Windows 10: Microsoft Enables Nvidia GPU Support for Windows 11 Local Language Model APIs

Microsoft has updated the Windows 11 local AI documentation to confirm that language model APIs can now run on PCs that are not equipped with Copilot+ but have supported GPUs. Previously, these APIs were only available on Copilot+ PCs that had dedicated NPU hardware.

The change is currently aimed at developers and is not yet available to end users. Supported hardware includes Nvidia GeForce RTX 30 series GPUs and newer models with at least 6GB of VRAM.

What Changed and What This Enables on Non-Copilot+ PCs


Microsoft's updated developer documentation and a GitHub post describe a recent change. They state that language model APIs now run on GPU-equipped PCs that are not Copilot+ models, expanding local language model capabilities to more Windows 11 devices.

This represents a shift from Microsoft's original focus on Copilot+ PCs launched on June 18, 2024, which emphasized NPUs as necessary for local AI features in Windows.

While NPUs are not the only chips capable of AI workloads, Microsoft previously limited most built-in AI features to devices with NPUs.

The current feature set centers on text-based tasks using Windows.AI.Text APIs. These capabilities include summarizing content, rewriting text, converting text into structured formats, and generating prompts.

Although the functionality resembles cloud-based AI tools, it runs locally on the device, reducing dependence on cloud processing and keeping data on the machine.

At the core is a small on-device model called Phi Silica. Instead of being pre-installed on all systems, Phi Silica can be downloaded through Windows Update when needed by an application. Once installed, the model runs locally and makes use of the GPU when available.

What’s Still Limited and Why GPU Support Matters for Windows 11 Users


GPU support is currently limited to the language model API layer. Other Copilot+ features remain tied to systems equipped with NPU, including:

• Windows Recall,
• Click to Do, and several other AI integrations branded under Copilot+.

This capability is primarily accessible at the developer level rather than directly for end users. Running these APIs requires applications that are built or adapted to utilize the Windows AI framework.

Users will see benefits as developers update their applications to support GPU-based local AI on non-Copilot+ PCs.

The update expands the range of Windows 11 devices that can run local AI features. PCs with mid- to high-end Nvidia GPUs that previously lacked NPU hardware now have a way to access Microsoft's built-in local AI tools.

Many gaming computers and workstations meet this new GPU requirement, enabling local AI capabilities for a sizable portion of Windows 11 users.

Microsoft has not indicated whether GPU support will be extended to Windows Recall and other features currently limited to Copilot+ hardware, or if AMD and Intel GPUs will get similar support in the future.

Thank you for being a Ghacks reader. The post Microsoft Enables Nvidia GPU Support for Windows 11 Local Language Model APIs appeared first on gHacks.

read more...

 

Windows 10 CPU usage 100% when Nvidia GPU enabled

I've had the same problem, and have exactly the same GPU model. I was able to set the drivers back to factory default, but any update I've tried (including the most recent) causes the massive cpu usage. I've contacted tech support for both Nvidia and Asus,
and followed all the steps they suggested, with no luck.

 

NVIDIA enables GeForce GPU Passthrough for Windows Virtual Machines

Source: GeForce GPU Passthrough for Windows Virtual Machine (Beta) | NVIDIA

 

Microsoft DirectX 11 Details Emerge

Microsoft has released a handful of details about DirectX 11, the latest version of the company's API.

• Full support (including all DX11 hardware features) on Windows Vista as well as future versions of Windows
• Compatibility with DirectX 10 and 10.1 hardware, as well as support for new DirectX 11 hardware
• New compute shader technology that lays the groundwork for the GPU to be used for more than just 3D graphics, so that developers can take advantage of the graphics card as a parallel processor
• Multi-threaded resource handling that will allow games to better take advantage of multi-core machines
• Support for tessellation, which blurs the line between super high quality pre-rendered scenes and scenes rendered in real-time, allowing game developers to refine models to be smoother and more attractive when seen up close

Source: Shacknews