Last year,we introduced the new changes that AI PCs bring compared to regular PCs through the AMD Ryzen AI PC.At that time,the AI experience was limited to only a few features and applications such as Windows Studio Effects and Topaz Photo AI,and everyone was looking forward to AI PCs bringing more groundbreaking changes.After more than a year of development,the concept of AI PCs has now become widely known.So,what new experiences can an AI PC offer if you were to purchase one now?

From Nothing to Something

From Cloud to Edge

What should an AI PC look like?Perhaps many people couldn't answer this question last year because,although the hardware era of AI PCs had already arrived,there were not many AI applications available at that time,and the entire market was in an awkward situation of "having hardware (NPU) but no applications." At that time,AMD was the first to launch processors with built-in NPU,but the AI software ecosystem was still in its infancy,and the AI applications that ordinary consumers could directly use were still the Windows Studio Effects launched by AMD in collaboration with Microsoft.It could add AI features such as background blur,automatic framing,and eye correction to the camera at the system level.

Coincidentally,generative AI applications began to take off in 2023.Although at the hardware device level,there were already AI PCs with built-in NPU on the market,the public's focus on generative AI applications was still on cloud-based large model services represented by ChatGPT,Wenxin Yiyan,and Tongyi Qianwen,and most people were still using cloud-based large model services accessed through web pages.

The change occurred in 2024.At the end of 2023,AMD and Intel successively launched new generations of processors with built-in NPU.Most of the new laptop products equipped with these processors were launched in the spring of 2024.With the launch of new products,chip giants and OEMs are working hard to promote the "AI PC" concept.In this process,chip giants,OEMs,and some ISV service providers are also continuously improving and developing the AI software application ecosystem,allowing AI applications to adapt to AI PCs to bring out the strength of AI PC hardware.As a result,we have seen ASUS,Lenovo,HP,Honor,Thunderobot,and other notebook manufacturers accelerating the creation of their own AI large model applications this year and embedding them in their notebooks.

ASUS launched a brand new "Dou Ding AI Assistant" in the a-Dou 14 Air notebook,which is supported by Byte's large model technology and provides users with AI functions such as communication,information retrieval,and text creation.Lenovo launched a personal intelligent entity "Lenovo Xiao Tian," which is driven by the embedded Tianxi large model and can bring a brand new AI experience in many scenarios such as work,study,and life through natural language interaction.HP launched a new generation of "Hui Xiao Wei" personal assistant,combining cloud-based large models and local large models to bring users a brand new AI experience.Honor launched a brand new "YOYO AI Super Assistant" on the MagicBook Pro16,which connects to online large models and integrates local computing power,supporting AI functions such as smart search,document summarization,smart recommendations,and text creation.Thunderobot not only launched the Thunderobot AI Book 15 Divine Edition with embedded local AI large model Euler at the press conference in May this year,but also enabled the personal assistant "Xiao Lei AI."Whether it's the AMD Ryzen 7040 series,the Ryzen 8040 series,or the recently launched Ryzen AI 300 series processors,these processors with built-in NPU can provide comprehensive AI computing power of CPU+GPU+NPU on PCs,thereby enabling common types of AI-generated content (AIGC) for graphics and text creation or handling complex graphic processing,large-scale data analysis,and other high-load tasks.With the continuous development of AI software ecosystems,the most significant change in the AI PC market today is the shift of AI application experience from the cloud to the edge,with industry participants striving to tap into the potential of local AI computing power.

What sets an AI PC apart from a regular PC if you were to purchase one now?The answer is that current AI PCs come with powerful NPU-based AI computing capabilities,which can efficiently support edge AI applications.Moreover,most of them come pre-installed with AI-related software and services,enabling many AI features accelerated by local hardware,such as AI video enhancement,AI voice dialogue,AI text generation,AI background removal,AI painting,etc.

A Comprehensive Experience of AI Applications on AMD AI PCs

Next,we will take a detailed look at an AMD Ryzen AI PC to see what sets a laptop pre-installed with AI-related software and services apart.The laptop we have chosen for this experience is the Lenovo Xiaoxin Pro16 AI Yuanqi Ryzen Edition,which is equipped with the AMD Ryzen 7 8845H processor.This processor features the AMD Ryzen AI NPU,with an NPU computing power of 16 TOPS.With the enhancement of the three major engines—CPU,GPU,and NPU—the overall computing power of the AMD Ryzen 7 8845H platform reaches up to 38 TOPS.

It should be noted that,as of the time of writing this article,new laptop models with the AMD Ryzen AI 300 series processors have not yet been released,and the overall computing power of the Core Ultra platform (34 TOPS) is not as high as that of the AMD Ryzen 7 8845H platform.Therefore,experiencing AI applications through the Lenovo Xiaoxin Pro16 AI Yuanqi Ryzen Edition is quite referential for everyone.For ease of reading,we will experience from the following major aspects one by one.

Personal Intelligent Entity Lenovo Xiaotian

This is a significant highlight of this AI PC.Nowadays,various notebook manufacturers have launched their own intelligent assistants,and the functionalities are essentially similar across the board.Therefore,we will mainly use "Lenovo Xiaotian" as an example here.Lenovo Xiaotian is built into the Lenovo Xiaoxin Pro16 AI Yuanqi Ryzen Edition and is driven by a local Tianxi large model.Through natural language interaction,it can bring a new AI experience in scenarios such as work,study,and life.Leveraging the user's local documents,Lenovo Xiaotian can also continuously deepen its understanding and cognition of the user,thereby providing more accurate and personalized services.Lenovo Xiaotian's core functionalities are centered around four key aspects: First,it possesses a local knowledge base; second,it generates local knowledge summaries,offering a "summary/induction" feature based on personally specified documents,with the ability to automatically match the output language; third,it ensures data security protection by providing chip-level security solutions combined with Lenovo's account management system,ensuring that the user's knowledge base can only be accessed by the local large model; and fourth,it fosters a natural interaction AI ecosystem,enabling natural interaction through handwriting recognition and voice recognition to create personalized images,content,and more.

Lenovo Xiaotian supports the management and Q&A of local personal knowledge bases,AI artist,AI PPT,and Q&A features.Let's first look at the management and Q&A functionality of the local personal knowledge base.This feature exclusively utilizes the local large model to process user-uploaded files such as PPTs,Word documents,and PDFs (image uploads are not currently supported),without the need for internet connectivity.This ensures greater data privacy and security.After uploading a file,users can engage in Q&A regarding the uploaded files,such as requesting Lenovo Xiaotian to translate,read aloud,summarize,or generate a PPT.I uploaded an English press release from AMD and asked it to translate it into Chinese; Lenovo Xiaotian provided an accurate translation.Since it's an offline large model handling the task,it can complete the translation even without an internet connection.Additionally,users can request Lenovo Xiaotian to summarize document content,which it processes with considerable accuracy.

The AI PPT feature of Lenovo Xiaotian requires an internet connection to operate.It can automatically generate a PPT in one click based on the user-uploaded materials,documents,and input themes.I asked the AI PPT to create a laptop recommendation PPT and uploaded a Word document.Lenovo Xiaotian was able to automatically summarize the PPT outline,which users can also manually adjust.Then,users proceed to select a PPT template.The system offers a variety of styles,including flat minimalist,business technology,artistic fresh,cartoon hand-drawn,and Chinese style,catering to the needs of most office workers.Afterwards,the AI PPT automatically generates the PPT for the user to download,and if not satisfied,users can go back and edit it again.

Overall,the AI PPT feature produces good-quality presentations and is definitely a lifesaver for those who are not proficient in creating PPTs or rarely do so.Even if you are a seasoned PPT user,it can serve as a great aid.By the way,the PPTs generated by AI PPT can also be downloaded in formats such as PPT,image,or PDF,which is very user-friendly for beginners.

Next,let's explore the AI artist feature.Lenovo Xiaotian's AI artist functionality is based on the cloud large model's text-to-image generation capability.Users can input textual descriptions,and the AI artist will create corresponding images.This feature is very helpful for office workers and designers,as it can trigger new inspirations at any time.Within Lenovo Xiaotian,the AI artist can generate an image of a cute white dog lying on a sofa as requested.The AI artist in Lenovo Xiaotian is not the full version of this feature; the full version is another software—Creator Zone.In this built-in software,users can set the parameters for text-to-image generation more finely,including style,proportion,resolution,and number of iterations,offering a richer and more comprehensive set of features.

Driven by a large model,Lenovo Xiaotian also supports knowledge Q&A,text creation,and other common functionalities of large model conversational assistants.Importantly,Lenovo Xiaotian supports both local and online cloud large models,making it usable whether offline or online,which is quite rare among other laptops.Users can switch between local and cloud large models through the icons above the dialog box.Let's explore the differences between local large models and cloud-based large models.When asking the same question,it can be observed that in an offline state,Lenovo Xiaotian,based on a local large model,can accurately answer the meaning of "tender as water,a date as sweet as a dream," albeit in a more concise manner.In contrast,under a cloud-based large model,leveraging the high computational power of the cloud,Lenovo Xiaotian provides a very accurate response,explaining the origin of the phrase and giving us a detailed introduction to its meaning.

Lenovo Xiaotian also features a mini-program store,indicating that Lenovo plans to integrate other AI applications through mini-programs to create its own AI ecosystem.The mini-program store now includes well-known applications such as Kimi AI Assistant,Meitu Clipping,Evernote's H5 web page generation,Tongyi Wanshang,Flash Resume,NetEase Youdao Translation,Baidu Dujia,etc.Users can use or activate the corresponding AI services through the one-stop mini-program store according to their actual needs,saving the trouble of downloading separate software or opening web pages.In addition,Lenovo Xiaotian also supports a "Hi Chat Mode," in which users simulate immersive voice calls with Lenovo Xiaotian,engaging in in-depth communication and discussing various topics.

Edge AI Large Models

When AI PCs were born last year,some people might have thought it difficult to successfully run generative AI applications on PCs,but now it's no longer a challenge.The edge AI large models we are referring to here are the deployment of various types of large models locally on AI PCs,enabling natural language processing and other operations even when disconnected from the internet.The edge AI large models mentioned here are divided into two categories: one is the smart assistants with built-in local large models developed by OEMs themselves,similar to Lenovo Xiaotian,which is based on the Tianxi large model and is a personal intelligent entity that we have experienced earlier; the other is software developed by other manufacturers that can help users deploy large models on PCs,such as the AMD Ryzen AI Assistant we are about to experience.

The AMD Ryzen AI Assistant is developed based on the self-developed optimized inference tool chain MLGuider for large language models (LLMs) by Qinghao Intelligence.It is developed and runs on the AMD Ryzen 7040 and Ryzen 8000 series AI PC platforms,capable of loading multiple types of edge large models,and can quickly experience generative AI without the need for an internet connection.It now supports functions such as regular Q&A text generation,code generation,and health assistance suggestions.

Thanks to the AMD Ryzen AI platform's ability to provide integrated AI computing power from CPU,GPU,and NPU,deploying local large models on AMD Ryzen AI PCs to provide AI application solutions has been practically implemented in both commercial and consumer scenarios.Currently,these large models cover various vertical fields such as medical imaging,programming,healthcare,education,and law.They can be integrated into users' workflows,helping users generate knowledge graphs,contracts,legal opinions,briefings,etc.,thereby improving users' work efficiency.For example,the high-performance large model developed by Wuwen Xinqiong Company — Wuqiong Tianquan,which excels in logical reasoning and long-text processing,supports text up to 256k characters and can directly process the entire "The Three-Body Problem 3: Death's End." It can achieve excellent inference acceleration on the AMD Ryzen AI PC.

On the Lenovo Xiaoxin Pro16 AI Yuanqi Ryzen Edition,we have deployed several mainstream large models through the Ryzen AI Assistant to see the actual performance of this machine running large models!These large models include Qianwen 14B-Chat,BaiChuan 2-13B-Chat,ChatGLM3-6B,Bianque 2,etc.

In actual experience,Qianwen 14B-Chat has the largest model parameters,and its accuracy is also higher.When running the Qianwen 14B-Chat model on the Lenovo Xiaoxin Pro16 AI Yuanqi Ryzen Edition,the processor power consumption can soar to about 51W-60W,and the processor occupancy rate reaches about 65%.After all,this large model has a large number of parameters and requires more powerful performance.ChatGLM3-6B is very responsive,but it seems to have limited knowledge of Chinese literary knowledge.BaiChuan 2-13B-Chat,compared to QianWen 14B-Chat and ChatGLM3-6B,provides overly concise answers but with a high degree of accuracy.CodeGeeX2-6B,on the other hand,is a lightweight open-source code generation model that can automatically generate code for multiple programming languages,boasting powerful capabilities in code generation and completion,adding comments,code translation,and technical Q&A.Once deployed locally on a PC,it can accurately answer questions like "implement a simple greedy algorithm in Java" even in an offline environment,which is quite helpful for the programmer community.

Stable Diffusion Image Generation

When it comes to the most popular AI applications currently,in addition to generative large models,image diffusion model applications represented by Stable Diffusion are also widely pursued.Compared to various online AI image generation tools,Stable Diffusion offers excellent freedom and a plethora of free resources.Deployed locally on a PC,it ensures security,privacy,and convenience for users.

With the computational power synergy of the CPU,GPU,and NPU in the AMD AI PC,it naturally can handle Stable Diffusion with ease.The Stable Diffusion Web UI is primarily driven by the GPU,and the AMD Ryzen 7 8845H processor's integrated AMD Radeon 780M graphics card,which uses the RDNA 3 architecture and is equipped with 12 CU units and 768 stream processors,provides a strong computational foundation.Therefore,the AMD AI PC also has an advantage in Stable Diffusion image generation.

From practical testing,with a 4GB video memory setting,the accelerated generation of a 512x512 resolution image through the AMD Radeon 780M graphics card takes 43 seconds.This efficiency can meet the usage needs of AIGC for groups such as new media operators.

Video Conference Enhancement

Video conference enhancement is a common feature for AMD Ryzen AI PCs,mainly referring to the Windows Studio Effects launched by AMD in collaboration with Microsoft.With NPU support,Windows Studio Effects can add AI features such as background blur,automatic framing,and eye correction at the system level for the camera.For ordinary users,it is convenient to enable Windows Studio Effects driven by the NPU in the "camera" settings of the AMD AI PC.It should be noted that the AMD Ryzen AI PC has also brought a new change— the well-known HWinfo software now supports monitoring the utilization of the AMD NPU,which was not possible last year.Windows Studio Effects is a feature that originates from the depths of the Windows operating system.It not only affects the "Camera" application within the Windows system but also supports any video calling software.For instance,when we engage in video calls through TIM software,which does not support any video effects,you might feel slightly embarrassed if your laptop does not support Windows Studio Effects,especially when the environment during the call is chaotic.With the aid of Windows Studio Effects,users can enjoy the charm of features such as background blur,automatic framing,and eye contact correction during video calls with TIM.Moreover,this feature is driven by the Neural Processing Unit (NPU),so activating it will not increase the usage of the CPU or GPU.

Gaming Assistance

AMD Ryzen AI PC has empowered a myriad of industries,and in addition to enhancing large-scale models on the edge and video conferencing,it has also demonstrated its value in the realm of gaming and entertainment.Currently,the gaming assistance tool "Game Plus" has integrated AMD NPU support for features such as "Super Moments," "Peripheral Intelligent Light Control," and "AI Gaming Coach." The "Super Moments" feature primarily utilizes AI technology to automatically recognize and record the moments of kills in games,assisting players in easily saving and sharing the highlights of their gaming experiences without the need for manual recording.The games currently supported by this feature include "Eternal Return," "Valorant," "PlayerUnknown's Battlegrounds: Battle Royale," "APEX Legends," and "Call of Duty: Modern Warfare 3."

Let's briefly experience the "Super Moments" feature driven by the NPU through the game "Eternal Return." First,in the settings interface,you can see that there are two AI recognition modes for this feature: General Mode and AMD Ryzen AI NPU Mode.In General Mode,implementing this feature requires additional resources from the GPU,whereas in Ryzen AI NPU Mode,the entire AI inference is completed by the NPU,thus not consuming additional GPU resources.This has a minimal impact on the actual game's frame rate,and AMD and game developers have also collaborated deeply in training the AI models,further enhancing the accuracy of AI recognition in this feature.

After enabling the AMD Ryzen AI NPU Mode in the "Super Moments" feature settings and entering the game,"Game Plus" can monitor the NPU's frequency and usage in real-time.After playing a round,the exciting Super Moments videos will be automatically saved on your computer,which is very convenient.

In conclusion,the integration of AMD Ryzen AI NPU into gaming assistance tools like "Game Plus" has significantly enhanced the gaming experience by providing features that are not only resource-efficient but also highly accurate,thanks to the collaboration between AMD and game developers in AI model training.After more than a year of development,AI PCs represented by the AMD Ryzen AI platform have reached a new stage.Nowadays,people have come to regard the CPU,GPU,and NPU as the three major computational engines of AI PCs.AMD is an innovative leader with advantages in all three areas,and by promoting the practical implementation of AI application solutions through the collaboration of the three,it truly unleashes the power of AI PCs.Because of this,we are witnessing the era where large-scale AI model applications on the edge are flourishing.

In fact,AMD's ISV service providers and ecosystem partners have exceeded 150,with more partners joining in the future.The AI application solutions suitable for the AMD Ryzen AI platform are far more than just the few features we experienced this time.From industries such as healthcare,education,and consulting to consumer scenarios like graphic design and gaming assistance,the AMD Ryzen AI platform has brought an unprecedented new experience to AI PCs.

From the cloud to the edge,this is an inevitable path in the development of AI PCs.On one hand,AI PCs inherently possess strong edge AI computing power with the collaboration of CPUs,GPUs,and NPUs,providing a foundation for the local deployment of large-scale model applications.On the other hand,as we can see from our experience and testing,large-scale model applications have become increasingly practical,truly helping users in their daily lives and entertainment,and the development of edge large-scale model applications has also lowered the barrier for ordinary users to experience AI applications.

Lastly,we must not overlook some of the advantages of edge AI applications,such as cost-effectiveness—deploying large models on the edge eliminates the need to purchase value-added services in the cloud,and local data processing also reduces the cost of cloud services; security—data remains in the hands of users,avoiding the hassle of privacy leaks; and convenience—local AI services can operate even in the absence of a network connection,avoiding network restrictions.

Overall,AI PCs represented by AMD Ryzen AI are no longer just a concept but have become a more valuable tool that ordinary users can perceive,experience,and that can truly help users.In the face of the tide of the times,it is time for everyone to embrace AI PCs and set sail towards the AI era.