The question pretty much says it all. Thoughts? Experiences?
Also, might as well through in CPU hyperthreading. I believe it’s recommended to turn off in the optimization guide (right?), but I’ve read different experiences and wanted to get more feedback. Thanks!
As far as I understand, it’s primary use is for simple tasks based on ‘deep learning’. The type of tasks that’s typically mentioned is noise suppression for audio (for example in video conferencing applications like Teams).
I don’t think it is of much benefit for the type of audio applications like GP. Probably too slow (my estimate) for processing 20 channels of CD-quality audio.
I don’t think it would harm the performance if is stays active: an application (or plugin) must explicitly use the api.
I also read GNA is going to be abandoned in favor of neural-VPU
My verdict: I Say not Yay, not Nay, but May 
That clears it up! Actually, I understand what you’re saying. Thanks for your input!