in 2025.

The second is preparation.

Intel believes that the AI ​​cockpit should have three capabilities:

• 
  

• Display: Supports 4K high refresh, 3A game scenes, covers full 3D, UI and other functions, and enhances user immersive experience

• 
  

• Perception: Capture the panoramic view outside the cabin and strictly protect the privacy inside the cabin

• 
  

• Thinking: Multimodal interaction, local large models and full-cabin AI perception give the smart cockpit the ability to think

Intel has configured 28 Xe cores, 28 ray tracing units, and up to 448 AI-oriented XMX/vector engines on this ARC A760-A chip, while inheriting 16GB of video memory and supporting eight independent 4K displays running simultaneously. These parameters are completely sufficient for the smart cockpit.

What's more, its AI computing power is as high as 229TOPS.

If we replace these hard terms with soft expressions, we can achieve the following on this chip:

• 
  

• Move the Internet cafe into the car and play the AAA game "Black Myth: Wukong" in the car;

• 
  

• Move the computer into the car and use AI tools to easily complete a series of office tasks such as meeting summaries;

• 
  

• Bring leisure space into the car and complete personalized interactive actions through multi-modal commands such as voice, gestures, movements, and emotions, including personalized news recommendations and summaries, music recommendations, health monitoring and other services;

……

It is worth mentioning that these functions can be run simultaneously without worrying about the multiplier effect of being too large.

At the chip launch conference, ThunderSoft co-founder and executive president Geng Zengqiang introduced the Dishui OS operating system and demonstrated the 3D HMI human-computer interaction interface, which covers the main application scenarios from car starting, 3D mapping to parking, etc.

Whether it is the scene of killing enemies outside the car, the reflection of the ground on a rainy day or the complete presentation of the car interior, diversified visual presentation can be achieved in the interaction.

It is reported that this large interactive scene demonstration rendering contains 6.5 million triangles, and based on Arc A760-A, it can achieve 60fps full frame operation for real-time interaction. However, this only uses 30% of the chip's computing power, and the remaining 70% can fully meet the needs of AI large models, so this chip can achieve simultaneous interpretation of 3D graphics scenes + AI large models.

In addition, while Intel extends the performance upper limit, it also ensures the lower limit.

As vehicle-side computing power is improved, large models on the end are installed in the vehicle. This means that vehicle-side functions do not need to rely on the network, and sensor data does not need to be uploaded, protecting user driving privacy. The vehicle no longer needs to rely on the cloud for direct and timely response, reducing safety risks. In addition, this can save a lot of traffic and cloud token fees, forming a more economical cost solution.

It is undeniable that Intel is definitely a powerful chip player, and it may be able to catch up by seizing the opportunity of generative AI.

02. AI computing power crushes Qualcomm and is better than MediaTek

In the past two years, Qualcomm has been the protagonist in the cockpit chip field.

Since cockpit intelligence places high demands on advanced chip processes (below 7nm), traditional automotive semiconductor manufacturers such as NXP and Texas Instruments were unable to meet them at the time, which gave mobile terminal chip companies huge opportunities.

As a result, Qualcomm, which emerged from the mobile field, quickly opened up the market and gained a firm foothold in the automotive cockpit field with its cockpit chip products such as Snapdragon 820A and Snapdragon 8155P.

According to the installed capacity ranking of smart cockpit suppliers from January to June 2024 released by Gasgoo Automotive Research Institute, Qualcomm leads the cockpit domain control chip market with an installed capacity of more than 1.55 million chips, accounting for 66.7% of the market share, and has established an unshakable industry leadership position with absolute advantage.

Qualcomm is striving to build a market perception that "Snapdragon is the best choice for automotive chips" and is penetrating from the B-end to the C-end. This leads consumers to consider Snapdragon chips as an important factor when evaluating whether the price of a car model matches its value.

Therefore, the fourth-generation cockpit solution launched by Qualcomm, which uses the Snapdragon 8295 chip with a 5nm process, has basically become the standard configuration for new mid-to-high-end cars, and has been mass-produced on the Jiyue 07, Ideal L6, NIO ET5, and Xiaomi SU7.

This situation where commercial value exceeds commodity value can easily lead to a high demand for goods, which, to put it bluntly, means they are expensive.

It is reported that Qualcomm can sell a Snapdragon 8295 chip for around US$300, plus licensing fees, service fees and other miscellaneous fees.

In fact, the industry does not want a monopoly, as this will not only lead to the loss of bargaining power, but also easily lead to operational risks of lacking a Plan B.

As a result, downstream players choose to rely on Qualcomm while also being eager to find a second option.

There are many disruptors in the market, and MediaTek is one of them.

As players that also emerged from the mobile terminal field, the war between Snapdragon and Dimensity has spread to smart cockpits.

The key to MediaTek's strategy lies in two points:

• 
  

• First, the price has been reduced, with chips generally being $30 cheaper than Qualcomm’s;

• 
  

• The second is performance. It is the first to apply the 3nm process to automotive-grade cockpit chips.

In April this year, MediaTek released three chips in the Dimensity series, including CT-X1 (3nm process), CT-Y1 and CT-Y0 (4nm process). Among them, the 3nm process automotive-grade chip is the world's first.

The reason why MediaTek controlled the process to 3nm is that it is consistent with Intel's thinking and believes that the way out lies in AI. Therefore, it chose to cooperate with NVIDIA and integrate NVIDIA's AI and graphics computing IP in the form of Chiplet.