Would very much like to be able to see the NPU on my laptop and desktop systems being used by PL alongside the GPU during processing and export. That little chip module represents about 25% of the processing potential of modern CPUs.
On my Tower NPU is used correctly by DxO PL.
with Intel chips, Select the “AI Boost” rather than GPU and the NPU will allocate AI tasks between the NPU and the GPU on your machine
D
Im not an export on that, but its may “just” decide, what acceleration is preferred? Not unusual, the AI provider give a choice for the Application: what do you want to use? Most fastest (for that workload) or the most energy efficient, or just NPU, and so on.
In ONNX at least the followings :
| OrtExecutionProviderDevicePolicy_DEFAULT | |
|---|---|
| OrtExecutionProviderDevicePolicy_PREFER_CPU | |
| OrtExecutionProviderDevicePolicy_PREFER_NPU | |
| OrtExecutionProviderDevicePolicy_PREFER_GPU | |
| OrtExecutionProviderDevicePolicy_MAX_PERFORMANCE | |
| OrtExecutionProviderDevicePolicy_MAX_EFFICIENCY | |
| OrtExecutionProviderDevicePolicy_MIN_OVERALL_POWER |
Check out this previous topic for more info :-