news

한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina

AI PC

Since its invention, PC (personal computer) has always been closely related to "productivity". It is precisely because of its attributes as a productivity tool that it can be accepted by consumers and has also become one of the few "must-have technology hardware".

However, in the past decade of development, the PC market has shown greater fluctuations.After 2008, affected by the impact of smartphones and the economic environment, global PC shipments began to slow down. By 2013, annual shipments had dropped from a peak of 350 million units to 310 million units.

Since then, due to market saturation and the popularity of mobile devices, the PC market has continued to be in a "cooling period". However, after 2019, due to the sharp increase in demand for remote work and online learning, the market reached a peak again in the short term. However, once the short-term demand boom passed, the market fell into silence again. In 2023, shipments fell to 247 million units, setting a "new historical low".

at the same time,The product categories in the PC market are also undergoing tremendous changes. Two typical examples are gaming laptops and AI PCs.The former has completed a rapid transformation from a "new species" to a "sales leader" in the past five years, and AI PC is considered to be the biggest driving force for the PC market in the next five years. According to Canalys' forecast, China's AI PC market will enter a period of explosive growth in 2024 and will reach 33 million units by 2028, accounting for 73% of the overall PC market shipments.

It has only been one year since the concept was proposed. Can AI PC really “turn the tide and save the collapsing building”?

AI+PC has never been a "new species"

Regarding the starting point of the development of AI PC, most people in the industry believe that it can be traced back to September 2023, when Intel CEO Pat Gelsinger first proposed "AI PC" in Silicon Valley. At the same time, with the launch of the Core Ultra series processors at the end of the year, related products were put into practice.

However, after an in-depth study of the relevant technologies, Titanium Media APP found that if we trace back the timeline, Nvidia actually launched RTX technology and the first consumer GPU chip (GeForce RTX) designed specifically for AI in 2018. According to Nvidia's technical engineers, "In Nvidia's definition, AI PCs are computers equipped with dedicated AI acceleration hardware, and on RTX GPUs, these dedicated AI accelerators are called Tensor Cores."

AI applications supported by Nvidia GPU chips

To put it more simply, the Tensor Core added by NVIDIA to the GPU chip is like a supercomputing "accelerator" that is specifically designed to process and accelerate certain types of mathematical calculations, especially computing tasks in deep learning.

Imagine you are cooking in the kitchen and have an ordinary knife to cut vegetables, but if you have a special vegetable cutter, the speed and efficiency of cutting vegetables will be greatly improved. Similarly, ordinary processors (CPUs) and graphics processors (GPUs) can handle a variety of computing tasks, but Tensor Core is like that special vegetable cutter, which is designed to handle specific "vegetables" (that is, mathematical operations such as matrix multiplication and convolution).

These operations are common in deep learning, such as a large number of matrix multiplications required when training neural networks, and Tensor Core can complete these tasks faster than traditional computing units. As a result, the use of Tensor Core can greatly accelerate the training and reasoning process of AI models, allowing these complex calculations to be completed in a shorter time.

The emergence of Tensor Core is actually to accelerate AI performance and introduce new AI functions that could only be run in the cloud before to PC users. For developers or deep AI users, NVIDIA has also launched the TensorRT developer kit to accelerate deep learning reasoning performance.

“Image-to-image” application completed by generative AI

According to NVIDIA technical engineers, TensorRT can accelerate popular generative AI models, including Stable Diffusion 1.5 and SDXL, and the new UL Procyon AI image generation benchmark also supports TensorRT acceleration. So when generative AI applications just emerged, NVIDIA professional graphics cards were hard to find. When GPU chips were in short supply, many consumer graphics cards such as RTX 4090 and RTX 4080 were also purchased in large quantities and applied to AI computing.

Nowadays, there are also a large number of AI applications for individual users that can be completed through NVIDIA RTX GPUs, such as text-to-image, image-to-image applications, and "intelligent voice assistants" that rely on learning from local models.Therefore, it can be said that from the application level and even the hardware technology level, "AI PC" itself is not a "new species".

Chip manufacturers take up the banner of "AI PC"

The reason why the concept of "AI PC" was widely recognized in the PC field after Intel proposed it was, on the one hand, because Nvidia's previous large number of related AI applications were more centered around the enterprise side, that is, applying professional graphics cards (or workstation graphics cards) to deep learning frameworks such as TensorFlow, PyTorch, and Caffe to conduct model training on large-scale data sets.

For example, in the training of large-scale neural networks (such as the GPT model, BERT model, etc.), research institutions and companies will use high-performance workstations or clusters such as NVIDIA DGX Station A100 to improve work efficiency.

Compared with these large-scale applications, the edge AI processing needs around personal computers (PCs) can be seen as a "by-product" brought by Tensor Core.Although related applications can be seen in games, image generation, language processing and other applications, compared with the cheap, easy-to-use and lower-threshold cloud computing power, purchasing a graphics card separately for AI applications is too luxurious for ordinary users.

On the other hand, compared with CPU, GPU is not in the "just-needed" ecological niche. According to data released by Jon Peddie Research, as of the first quarter of 2024, the penetration rate of discrete graphics cards in the PC market is less than 20%, because relying on the integrated graphics cards built into some CPUs can still meet many office, daily entertainment and basic productivity needs.

Therefore, even though NVIDIA has already started to lay out related plans around AI applications earlier and has a first-mover advantage in terms of ecological matching,But the limelight of "AI PC" was seized by Intel.

It is worth noting that when Intel proposed the concept of "AI PC", it also aimed at individual consumers. Gao Yu, general manager of technology for Intel China, said in an interview with Titanium Media APP: "AI PC will usher in a new era of PC. Since the concept of AI PC was proposed, Intel has also been working with terminal manufacturers to help AI PC continue to penetrate consumers' minds by promoting product innovation, building an AI ecosystem, and implementing AI applications."

AI applications for individuals

Another point that needs to be explained is that Intel has a relatively clear definition of the AI ​​PC category. Gao Yu mentioned: "Intel's definition of AI PC specifically refers to a thin and light notebook with CPU+GPU+NPU. The scenario application it emphasizes is actually to use AI to work closely with the cloud and PC, or to run large language models independently on the computer side to provide users with rich AI application scenarios."

In other words, the "AI PC" proposed by Intel itself excludes high-performance desktop computers and workstation products that individuals may use. This also means that the "AI PC" defined by Intel does not completely overlap with application scenarios that rely heavily on computing power. It emphasizes more on how AI can provide services to individuals and improve work efficiency, rather than creating large-scale applications like ChatGPT.

In the view of Titanium Media APP,Whether it is a notebook product equipped with Ultra series processors jointly promoted by Intel, Lenovo, Asus and other OEM manufacturers, or a personal computer with AI computing power using Qualcomm and Nvidia chips, they can all be counted as "AI PCs".。

The definition of "AI PC" should come from the experience it can provide, rather than which brand of chip or manufacturer's software it uses.

As Fan Zijun, vice president of HP China and general manager of the China Consumer Products Division, said, "The definition of AI PC may be different for each brand and each person. For example, if AI software is installed on a computer, does it become an AI PC? Or if the product has an independent button to activate the AI ​​function, does it mean it is an AI PC? I think neither is true."

The core of AI PC is actually to use AI functions to improve efficiency and create value.On this basis, another major feature of AI PC is that it emphasizes the driving force for efficiency upgrades, which comes from the support of local computing power and local deployment capabilities.

How will AI change the PC?

After understanding the origin and definition of AI PC, from the current perspective, what impact will the addition of AI have on the PC field?Titanium Media APP analysis believes that it will mainly focus on two aspects: promoting changes in the market and competition landscape and updating the definition of product functions.

From a market perspective, "AI PC" seems to have revitalized the PC market, which has been dormant for many years. According to data released by Canalys, in the second quarter of 2024, global PCs continued to grow, with shipments of desktops and notebooks reaching 62.8 million units, a year-on-year increase of 3.4%.

Among them, the shipment volume of AI PCs was 8.8 million units, accounting for 14% of the total PC shipments this quarter. However, it should be noted that the AI ​​PC devices in Canalys' statistics include desktops and notebooks. The basis for defining the category is whether it is equipped with a chipset or module for dedicated AI workloads, such as NPU. Therefore, the sales volume of this part of AI PCs will not only include notebook products equipped with Intel chips, but also PC products using Nvidia graphics cards.

Industry insiders have previously analyzed that: "With the transition to Windows 11 and the adoption of AI PCs, the product update cycle will accelerate in the next four quarters. The integration of AI functions not only improves the performance of the device, but also brings new application scenarios and user experience, especially in improving productivity and entertainment experience. These features make AI PC an important driving force for future market growth."

Among many OEM manufacturers, Lenovo is the first to focus on "AI". After Intel launched the Core Ultra series processors in early December last year, Lenovo released two AI Ready AI PC products, the ThinkPad X1 Carbon AI and the Xiaoxin Pro 16 AI Core Edition, on December 15.

Earlier in October, at the 9th Lenovo Innovation Technology Conference (2023 Lenovo Tech World), Lenovo also took the lead in showcasing AI PC products. In addition, Lenovo also released the industry's first "AI PC Industry (China) White Paper" with IDC.

As a PC brand with a single-quarter shipment volume of 13.7 million units, HP is also one of the earliest manufacturers to enter the AI ​​PC market. So far, it has not only launched AI PC terminals including the StarBook Pro 16 and StarBook Pro 14, but also released the Hui Xiaowei Smart Assistant 4.0 with integrated one-stop AI functions.

On May 10 this year, Thunder Technology released its first extended-range AI PC, Thunder aibook15, which is equipped with Intel Core Ultra 7 processor and supports CPU+GPU+NPU to provide AI hybrid computing power. In addition, OEM manufacturers including MSI, ASUS, Honor, etc. have also completed the evolution to "AI PC" with the help of Intel CPU iteration.

Regarding the achievements of AI PC in the current market, Titanium Media APP believes that:This is more because the iteration of AI PC depends on the iteration of hardware centered on chips.In other words, when many consumers purchase products, they are not faced with the difficult choice between "AI PC" and "non-AI PC". Instead, as long as users buy PC products that use the latest chips, they must be "AI PCs".

Wang Rui, chairman of Intel China, said that as of June this year, 8 million devices have been equipped with Intel Core Ultra processors since the release of the processor. From a statistical perspective, this is 8 million AI PCs, regardless of whether AI applications are running on them or whether users know that the product is equipped with local AI computing power, it is the sales volume of "AI PC".

Regarding how AI functions can boost PC product sales, Titanium Media APP believes that Fan Zijun's viewpoint is more reliable: "As people around users use AI PCs to achieve higher and higher efficiency and better experience, more people will definitely learn about and purchase AI PCs, and new demands will emerge."

After all, compared with traditional PCs, AI PCs have no difference in appearance, design, or even system UI. Therefore, it is impossible to let users understand the product innovations through superficial market education like gaming laptops and handheld game consoles.At this stage, with the addition of AI functions and local computing power, a "killer application" is urgently needed to break through the circle.Only by allowing consumers to perceive the value brought by AI can they ultimately promote purchases.

It can be seen that when will there be killer applications suitable for AI PCs has become the core. In the past, popular AI applications such as ChatGPT, Stable Diffusion, and Pika have either utilized larger cloud computing support or required independent GPUs with high computing power. In the view of Titanium Media APP, the main advantage of AI PCs at this stage will be reflected in the deployment of these high-computing power applications on mobile terminals.

Gao Yu also mentioned this point: "In AI PC, the cloud represents the upper limit of computing power, and the end-side represents the lower limit of computing power. Cloud AI and end-side AI jointly present the complete AI application experience to PC users. The two coexist on the platform. At the AI ​​PC level, in order to promote the implementation of AI applications, not only powerful hardware but also optimized software is needed."

An industry analyst told Titanium Media APP: "At present, the speed of AI PC popularization is largely affected by the price of the equipment and the education level of users. In addition, the experience difference brought by the products is not obvious enough. Therefore, this type of equipment is mainly aimed at the high-end market and professional users at this stage, and the acceptance of the mass market is still improving."

In terms of market influence, AI’s actual empowerment of PC sales is very limited at this stage, and more users only pay for PCs rather than AI. If we extend the time dimension, as more AI applications emerge, especially those that can rely on local AI computing power,Only when some users begin to perceive the improvement in life and work efficiency brought about by AI will people want to understand and purchase "AI PCs."

When talking about the future competition direction of AI PC, Fan Zijun believes that the competition at the hardware level will still be the core, such as CPU performance, heat dissipation capacity, screen and keyboard, and body design, etc. However, the addition of AI has indeed brought some changes to the product's competitive thinking.

"Based on hardware, I believe that in the AI ​​PC era, security, reliability and ease of use will become new dimensions of consideration for products. Security is mainly about privacy, especially in the AI ​​era where the end and the cloud are closely integrated. Reliability is about ensuring that the product can run stably under various working conditions as a productivity tool. Ease of use refers to how to fully tap the capabilities of AI to achieve a balance between low thresholds, practicality and high efficiency." said Fan Zijun.

Triggering a change in the landscape, the addition of AI triggers a "chip war"

If it is too early to talk about the innovation brought by AI PC and whether it will become a key factor affecting consumers' purchase of products,With the emergence of "AI PC", changes in the industry's competitive landscape have already arrived.

Taking the 2024 Taipei International Computer Show as an example, we can see traditional chip brands Intel, AMD, and Nvidia competing around AI, and we can also see "PC new forces" such as Qualcomm competing to deploy AI PCs, and many terminal manufacturers are also simultaneously focusing on AI applications.

In fact, Qualcomm's layout for the PC platform began as early as 2016. From the Snapdragon 835 on PC to the launch of the Snapdragon 8cx computing platform for PC to the current Snapdragon X computing platform and two processors, Qualcomm has been in the running-in period with "PC" for 8 years. The introduction of the AI ​​PC concept has also allowed Qualcomm to see the opportunity to reshuffle the industry and take the opportunity to enter the market.

Qualcomm and Microsoft team up to build AI PC

Entering 2024, Qualcomm's layout in the PC field is also accelerating, and one of the key events is the cooperation with Microsoft. In the early morning of May 21, Microsoft launched the new Surface Pro and Surface Laptop, which can be seen as the beginning of Microsoft's hardware in the AI ​​era. The new machine is equipped with a Qualcomm processor. At the same time, in order to distinguish it from the AI ​​PC concept proposed by Intel, Microsoft also gave the new product the name "Copilot+PC".

Previously, when Titanium Media APP was communicating with the technical director of Qualcomm, the other party mentioned that related suites such as Microsoft Office 365 have been ported to Windows PCs equipped with Snapdragon in the form of native applications. Others including Adobe's entire family of applications have also been adapted. The browser also provides support for Edge and Chrome. However, there is currently relatively little support for games, and the current adaptation is mainly focused on casual games.

Qualcomm President and CEO Amon said: "The PC is being reshaped. The entire system of Copilot+ PC equipped with Snapdragon X Elite is integrated with AI. Windows Copilot+ PC powered by Snapdragon chips is jointly defining the personal computing experience and creating applications for the next generation of PC products."

Not only Qualcomm, Apple will also be the "catfish" in the AI ​​PC eraWith its self-developed architecture, the latest process technology and Mac OS, Apple's M series chips and MacBook products have achieved success in a short period of time. But what many people overlook is that Apple is also one of the first manufacturers to introduce AI acceleration modules in chips.

Apple first added a dedicated AI acceleration module, the Neural Engine, to its mobile phone chips in the A11 Bionic chip released in 2017. The A11 Bionic chip is equipped with Apple's first neural engine, which is specifically designed to handle machine learning tasks such as facial recognition (Face ID) and augmented reality (AR) applications.

Apple's M series chips also integrate a dedicated Neural Engine, which is specifically designed to handle AI and machine learning tasks, such as image recognition, natural language processing, etc. Therefore, for Apple, a certain hardware foundation has been laid for the deployment of AI applications.

Titanium Media APP believes that after entering the AI ​​PC market, all chip manufacturers have once again been brought to the same starting line.Judging from the current competition, NVIDIA does have certain advantages in the competition for AI PCs, thanks to its unique architecture, relatively complete ecosystem, and applications from the enterprise side, while Intel follows closely behind with market awareness, product first-mover advantage, and leading product volume.

Other manufacturers, whether AMD, Qualcomm, Apple or even other entrants, have maintained a consistent pace and equal development opportunities. In the AI ​​PC era, it has become more difficult for Intel to continue to maintain its dominance in the CPU field.

In general, AI PC is still in its early stages of development. How to complete market education faster and more efficiently, and at the same time find the real "killer feature" of AI PC compared to traditional PC is a common problem facing terminal manufacturers, chip manufacturers, and software manufacturers.

In the long run, AI PC can better complement cloud computing power by relying on the advantages of local computing power. The former is responsible for providing AI applications with strong privacy, high computing efficiency, and good experience consistency, while the latter will play a role in application scenarios with higher versatility, scalability, and computing power requirements. Only then can AI PC get rid of the label of "gimmick" and truly get on the fast track of AI.