Author: One.
Editor: Meimei.
The AI PC we see now may just be a marketing gimmick.
As the wave of AI sweeps the world and profoundly changes all walks of life, the combination of AI and hardware, especially consumer electronics, such as mobile phones and PCs, is becoming more and more fascinating. In September 2023, Intel CEO Pat Gelsinger first proposed the concept of AI PCs in Silicon Valley.
Intel believes that the so-called AI PC is a PC that can play with AI functions. AI can fundamentally change and reshape the user experience of PCs and unleash people's productivity and creativity through close collaboration between the cloud and PCs, or by running large language models independently on the computer.
When AI PC is popularized, people can use AI functions more simply and conveniently, everyone can show their talents in painting, writing and ** through AI, and can also improve work efficiency through AI, it seems that AI PC can do anything.
After this concept was proposed, Lenovo, Samsung, Google, Nvidia and many other related companies in the PC industry chain chose to follow up. All of a sudden, some "AI processors" that integrate dedicated NPUs and carry them"AI PC" has become the focus of the marketAI seems to have become a "fashion item" for these manufacturers, and it is embarrassing to release new products without having anything to do with AI.
However, behind this hot speculation on the whole network, we can't help but ask: does AI PC really represent the future of the PC industry? Or is it just a marketing gimmick, a concept created to stimulate sales?
The market is looking forward to the new concept of "AI PC".
If there is any difference between an AI PC and an ordinary PC, then it is equipped with AI-related software and services from the factory, which can realize many AI functions based on local hardware acceleration, which may be its most prominent feature. These AI functions include, but are not limited to, AI voice dialogue, AI enhancement, AI matting, AI painting, and AI article summarization.
After a period of rapid growth during the pandemic dividend period, global PC shipments have declined for several consecutive quarters since the fourth quarter of 2021 and have experienced nearly two years of shipments**. Therefore,For the entire PC industry chain, they need a new growth pointAnd 2023 happens to be the year when the AI wave is sweeping, and it is natural to combine AI with PC. Therefore, the industry is hopeful about the concept of "AI PC", trying to attract consumers' attention through AI-driven features and open up a new market area.
After the emergence of the concept of "AI PC", global PC shipments in the third quarter of 2023 increased by 1071%, down 76%, although it is still the first trend compared with the same period in 2022, the decline has narrowed significantly.
From the consumer's point of view, AI-driven features such as AI voice assistants, intelligent image processing, and automated document analysis brought by AI PCs are indeed novel and interesting enough. And with the deepening of AI large models in various industries,Low latency + privacy protection is becoming more and more importantTherefore, the localized large model seems to be the optimal solution, and the AI PC that can solve this problem also has the potential to stimulate consumer demand.
It can be said thatThe market is also looking forward to the concept of "AI PC".
Is an AI PC really better than an old computer?
Although AI PC may seem novel in terms of functionality, in fact,The application of AI technology on PCs is not new。Long before the concept of AI PCs was proposed, people have been using AI technology on PCs. Many software is already taking advantage of the CPU's specific instruction set or the GPU's computing power to perform AI tasks. For example, AI filters in Photoshop, AI noise reduction in game streaming software, and AI drawing programs, etc.
Older computers like desktops are not blank when it comes to the application of AI technology。Many older computers already have the ability to run AI programs, thanks to the CPUs and GPUs they carry. For example, the DLBow instruction set and the X512 instruction set in the CPU, as well as the Tensor Core cores of NVIDIA graphics cards and the GPGPU general computing power of AMD graphics cards, are all key technologies for old computers to play a role in the field of AI.
And from a performance point of view,The performance of old computers in terms of AI computing power can often outperform that of newly launched AI PCs。Even the latest generation of AI PCs has a lot to offer with their integrated NPU computing power, which is hard to compete with high-end graphics cards. For example, the RTX 2060 mobile graphics card has reached 36 AI unit computing power9TOPS, and the AI computing power of the RTX 4060TI 16GB desktop version graphics card is as high as 353TOPS. This performance gap is especially noticeable when dealing with complex AI tasks, where older PCs may excel.
WhileThe advantage of today's AI PCs over these older PCs may lie in their all-in-one design and optimization。They usually come pre-installed with software optimized for AI functions, allowing users to enjoy the convenience of AI without additional configuration. This kind of out-of-the-box convenience is difficult to match with old computers, so in the field of thin and light laptops, AI PCs may have advantages in mobile office and daily use.
But,Do users really need to use AI on thin and light laptops?
AI PCs with somewhat awkward market positioning
Although the concept of AI PC is currently hot, it obviously still faces some limitationsThe market positioning is a bit awkward
As mentioned above, although the AI performance of AI PCs has improved in theory, in practical applications, its performance may not be as expected. Since there is still a large gap between the computing power of AI PCs and high-end discrete graphics cards, this is especially obvious when dealing with high-load AI tasks. For example, for applications that require a lot of parallel processing power, such as deep learning training or complex image recognitionAI PCs may not be able to offer performance comparable to professional-grade hardware
So in this case,The energy efficiency benefits of AI PCs may be more pronounced on thin and light laptops. But in reality, in the broader PC market, this advantage may not make it stand out as the first choice for consumers. In terms of user habits and market acceptance, many AI user groups have become accustomed to the existing hardware and software environment, and when dealing with AI-related tasks, many users may prefer to choose a more powerful desktop or gaming notebook, which has better performance in terms of heat dissipation and performance, while the thin and light design of AI PCs may not be able to meet these needs well when AI computing.
Therefore,There is also a certain ambiguity in the market positioning of AI PCs。They try to find a niche between traditional PCs and professional AI workstations, but this positioning can lead to a fragmentation of the target user base. For professional users who pursue extreme performance, AI PCs may not provide sufficient hardware support; For the average consumer, the additional cost of AI PCs may not be offset by the convenience it brings.
In addition, the software ecosystem of AI PCs is also a challenge. AI apps are pre-installed, but their usefulness, compatibility, and update frequency take time to verify. It may take a while for users to get used to the new AI features, and it will take time for developers to optimize their software for these new hardware.
In general,The current AI PC may still be a pseudo-conceptBecause it doesn't use AI to achieve new features that current desktops and traditional PCs can't do, in other words, today's AI PCs are just an "old computer" that has been repackaged, not something new.
Looking to the future, AI PCs still have a long way to go to democratize AI technology. They need continuous optimization and innovation in terms of performance, market positioning, software ecology, etc., in order to truly overcome these limitations and achieve their breakthrough in the PC market.