MS Ai Insider

Title: Revolutionizing AI on Windows: DirectML Expands NPU Support for Enhanced Performance on Copilot+ PCs

Posted by

ailona

–

August 29, 2024

1. ** **DirectML is enhancing AI capabilities on Windows by expanding support for NPUs, particularly on Copilot+ PCs powered by Qualcomm’s Snapdragon X Elite platform. This update allows developers to leverage local hardware acceleration for AI innovations, making it easier to port models and create efficient applications. Additionally, the WebNN Developer Preview enables web developers to harness NPU power for machine learning in browsers, promising a new era of AI-driven web experiences.2. ** in HTML:**

“`html

DirectML Expands NPU Support: A Game-Changer for Developers

DirectML is making waves in the tech community by expanding its support for Neural Processing Units (NPUs). This development is particularly exciting for developers working on AI innovations for Windows devices.

What’s New?

DirectML now supports Copilot+ PCs, which are powered by the Snapdragon® X Elite Compute Platform. This platform features Qualcomm® Hexagon NPU, delivering remarkable performance and energy efficiency.

“With Snapdragon X Elite, we introduced industry-leading NPU with 45 TOPS of AI performance at incredible power efficiency.”

This enhancement allows developers to create AI applications that can scale seamlessly across Windows devices.

Major Updates to DirectML

With the recent updates, developers can now leverage DirectML as a foundational component of the Windows Copilot Runtime. This means they can easily port models from GPU to NPU with minimal effort.

To get started, developers need to ensure they have the correct versions of DirectML, ONNX Runtime, and the Qualcomm Hexagon NPU driver. The minimum requirements include:

DirectML version 1.15.2 (ARM64)
ONNX Runtime version 1.18 (ARM64)
Windows 11, version 24H2 or newer
Qualcomm Hexagon NPU Driver version 30.0.31.250 or newer

What’s Important to Know?

While DirectML is rapidly expanding, some limitations exist. For instance, C++ developers must include DirectML 1.15.2 as a redistributable package within their applications. Additionally, NPU selection through the Python API for ONNX Runtime is not yet available but is on the way.

“DirectML uniquely simplifies how developers can scale their AI innovations, through a single, cross-hardware DirectX API.”

Moreover, developers may encounter error code DXGI_ERROR_UNSUPPORTED when using models outside the supported list. Testing models outside this list may lead to undefined behavior, so caution is advised.

WebNN Developer Preview: A New Frontier

The WebNN Developer Preview is another exciting addition. This release allows web developers to harness the power of NPUs on Windows devices. WebNN defines how to run machine learning models in the browser, paving the way for innovative AI applications.

As Intel Fellow Moh Haghighat stated, “Intel is excited that WebNN and DirectML can now provide countless web developers the ability to harness the power and efficiency of Intel’s NPU.”

This is just the beginning. Future updates promise broader model coverage, including generative AI models. Stay tuned for more developments in this rapidly evolving space!

“`

DirectML now supports NPUs on Copilot+ PCs, enhancing AI performance and efficiency.

Developers can easily port models from GPU to NPU with minimal effort using DirectML.

WebNN Developer Preview allows web developers to utilize NPUs for local AI processing.

DirectML simplifies access to various local accelerators across Windows devices.

Upcoming releases will expand model coverage and functionality for both DirectML and WebNN.

“`

From the Windows Blog