AI BriefingMicrosoft testing Phi Silica on Nvidia GPUs for Windows AI
Key point
Microsoft is testing the ability to run the Phi Silica model not only on Copilot+ PC's NPU but also on Nvidia RTX GPUs.
Details
Microsoft is expanding support for its small language model (SLM) Phi Silica to run on Nvidia RTX GPUs. Previously optimized for NPU (Neural Processing Unit)-based Windows Copilot+ PCs, Phi Silica will now also be usable in general GPU environments, with the goal of increasing local AI utilization.
Currently, this feature is provided at an experimental stage targeting developers. The following conditions are required to use the GPU path:
- Nvidia RTX 30 series or higher GPU (minimum 6GB VRAM required)
- Windows Insider Experimental Channel and Developer Mode enabled
- Windows App SDK 2.2.2-experimental9 or later version and the latest GPU drivers
This test is expected to open a path for local AI features to be used even on regular Windows 11 devices, not just Copilot+ PCs. However, NPU-exclusive features such as Prompt Compression or Speculative Decoding may still be limited when running on GPU.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.
Related
MS Paint and Photos Insert GUID-Based Invisible Watermarks into Locally Generated Images
NVIDIA Open-Sources Local AI Inference Router 'PAIR'
Microsoft AI CEO Criticizes Anthropic: 'Model Welfare' Training Is Dangerous