AI Discussions

[androidmarshmallow]

Hello,

I've tested the live captions feature that came with Windows 24H2 and it worked well to show captions while a youtube video plays.

I was curious to see how NPU usage changed while this AI feature was working but NPU didn't show up on task manager, so I installed the NPU driver from AMD and now NPU shows up both on task manager and device manager.

The problem is its usage always shows 0%. I read the NPU monitor is not yet fully implemented for AMD, only for some Intel processors. When will the AMD implementation get released? Any other way to see NPU usage?

Either live captions is not using NPU at all right now, or it uses NPU but the usage is not being monitored by Windows task manager.

[Uday_Das]

Youtube video caption not using NPU, I think.I am not sure which built-in application using NPU in your 8600G.  You may run any model directly on NPU using Ryzen AI software to see NPU activity. 

[Shanemike]

To monitor the NPU (Neural Processing Unit) usage on an AMD Ryzen 8600G (which is a part of AMD's Ryzen 7000 series with integrated graphics), you would typically follow these general steps:

1. Check System Performance Using AMD's Tools:

   • AMD Radeon Software: AMD’s official Radeon software allows you to monitor the usage of integrated graphics, but it might not provide detailed NPU stats if the 8600G has an NPU component integrated specifically for AI tasks.
   • AMD Ryzen Master: For detailed CPU and GPU information, you can use the Ryzen Master utility. It may give insights into performance and workload distribution, although it's primarily for CPU usage.

2. Using Third-Party Monitoring Tools:

   • HWMonitor: This tool can give you a good overview of the system’s hardware, including CPU and GPU usage. However, NPU-specific stats may not be directly available here unless the NPU usage is routed through the GPU.
   • GPU-Z: A more specialized tool for monitoring graphics card performance. If the NPU is part of the integrated GPU, you might be able to see general GPU usage, but specific NPU stats could be harder to find unless the NPU is actively being used for tasks.

3. For Machine Learning and AI Workloads:

   • If you’re running machine learning workloads on your system, tools like TensorFlow or PyTorch (if they support the hardware) may provide information on NPU usage.
   • OpenCL or Vulkan: If the 8600G integrates NPU with these APIs, you might be able to extract detailed stats using tools or frameworks that allow OpenCL or Vulkan debugging.

4. Check for Specific NPU Monitoring Tools:

   • If the 8600G’s NPU is part of a custom AMD architecture for AI workloads (like a mobile APU with machine learning acceleration), you might need to look for specific software or drivers from AMD that expose NPU usage. Sometimes these tools come with SDKs or special developer utilities that are only available for developers working with AI acceleration.

5. Command-Line Tools:

   • On Linux-based systems, tools like nvidia-smi (if you're using an NVIDIA GPU for AI work) or other platform-specific utilities could provide stats. However, these may not work on AMD hardware, so it’s worth checking if AMD has command-line tools for this purpose.

6. Look for Official Documentation:

   • The best way to get specific details on how to track NPU usage would be through AMD’s official documentation or community forums, as NPU support and monitoring can vary depending on the architecture and software environment.

So, to sum it up: if you're looking for NPU-specific usage stats on the 8600G, you'd need to check if AMD provides a monitoring tool for that specific component. Otherwise, general GPU and CPU monitoring tools may give you an indirect look at the load on the NPU, especially if it's shared with the GPU for AI processing.