The personal computer (PC) has undergone nearly 40 years of development. From its inception in the 1970s to the PC boom of the 1980s and 1990s, and into the relatively stable maturity period in the new century, the PC industry has been a significant driver of technological advancement.
However, things have changed since 2020. According to research data, global PC shipments in 2021 totaled 345 million units, a 5% decrease from 2020. This figure further declined to 332 million units in 2022, marking two consecutive years of negative growth. This trend indicates that the personal computer industry is currently experiencing a relatively sluggish phase.
At this critical juncture, the rise of artificial intelligence (AI) technology has brought new opportunities for the future development of the PC industry. Traditional PC processor giants such as Intel and AMD have accelerated the release of new generations of PC products equipped with AI acceleration chips.
They hope to inject new vitality into the PC market and seize market opportunities with the support of AI technology. Meanwhile, mobile chip manufacturers like Qualcomm have also entered the PC market this year with products like the Snapdragon X Elite processor, actively competing with established brands in the personal computer arena.
Faced with this emerging AI PC market, different manufacturers have put forward their own definitions and interpretations. For ordinary users, what exactly constitutes a "true" AI PC? What revolutionary impacts can these new computing devices bring to our daily lives and work? Let's delve into the world of AI PCs and explore what they're all about.
As a dominant player in the personal computer industry, Intel naturally has significant influence over the definition of AI PC. Intel has three core requirements for AI PCs.
Equipped with dedicated Neural Processing Units (NPUs), Central Processing Units (CPUs), and Graphics Processing Units (GPUs), this hardware combination provides the necessary computational resources for AI applications.
Support for Microsoft's Copilot feature, with a dedicated Copilot physical key on the keyboard.
Possess AI-specific acceleration capabilities, meaning that each component, including the CPU, GPU, and NPU, is optimized for AI tasks to enhance efficiency and performance.
The key here lies in the integration of Intel's Neural Processing Unit (NPU). This hardware accelerator, tailored by Intel specifically for AI and machine learning scenarios, significantly enhances the computing power of PCs in tasks such as voice interaction and image processing, providing fast response times locally. It's also more energy-efficient compared to cloud computing services, thus contributing to prolonged device battery life.
For more complex AI tasks, a combination of GPU and CPU may be necessary as the NPU might not suffice to handle these high-intensity workloads. In certain cases, the CPU, NPU, and GPU can collaborate to execute large-scale language processing models.
AI models also place high demands on memory capacity and speed. Larger memory capacity allows for more complex and precise models, while faster memory speed enhances overall performance. Although Microsoft has not yet set specific minimum memory requirements, Intel notes that certain workloads may necessitate at least 16GB or even 32GB of memory.
Intel is leading the charge in bringing the AI PC concept to laptops, unveiling a new generation of Core Ultra processors equipped with NPUs. The company aims to deliver over 100 million PCs with AI accelerators by the end of 2025. Reports indicate that Intel has partnered with over 100 AI Independent Software Vendors (ISVs) and is set to roll out more than 300 AI-accelerated applications by the end of 2024.
AMD's definition of an AI PC lies in its integration and application of AI technology, aimed at transforming personal computers into the smartest and most personalized devices possible. According to AMD, an AI PC should possess the following characteristics:
Embedded with a natural interactive personal intelligent agent based on large-scale models, allowing the device to understand and respond to user needs, providing a more personalized user experience.
Embedded with a personal knowledge base, enabling the AI PC to store and manage user data and information for delivering more accurate services.
Equipped with local heterogeneous computing power consisting of CPU+GPU+NPU. This mixed architecture fully utilizes different types of processing cores to achieve efficient AI computing.
Connected to an open AI application ecosystem, meaning the AI PC should seamlessly integrate with various AI applications and services, forming an interconnected environment.
Ensuring the protection of personal privacy and data security, ensuring that users enjoy the convenience brought by AI while their data and privacy are properly protected.
In terms of product release timeline, AMD is slightly later than Intel. Apart from targeting laptops, the Ryzen 8000 series processors integrated with NPU mark the first integration of NPU chips into desktop processors. Its computational core integration capability reaches 39TOPS.
AMD also offers three computing engines to meet the differentiated needs of AI PCs, namely Zen4 architecture CPU, RDNA 3 architecture GPU, and XDNA architecture NPU. Zen4 architecture CPUs are mainly used for general processing and AI inference, RDNA 3 architecture GPUs are suitable for gaming and AI-based content creation, while XDNA focuses on low-power AI applications with adaptive data flow architecture, reducing external memory access to improve performance and efficiency.
Qualcomm officially enters the PC market with its Snapdragon X Elite processor. This processor is equipped with Qualcomm's independently developed AI acceleration engine, providing excellent machine learning computing performance for laptops.
Summarizing Qualcomm's publicly available information, Qualcomm believes that an AI PC should have the following key characteristics:
Equipped with a heterogeneous computing architecture optimized specifically for AI tasks, including CPU, GPU, and AI accelerators working together.
Supporting industry-leading AI frameworks and models, capable of smoothly running various intelligent applications.
Demonstrating outstanding energy efficiency performance, maintaining long battery life even in intensive AI computing scenarios.
Providing instant responsive user interaction experiences, such as natural language interaction and intelligent assistance.
While laptops equipped with the Qualcomm Snapdragon X Elite processor aren't yet on the market, official data from Qualcomm indicates impressive performance. In various benchmark tests, it outperforms products from Intel and AMD, boasting superior power efficiency, particularly in AI performance, purportedly providing up to 75 TOPS of AI computing power. This capability allows the Snapdragon X Elite processor to rapidly process AI models locally, such as Stable Diffusion, which generates images from text prompts. Additionally, the NPU of the Snapdragon X Elite processor can handle all AI tasks independently or collaborate with the CPU or GPU when needed.
Different manufacturers have varying definitions of the AI PC concept, but they've reached consensus on several key points:
Firstly, AI PCs must have robust AI processing capabilities. This entails dedicated Neural Processing Units (NPUs) and the ability to collaborate with CPUs and GPUs to support various intelligent applications.
Secondly, AI PCs should be capable of running complex AI models locally without relying entirely on cloud computing resources, thus providing real-time responsiveness and smooth interaction.
Moreover, AI PCs require ample memory and storage space to accommodate and run increasingly large AI models. A robust software ecosystem is also essential to support the development and deployment of AI applications.
Why the need for NPUs?
NPUs are processors designed specifically for AI and machine learning scenarios. Unlike general-purpose Central Processing Units (CPUs) and Graphics Processing Units (GPUs), NPUs are optimized in hardware architecture for tasks like neural network inference, focusing on AI-related computations.
In contrast, CPUs excel at handling general computing and control logic, while GPUs have parallel computing advantages, particularly in tasks like graphic rendering. As a specialized neural network accelerator in AI PCs, NPUs efficiently support various AI application scenarios.
For example, in video conferencing, NPUs enable intelligent background segmentation. Previously reliant on GPUs for complex image segmentation and synthesis, NPUs now efficiently handle these tasks locally, providing faster response times and reduced power consumption, resulting in smoother video calls.
In image processing, NPUs efficiently run AI models for tasks like intelligent noise reduction and super-resolution. Previously, these operations relied heavily on the GPU's powerful parallel computing capabilities. Still, NPUs now offer similar image quality with lower power consumption.
Even in system management, NPUs play a significant role, running AI models for real-time monitoring and analysis, intelligently adjusting CPU frequencies, fan speeds, and other parameters for more intelligent, energy-efficient power management, particularly crucial for mobile devices like laptops.
Additionally, NPUs empower fields like speech recognition and natural language processing, providing users with more natural, intelligent interaction experiences. However, for complex large-scale AI models, strong computational power from CPUs and GPUs is still necessary support.
In summary, NPUs, as specialized neural network accelerators in AI PCs, excel in various scenarios like video, image processing, and system management, delivering excellent AI acceleration capabilities for users, resulting in more intelligent, efficient, and energy-saving computing experiences. This is why major manufacturers are integrating NPUs into AI PCs.
Is AI impossible without NPUs?
Many may wonder if a computer lacking an NPU means it can't use AI technology. In reality, that's not the case.
NPUs, as specialized neural network accelerators, primarily enhance the efficiency and experience of AI applications on personal computers. For instance, in scenarios like video conferencing background segmentation or intelligent image denoising mentioned earlier, NPUs significantly boost performance and reduce power consumption. However, this doesn't mean AI functions are impossible without an NPU.
For some cloud-based AI services, users only need an internet connection to access support without requiring local hardware acceleration. Applications like smart voice assistants or image-to-text converters can leverage cloud-based AI models for computation without depending on local NPUs or GPUs.
Of course, for running more complex large-scale AI models locally, such as large language models or extensive generation models, stronger computational support is still necessary, including dedicated GPUs or multicore CPUs. Even with an AI PC equipped with an NPU, GPU support is often required for most computations in these applications.
Therefore, while NPUs are crucial components of AI PCs, significantly enhancing the performance and efficiency of common AI applications, they're not mandatory for all AI applications. Cloud-driven intelligent services can be supported on regular PCs, while AI-heavy applications requiring substantial local computing power may still rely on GPU and other hardware support, even in AI PCs with NPUs. Do we really need to switch to an AI PC?
Most users probably haven't embraced AI PCs yet, but we're already experiencing a plethora of online AI tools. These products and services tap into powerful cloud-based AI models, making AI applications on various devices more convenient without relying on local computing power.
However, we must also recognize AI PCs as a significant trend in personal computing development. The transformation brought about by AI PCs goes beyond just enhancing certain efficiencies or empowering a few AI applications. It fundamentally delivers unprecedented quality experiences for users leveraging artificial intelligence technology.
For instance, intelligent retrieval and management of local files, or AI-optimized power management strategies that significantly enhance battery life and quietness for users. AI PCs integrate artificial intelligence into various aspects of personal computing, revolutionizing human-machine interaction and user experience.
Looking further ahead, AI-powered wearable devices, mobile terminals, and other new types of smart hardware are on the rise, blurring the boundaries between them and AI-powered PCs. Intelligence and interactivity are common features of these devices, with AI being the key driving force behind realizing these features.
Therefore, although there are various ways to obtain AI capabilities at present, AI PCs, as the deep integration of artificial intelligence in the personal computer field, will fundamentally reshape our usage experience and cognition of computing devices. In the future, the boundaries between multiple devices will gradually disappear due to AI, and only through seamless integration of artificial intelligence with hardware can the full potential of AI be maximized.
