When is predictions method in Core ML 2 effective?

Hi.

I tried predictions(from:options:) method in Core ML 2.

When I applied multiple inputs, it achieved 1.5 times to 2 times faster performance.

I also tried prediction(from:options:) method in mutlithreaded manner and got 2 times faster in the case of 2 inputs and 3 times faster in the case of 3 inputs.

I understood that it depends on capacity of computing resources.

But actually, In what case is predictions(from:options:) method in Core ML 2 more effective than multithreaded one?

Could you tell me your experiences?

Thanks.