In 2024, a significant number of Windows laptops have entered the market equipped with dedicated Neural Processing Units (NPUs), marking a pivotal shift in computing capabilities. Microsoft has been proactive in promoting the advantages of NPUs, which are designed to alleviate the processing burden from traditional CPUs and GPUs by handling AI workloads more efficiently. If you’ve recently acquired a Windows laptop featuring an Intel Core Ultra processor, Qualcomm Snapdragon processor, or an AMD 7040 series or later, you’re already harnessing the power of an NPU. However, the question remains: how can you fully leverage this technology?
Currently, the landscape of on-device AI computing on Windows is still in its infancy, as the number of applications utilizing NPUs remains limited. Major CPU manufacturers and Microsoft are essentially painting a picture of a future where dedicated AI hardware enhances the Windows experience. For those interested in training AI models, the GPU continues to reign supreme. Yet, when it comes to AI inferencing with pre-trained models, NPUs can offer a more power-efficient and faster alternative.
Windows Studio Effects
One of the standout features available to users is Windows Studio Effects, which does not require a Copilot+ PC. For example, the Dell XPS 14 9440, powered by an Intel Core Ultra 7 155H processor, supports this feature. Accessing it is straightforward—simply navigate to the control panel pop-up menu in Windows 11. If your laptop is equipped with an NPU, you’ll find the “Studio effects” toggle next to the Bluetooth button. This feature allows users to enable background blur, eye contact, or automatic framing during video calls. As you activate these effects, you’ll notice an uptick in NPU usage in Task Manager, confirming that the NPU is efficiently processing these enhancements in real-time. The beauty of this setup lies in its energy efficiency; the NPU consumes significantly less power compared to offloading these tasks to the CPU or GPU, ultimately preserving battery life.
Amuse 2.1
Amuse stands out as one of the few generative AI applications capable of performing text-to-image generation directly on-device, powered by an NPU. Utilizing the FLUX.1 Schnell model for image generation on AMD hardware, it also supports various Stable Diffusion models. Running Amuse 2.1 on an Asus Zenbook S 16 with an AMD Ryzen AI 9 HX370 processor showcases the built-in NPU’s capabilities. However, on the XPS 9440 with a Core Ultra (Series 1) processor, the NPU is not currently utilized, highlighting the need for Amuse to integrate support for Intel’s NPUs in future updates.
Select features in Adobe Premiere Pro
Adobe has been at the forefront of incorporating AI-driven features into its software, with Premiere Pro being a prime example. The software includes several functionalities such as Adobe Scene Edit Detection, Auto Reframe Sequence, and Text-Based Editing, all of which leverage Ryzen AI hardware for enhanced on-device processing. Additionally, Adobe recently unveiled a beta feature called Generative Extend during its Adobe Max annual conference. This innovative tool utilizes Adobe Firefly to generate additional seconds for video clips, filling gaps in editing timelines. While this processing is primarily cloud-based for now, there is potential for a streamlined, on-device version in the future.
GIMP with OpenVINO plugin
For those with an Intel Core Ultra CPU, GIMP offers an intriguing opportunity for text-to-image generation through the OpenVINO plugin. A comprehensive setup guide is available on GitHub for both Windows and Ubuntu users, though some technical expertise is required. During the installation, users can select the type of processing to offload to the NPU. The installation package includes all necessary files and drivers to ensure GIMP can detect the NPU on your PC, making it a rewarding weekend project for tech enthusiasts.
Plenty more on-device Windows AI features coming soon
For owners of Snapdragon-powered Windows laptops, a suite of features is already accessible, including live captions with translations on YouTube, Cocreator in the Paint app, and image enhancement options in the Photos app. Users with Intel or AMD laptops that are Copilot+ ready can expect these features to roll out through a software update in November. Looking ahead, Windows 11 is poised to unveil a wealth of additional AI capabilities by 2025, further unlocking the potential of NPUs and enriching the user experience.
