Intel & Microsoft Set AI PC Standards: NPU + Copilot Keys

On March 27th, Intel unveiled two new initiatives under its AI PC Acceleration Program, aimed at fostering innovation and standardization in the realm of electronic components. These initiatives include the New PC Developer Program and the Independent Hardware Vendor (IHV) Program. Additionally, Intel partnered with Microsoft to introduce standardized definitions for AI PCs.

In the landscape of AI PCs, various manufacturers including Intel, AMD, Apple, and Qualcomm have introduced chips with dedicated AI processing units. However, the absence of standardized definitions has led to discrepancies in understanding AI PCs across the industry. To address this, Intel and Microsoft collaboratively established criteria for AI PCs. These criteria stipulate the necessity of NPUs, CPUs, and GPUs, alongside support for Microsoft Copilot, featuring dedicated Copilot physical keys on the keyboard. Consequently, devices lacking Copilot keys, such as Meteor Lake and AMD Ryzen (7040/8040) laptops, do not meet these standards.

Intel also proposed a simplified definition for AI PCs, advocating for each component—CPU, GPU, and NPU—to feature dedicated AI acceleration capabilities. This aligns with Intel's design philosophy, exemplified in products like Meteor Lake. Components can efficiently allocate AI tasks based on computational requirements, with NPUs excelling in low-intensity tasks such as photo, audio, and video processing, thereby enhancing performance and ensuring data security through on-device processing.

Furthermore, Intel emphasized the importance of memory capacity and speed for executing AI models, particularly large language models. Tasks may necessitate memory capacities of 16GB or even 32GB, albeit with associated cost implications. However, Microsoft has yet to define minimum memory requirements.

Looking ahead, Intel aims to deploy over 100 million AI PCs in client PCs by 2025. The company has forged partnerships with numerous software vendors, including Adobe, Zoom, and Autodesk, to develop over 300 AI-accelerated applications by the end of 2024.

In addition to these announcements, Intel introduced the ASUS NUC Pro 14 development kit featuring the Core-i Ultra Meteor Lake processor. This kit equips developers with tools, documentation, and training resources to harness Intel's technologies effectively for AI and machine learning applications. Pre-installed software tools include XPU drivers, Visual C redistributable, Cmake, Python, OpenCV, Git, VS Code, OpenVINO, and Chrome Canary. Intel's OpenVINO model zoo boasts over 280 optimized pretrained models, including 173 ONNX models and 150 Hugging Face models.

To facilitate knowledge dissemination, Intel has established a dedicated knowledge center on its official website. This resource offers AI PC-related training videos, documentation, and sample code, empowering developers to leverage AI PC technologies efficiently.