From the beginning of 2023,when AMD introduced the first x86 platform processor with a built-in NPU,to the end of 2023 with giants like Intel and Qualcomm joining the fray,and into 2024,all participants in the PC industry,from chip manufacturers to OEM terminal manufacturers,are envisioning the future of the AI PC concept.However,for most consumers,what exactly is an AI PC?

AI PC is a concept

There has never been a clear definition in the industry for what constitutes a true AI PC.If we compare the differences between ordinary PCs and current AI PCs from the processor level,we can find that devices claiming to be AI PCs are all equipped with a dedicated NPU (Neural Processing Unit) to handle most of the AI workload on the device.Therefore,from this perspective,one could argue that "without an NPU,there can be no AI PC," and it is the NPU that makes AI applications on AI PCs more efficient.

Previously,NPUs have become quite common in smartphones,as they are specialized hardware accelerators designed specifically for AI and machine learning tasks.Unlike the "general" computing processor roles played by CPUs and GPUs,NPUs are more specialized and optimized to handle the complex computations required for various deep learning algorithms,making them more efficient in processing AI tasks.Additionally,NPUs can work in tandem with CPUs and GPUs to handle complex task requirements,providing faster processing speeds.On the other hand,due to the powerful AI computing power of NPUs,coupled with their collaboration with CPUs and GPUs,PC devices can run AI applications such as generative large models locally.

In the x86 platform,AMD was the first to bring an NPU to PCs.Following AMD,Intel and Qualcomm have successively integrated NPUs into their PC processors.It is the presence of NPUs and their ecosystem synergy that allows us to achieve more efficient and personalized end-side experiences when running AI technologies such as deep learning,natural language processing,and computer vision on PCs.

Tech giants have their own definitions for AI PCs.AMD believes that an AI PC should be a combination of CPU,GPU,and NPU.Microsoft,not long ago,introduced "Copilot+PC" and proposed that "any computer capable of running generative AI can be called an AI PC." Microsoft also defined its own AI PC standards: it must be equipped with an NPU (with a computing power of more than 40 TOPS),CPU,and GPU,and support Microsoft Copilot,with a physical Copilot button on the keyboard.Intel,on the other hand,believes that an AI PC is simply a PC that can handle AI functions.

Regardless of how the giants define it,in our view,an AI PC is inseparable from two key points: NPU and generative AI.Therefore,we believe that a PC equipped with an NPU and capable of running generative AI applications locally is an AI PC.What platforms currently support NPU?

Let's start with AMD.In early 2023,AMD introduced the first independent NPU for the x86 platform with the release of its Ryzen 7040 series processors,officially ushering in the era of AI PCs.Although the computing power of AMD's first-generation NPU is only 10 TOPS,it has brought consumers an exclusive AI application experience on the x86 platform and laid a solid foundation for the further development of software and operating systems in the future.

In early 2024,AMD's all-new Ryzen 8040 series processors were officially launched.Compared to the Ryzen 7040 series processors,the NPU computing power of the Ryzen 8040 series has increased from 10 TOPS to 16 TOPS,a 60% improvement,and the overall processor computing power has increased from a maximum of 33 TOPS to a maximum of 39 TOPS,achieving a 40% performance improvement in running language and vision large models.

Following the Ryzen 7040 series and Ryzen 8040 series processors,AMD released its third-generation AI PC processor in June this year: the Ryzen AI 300 series.This series not only includes a new generation of NPU architecture but also significant upgrades and changes in the CPU and GPU components.The CPU of this series adopts the brand-new Zen 5 architecture,with an official claim of an average IPC performance increase of up to 16%,and the GPU has been upgraded to the RDNA 3.5 architecture,with a very considerable performance improvement.The NPU part has been upgraded to the brand-new AMD XDNA 2 architecture,with AI processing power soaring three times compared to the previous generation,reaching an astonishing 50 TOPS.This not only significantly exceeds the Microsoft Copilot+ PC standard but also can be considered the strongest NPU among global AI PCs.

Next,let's look at Intel.Intel first integrated an NPU in its Core Ultra series processors launched at the end of 2023.However,Intel did not disclose the computing power of the NPU at the time,only emphasizing that the overall computing power of the Core Ultra platform reached 34 TOPS.In June 2024,Intel officially announced its next-generation mobile processor codenamed "Lunar Lake." The biggest change in the Lunar Lake processor is the integration of a brand-new fourth-generation NPU core (NPU 4),which has 6 Neural Compute engines,12 enhanced SHAVE digital signal processors (DSPs),and 9MB of cache,providing 48 TOPS of AI computing power.

Last but not least,we must mention Qualcomm.Qualcomm has been quietly working on ARM PCs in recent years.The company launched a brand-new Snapdragon X Elite series processor for the desktop platform in October 2023,joining the current AI PC market as a third force.The Snapdragon X Elite series processor integrates a brand-new custom-designed Qualcomm Oryon CPU,Adreno GPU,and an independent Hexagon NPU.The biggest highlight of this processor is its NPU computing power of 45 TOPS,which significantly surpasses other mobile processors of the same period.Even now,the NPU computing power of Snapdragon X Elite is not much different from that of AMD's Ryzen AI 300 and Intel's Lunar Lake,fully meeting the Microsoft Copilot+ PC standard.

In summary,with the launch of new processors from AMD and Intel,as well as the entry of Qualcomm from the Arm camp,the AI PC on the Windows platform is about to explode.Every laptop we buy in the future may have an integrated NPU and be able to run generative AI applications,making them AI PCs.After understanding AI PCs,some may wonder what the use of AI PCs is.Is it just limited to call noise reduction,camera auto-focus tracking,and automatic background blurring?Obviously,it goes beyond that.Led by upstream chip manufacturers,the software ecosystem of AI PCs today has flourished,and AI applications have been implemented in various industries."Microcomputer" has been continuously tracking and reporting on AI PCs.Looking at the industry,AMD,which was the first to launch and has already released three generations of AI PC processors,undoubtedly has a first-mover advantage,and its products are more representative and diverse.Next,we will take AMD's Ryzen AI PC as an example to show everyone the new experiences that AI PCs bring today.