How far are we ordinary people from the big model? AI has developed to 2024 and has been attached to various smart hardware, such as AI PCs, AI mobile phones, AI Bluetooth speakers, AI learning machines, AI mice, etc., but in fact, although many products have added the name of AI, they always make people want to scold "artificial intellectual disability". The "Xiaoyi Xiaoai" in the mobile phone and speaker can only do some things within the framework of conventions, and he doesn't feel like he can learn it when he uses it at present. But AI PC seems to be able to complete the "last mile" between me and the large model, so that every ordinary person can have an expensive large model. Inventory and application of mainstream AI PC chipsIn September 2023, the concept of AI PCs was first proposed, and by February 2024, Intel said at MWC 2024 that it will work with companies such as Acer, Asus, Dell, Dynabook, Fujitsu, HP, Lenovo, LG, Microsoft Surface, NEC, Panasonic, Samsung, and VAIO to launch more than 100 business laptops, desktops, all-in-ones, and entry-level workstations powered by Intel's latest AI PC chips. After the concept of AI PC was gradually recognized by the market, the major old chip giants followed suit to launch new products. In October, AMD Ryzen 7040 series laptops were equipped with a dedicated Xilinx IP-based AI engine called "Ryzen AI" that accelerates machine learning frameworks such as PyTorch and TensorFlow; Also in October, Qualcomm released the X Elite chip for AI PCs; In January 2024, NVIDIA unveiled three new graphics cards loaded with additional devices, saying that users can use AI features on "AI PCs". At present, Intel, Qualcomm, AMD, Apple and other chip processor manufacturers have launched processors for AI PCs, and brand manufacturers such as Lenovo, Asus, Dell, and Honor are also actively promoting the development of AI PCs. 
Although there are various AI PC processors, the general computing architecture of each vendor basically adopts a heterogeneous solution of "CPU + GPU + NPU". 
Caption: The local operation of large models of AI PC processors from Qualcomm, Intel, and AMD requires powerful AI computing power, so an embedded neural network processor - NPU is added to the traditional CPU + GPU solution. NPU is a computer microprocessor specially designed to process artificial neural network calculations, which is used for deep learning and is responsible for the work of neural networks. The combination of the CPU, the GPU, and the NPU for AI acceleration can achieve faster and more efficient edge AI model inference. The high-general Snapdragon X Elite has entered the PC field, and Intel has a lot of momentum to occupy the AI PC marketMicrosoft has announced a minimum hashrate of 40 tops for its AI PCs, which neither the Intel Core Ultra (34 TOS) nor the AMD Ryzen 8040 (39 TOPS) currently on the market can meet. Aside from laptops with AI-specific graphics, only Qualcomm's Snapdragon X Elite (45 tops) can meet this performance requirement at the moment. It is expected that the AMD Ryzen 8050 and Intel Ultra 200 processors, which will be launched in the second half of 2024, will also meet Microsoft's 40 tops. From the perspective of computing power, Qualcomm is indeed one step ahead of other AI PC processors at present, but Intel's ecological position in PC chips is difficult to shake, and before Snapdragon X Elite, Qualcomm has not made much noise in the PC field. However, at the Snapdragon X Elite press conference, Qualcomm was very frenzied, pulling out Apple and AMD to compare, shouting "Our single-threaded performance surpasses the Apple M2 Max!" And the power consumption of the same performance is 30% lower than Apple! "It's the same with Intel" "80% better than AMD's GPU!" ”
Caption: At the Qualcomm press conference, the Intel i9-13980HX was compared with the power consumption
Caption: At the Qualcomm press conference, AMD Ryzen 9-7940HS was compared with GPU performance PPT data is very powerful, and there are also many PC OEMs that will carry it, Microsoft, Lenovo, HP, Acer, Asus, Samsung and other traditional notebook players, as well as Xiaomi and Honor. There are also many PC OEMs on Intel, Intel is unveiling the new vPro platform at MWC 2024, covering **Core Ultra and 14th Gen processors, introducing is a commercial computing platform, mainly for enterprise users, and will launch more than 100 in partnership with companies such as Acer, Asus, Dell, Dynabook, Fujitsu, HP, Lenovo, LG, Microsoft Surface, NEC, Panasonic, Samsung, and Vaio Business laptops, desktops, all-in-ones, and entry-level workstations powered by Intel's latest silicon. The merchants and data listed here have the ambition and momentum to occupy the PC market. AMD Ryzen AI products have already begun to shipAMD's AI portfolio spans cloud, high-performance computing, enterprise, PC, embedded and more, and its AI strategy is based on three key elements: product, architecture, and open ecosystem. Mobile CPUs such as Ryzen with integrated AI engines are part of the product portfolio, among which the Ryzen 8040 series has already begun to ship, and related terminals will be available in Q1. In addition, AMD also has the Ryzen 8000G series of desktop APU products, which are also based on TSMC's 4nm process, all of which provide up to 8 CPU cores of the ZEN4 architecture, integrate the RDNA3 architecture core graphics with up to 768SP configuration, and can support hardware ray tracing and AFMF technology. 
At present, products based on Ryzen AI technology have been used in Lenovo, Asus, Acer and other brand computers, these AI PCs provide low-power and high-efficiency AI computing power through the built-in neural processing unit (NPU) of Ryzen series products, and accelerate the processing speed of AI-related local applications such as **and** processing, speech recognition, text transcription, etc., thereby improving users' productivity and entertainment experience. 
The ARM architecture may benefit from the AI PC waveCompared with the x86 architecture, ARM's advantages lie in low cost, low power consumption, increasingly enhanced performance and system versatility, and a massive terminal device ecosystem, which are expected to be brought into play on mobile terminal AI PCs. Counterpoint Research**, with the iteration of ARM chips and the continuous improvement of performance, Apple, Samsung, MediaTek and other international manufacturers have tested the waters of PC chips based on ARM architecture, and Arm is expected to occupy 25% of the global market share of notebook CPU SoC shipments in 2027 in the future. The aforementioned Qualcomm Snapdragon X Elite processor, its internal Oryon CPU is a fully custom-designed processor core developed by Qualcomm based on the ARM instruction set. Unlike the traditional Arm Cortex public architecture, Oryon is a new CPU core introduced by Qualcomm after the acquisition of Nuvia, which is designed to improve the performance and power efficiency of devices such as mobile computing and laptops. It is also reported that NVIDIA and AMD have already started designing an ARM-based PC chip that will run the Windows operating system and launch it as early as 2025. Prior to that, Microsoft's first Surface tablet, Surface RT, used NVIDIA's ARM-based Tegra processor, but Windows RT failed to generate much market response. According to a report by foreign media Windows Central, Microsoft is preparing to upgrade its Surface product line with new AI features, or the Surface Pro 10 and Surface Laptop 6 released in 2024 will be equipped with a next-generation neural processing unit (NPU) and have two options for Intel and ARM processors. Further afield, in 2025, there may also be an updated Surface Laptop Studio. Microsoft has long supported ARM processors in its products, Windows RT is Microsoft's earliest attempt to run Windows on ARM architecture, the latest Windows 11 continues to support ARM architecture, and has made improvements in compatibility, performance and user experience, supporting more applications to run natively on ARM devices, while enabling more x64 applications to run on ARM devices through emulation technology. Microsoft has also partnered with hardware manufacturers such as Qualcomm and Samsung to launch a number of devices with ARM processors running Windows 10 on ARM, such as Always Connected PCs with Qualcomm Snapdragon processors, which combine the long battery life and always-on features of smartphones with PC productivity tool attributes. In addition to extensive hardware compatibility in the AI PC ecosystem, Microsoft also has a relatively important product, that is, the AI assistant Copilot, which gives users a new way of interaction, so that AI is more closely integrated into the operation and use of personal computers, making AI no longer an abstract concept, but a tool to improve productivity and convenience. SummaryAI PCs have created opportunities for processor chip manufacturers and PC manufacturers, and one of the more significant innovation trends is to provide higher computing efficiency and lower energy consumption by integrating AI acceleration units in CPUs, GPUs, or dedicated NPUs. AI application scenarios are relatively abundant, so another trend of processors is customization, which requires processor manufacturers to work closely with operating system developers, AI framework developers, and ODM OEMs to jointly promote AI applications and services, create unique selling points, and increase product added value.