March 29, 2024 By 0

Intel and Microsoft Jointly Define AI PC: Must Have Copilot Physical Buttons

AI PC is still on the rise, but there is no unified industry standard for what exactly AI PC is, and everyone is advocating their own concept. Intel and AMD, as platform level manufacturers, both believe that an AI PC must have a CPU, GPU, and NPU at the same time, while NVIDIA, as a graphics card giant, emphasizes that RTX GPU is the soul of AI PC.

In fact, Microsoft also has a say in AI PCs, as the true implementation of AI experience requires support and optimization at the operating system level. At an AI PC summit, Intel released a requirement jointly defined with Microsoft for an AI PC:

On the one hand, it is equipped with processor chips that include brand new NPUs, GPUs, and CPUs, such as Intel’s Core Ultra (and of course, AMD’s Ryzen 7040/8040 series).

On the other hand, it is necessary to support Microsoft Copilot and equip it with Copilot physical buttons.

The first point has nothing to say, but the second point is a bit troublesome, mainly because it is quite unique in China and cannot be used through conventional methods.

Perhaps in the future, Microsoft will bring Copilot to China through special methods, such as developing customized versions or collaborating with domestic manufacturers for customization, but it will definitely not work in the short term and will also cause fragmentation. Moreover, existing AI PCs are not equipped with Copilot buttons, are they not real AI PCs? Perhaps this is Microsoft’s definition of the future AI PC.

Of course, the current top priority is to encourage more manufacturers and developers to invest in AI PCs, bringing more products and applications. For this reason, Intel launched the AI PC Acceleration Program in October last year, and has now undergone two major upgrades: one is to increase the AI PC developer program, and the other is to attract independent hardware developers to join.

According to the AI PC Developer Program, Intel will provide developers with rich support resources, including frameworks and libraries such as OpenVINO, ONNX runtime, DirectML, WebNN, AI tools, workflows, code examples, and use cases, AI models optimized for Intel XPU, hardware development toolkits, one-stop knowledge centers such as documentation, training, and technical blogs.

Among them, the AI PC development kit currently provides an Asus NUC Pro 14 mini computer that is already AI ready, with a Core Ultra processor on the hardware, four USB 3.2 and two Thunderbolt 4 interfaces, and can expand up to four displays and 96GB of memory without any performance bottlenecks. The software is pre installed with AI stack, XPU driver, and a complete development environment.

For the support of IHV independent hardware developers, Intel provides open laboratory support in Chinese Mainland, Taiwan, China and the United States, and provides access to resources and early reference hardware 24/7. During the pilot phase, over 100 IHV manufacturers have already developed over 200 components that are compatible with AI PCs, and there will be significant additions, optimizations, and enhancements in the future.

For example, for screens, it is best to use OLED, which is more energy-efficient and can also support lower refresh rates. For example, cameras need to provide high-quality images and videos. For example, memory and storage need to be optimized for AI PC platforms, and large models have significantly increased their requirements in this area. For example, visual sensing requires the integration of factors such as facial recognition, gesture manipulation, user posture, and human eye tracking, all of which are indispensable foundations for AI experience.

Although these are not mandatory requirements yet, perhaps in the future, Intel will create a hardware standard and experience certification for AI PCs, just like the current Intel Evo strict certification.

At present, more than 100 ISV independent software developers have participated in the optimization of AI PCs in the industry, and more than 300 AI accelerated software functions will be implemented within this year. It is expected that there will be over 100 million AI PCs worldwide by 2025.