When it comes to Windows PC processors, Intel and AMD are the first thing that comes to mind for most people, and Qualcomm may break this inertia.
At the 2024 Lenovo Tech World held last week, Ammon Qualcomm President and CEO, Pat Gelsinger, CEO of Intel, and Lisa Su, Chairman and CEO of AMD, shared the stage.
Lenovo is the world's No. 1 company in PC shipments, and it is not uncommon for Intel and AMD CEOs to appear at Lenovo's important events, and it is rare for Qualcomm CEOs to appear at the same time.
Even an AMD vice president lamented at the Lenovo Innovation and Technology Conference, "So far, this is the first time I have promoted AI PCs with people from Intel and Qualcomm." ”
AI PC has gathered the CEOs of Qualcomm, Intel, and AMD, which is enough to show the great potential of the AI PC market in the future. It also seems to indicate that the PC market landscape of the past few decades may be reshaped by generative AI.
The new opportunities of AI PCs are rare to gather Qualcomm, Intel, and AMD
Even though AI PCs are becoming more and more widely spread, many people still don't know the difference between AI PCs and traditional PCs.
Lenovo showed several good examples last week, such as the famous director Lu Chuan using AI to generate a film, and young people with movie dreams can use AI PCs to make movies beyond the financial threshold.
For those who work, the intelligent body equipped with the AI PC, after learning the personal knowledge base, has become an all-round work assistant, greatly improving work efficiency, and even answering the professional questions of Li Yuansheng, the first director of China's Antarctic Kunlun Station and a pioneer of China's deep ice core science.
For students, with the help of generative AI functions such as language generation pictures and knowledge quizzes of AI PCs, AI can help children make PPT efficiently, significantly reducing the burden on parents.
Productivity, learning, and entertainment are the three major scenarios for AI PC applications, and these scenarios will give rise to killer applications, bringing efficiency improvements and better entertainment experiences.
Bringing a new AI PC experience to consumers is a new opportunity for AI PC and AI terminal companies.
Ablikim Ablimiti, vice president of Lenovo Group and chief strategy officer of Lenovo China, believes that AI terminals are a new opportunity after various terminals such as feature phones and home computers, as well as smartphones, laptops, wearable devices, and smart homes in the past 30 years.
"New opportunities will give birth to a new terminal ecological landscape. "AI terminal manufacturers will develop personal agents and deliver AI computing power to users through a reliable, effective and open platform." In this process, AI computing power providers will upgrade from the original computing architecture with general computing as the core to a hybrid AI computing architecture that includes CPU, GPU, NPU. ”
Alex Katuzan, senior vice president and general manager of the Mobile, Compute and XR business at Qualcomm Technologies, said when he unveiled the Snapdragon X Elite last year, "In the future, users will consider the performance and application of AI in addition to CPU and GPU when making PC purchase decisions." ”
Snapdragon X Elite is a high-performance and low-power PC computing platform including CPU, GPU, and NPU, and Qualcomm is also relying on Snapdragon X Elite to enter the AI PC market with huge potential with Intel and AMD.
There are two core parameters of an AI PC: AI performance and battery life
Hybrid AI is the future of AI is almost an industry consensus.
"Hybrid AI leverages AI on both the device side and in the cloud to distribute and orchestrate workloads through high-performance connectivity, resulting in smarter, personalized user experiences. "To meet the demands of generative AI-centric experiences, industry-leading connectivity and high-performance, low-power computing are more important than ever." ”
Qualcomm, Intel, and AMD are all highlighting the AI capabilities of AI PC computing platforms, but only Qualcomm's Snapdragon X Elite has a significant low-power advantage.
The 4nm Snapdragon X Elite, released in October last year, has the fastest NPU for the PC platform, with a computing power of up to 45TOPS, plus Qualcomm's self-developed Oryon CPU and Adreno GPU, which achieves AI performance of up to 75TOPS on the Snapdragon X Elite platform through a heterogeneous computing architecture.
The heterogeneous computing architecture of the Snapdragon X Elite can fully leverage the advantages of each processor to meet the diverse requirements and computing needs of generative AI.
Only powerful AI computing power can bring excellent AI experience.
At this year's MWC, Qualcomm showcased two laptops equipped with Snapdragon X Elite and X86 chips, running GIMP (a popular image editor) with integrated Stable Diffusion plug-in to generate AI images, Snapdragon X Elite can generate an image in just 7.25 seconds, which is 3 times faster than X86 competitors (22.26 seconds), and from the generation results, Snapdragon X Elite generates higher image quality and more detail.
Qualcomm's AI performance advantage is due to its long-term accumulation in the field of device-side AI. Of course, while the Snapdragon X Elite has become the industry benchmark in key AI performance, the self-developed Oryon CPU and Adreno GPU have also brought surprises and shocks to the entire PC industry.
When the Snapdragon X Elite was released, it had a significant performance advantage over Apple's latest M2 processor and x86 processor at the time. Now, in the face of Apple's latest M3 processor, as well as Intel's new generation of Core Ultra 7 155H, the Snapdragon X Elite released last year still has a clear advantage.
In the Geekbench multi-threaded test, the Snapdragon X Elite has a 28% performance advantage over the Apple M3.
Compared to the Core Ultra 7 155H, the Geekbench v6 single-threaded test shows that the Snapdragon X Elite consumes only 65% of the Core Ultra 7 155H when reaching the same peak performance. At the same power consumption, the Snapdragon X Elite delivers 54% more CPU performance than the competition.
In the Geekbench v6 multi-threaded test, the Snapdragon X Elite achieves the same peak performance as the Intel Core Ultra 7 155H with 60% lower power consumption, and the Snapdragon X Elite CPU performance is 52% better than the competition at the same power consumption.
The GPU performs the same as the CPU, with the Snapdragon X Elite's Adreno GPU reaching the same peak performance as the Intel Core Ultra 7 155H at half the power, and the Snapdragon X Elite's GPU performance is 36% higher than the competition at the same power consumption.
"These are two very important dimensions of reference, on the one hand, that our products are significantly better at the same power consumption, and that our products consume significantly less power than the competition at the same peak performance." ”
High-performance, low-power CPUs and GPUs deliver longer battery life and faster application response times. For example, Office 365 apps, local video playback, web browsing, YouTube streaming, and Teams video calling, the Snapdragon X Elite has at least 40% longer battery life than the Intel Core Ultra 7 155H, and at most more than 2x better.
Responsiveness is also an important indicator of user experience, and the shorter the latency when launching an app and the time it takes to open it for the first time, the better. A very intuitive statistic is that the three web browsers on the Snapdragon X Elite, Google Chrome, Microsoft Edge, and Brave, are 20%, 57%, and 15% more responsive, respectively, compared to the Intel Core Ultra 7 155H.
At this time, it is easier to understand that computing power determines the experience of the application, and power consumption determines the battery life, which are the two key parameters for choosing an AI PC.
In the past, it was difficult for PCs to combine high performance, thin and light form factor, and long battery life, because the high performance of the chip means high energy consumption, just like a sports car with a large displacement engine is difficult to achieve low fuel consumption.
With its self-developed Oryon CPU, Adreno GPU, Hexagon NPU, and the Snapdragon X Elite, which gives full play to the advantages of heterogeneous computing, has stronger performance than traditional PCs and supports new AI features, who can say no to such a Window PC?
PCs with Snapdragon X Elite, which will be launched in the middle of this year, could reshape the PC market once they are widely accepted by consumers.
Two soft powerhouses of the AI PC boom
The reshaping of the PC market pattern will be accompanied by the maturity of AI PCs.
The maturity of AI PCs requires the most powerful hardware platform, the best tool chain for developers, and more importantly, the need to help software partners develop products to achieve business closed-loop.
The easy-to-use and easy-to-deploy toolchain tests the soft power of various chip companies and is the key to lowering the development threshold.
Qualcomm has launched a unified software stack across platforms, terminals, and operating systems – the Qualcomm AI Stack.
Wan Weixing, head of Qualcomm's AI product technology in China, said, "Qualcomm's AI software stack supports all current mainstream training frameworks and execution environments, and we also provide developers with different levels and levels of optimization interfaces, as well as a complete compilation tool chain, so that developers can complete the development, optimization and deployment of models more efficiently on the Snapdragon platform." ”
There is also the Qualcomm AI Hub, which makes it easier for developers to deploy AI, allowing developers to run these models on cloud-hosted terminals powered by the Qualcomm platform with just a few lines of code.
Because Qualcomm's optimized models, such as Whisper, ControlNet, Stable Diffusion, and Baichuan-7B, can take full advantage of the hardware acceleration capabilities of all cores (NPU, CPU, and GPU) in Qualcomm's AI engine, increasing inference speed by 4x, and reducing the memory bandwidth and storage space occupied by the model.
According to Leifeng.com, there were more than 75 optimization models when the Qualcomm AI Hub was announced in March, and now it has provided about 100 optimized AI models for Snapdragon and Qualcomm platforms.
Qualcomm has invested resources to keep up with the development of AI technology, such as cooperating with Meta to optimize and take the lead in supporting the recently open-sourced Meta Llama 3 large language model, and developers can use the resources and tools provided by Qualcomm AI Hub to optimize the operation of Llama3 on the Snapdragon platform.
In addition to continuously optimizing their own software tool chains, companies such as Qualcomm, Lenovo, Intel, AMD, Alibaba Cloud, Baidu, and Baichuan have also proposed joint incentive plans for AI application developers to jointly promote the landing and prosperity of AI PCs, and a new AI PC ecosystem is also forming.
While AI PCs are booming, other AI terminals, including AI phones, will also develop rapidly. In order to create a more intelligent AI terminal experience, a seamless cross-terminal experience is indispensable, and both terminal equipment manufacturers and chip providers have already laid out their plans.
Qualcomm has introduced Snapdragon Seamless, a cross-platform technology that allows Snapdragon devices using Android, Windows and other operating systems to discover each other and work as if they were using a unified integrated system to share information.
Shaping a responsible AI future is also even more important at this time. Qualcomm's device-side AI solutions are designed with user security and privacy in mind to support a more trusted AI ecosystem.
Thanks to Qualcomm's extensive ecosystem layout and unified technology roadmap, as well as its powerful connectivity and computing products, Qualcomm will be able to occupy an important position in the process of the rise of AI terminals, and is also expected to reshape the market structure in the process of PC market shift to AI PC.