ippiDilateBorder_8u_C1R performing worse on faster Intel proccessor - IPP 2017 Upd 3

We are running into a strange issue with ippiDilate/Erode. It runs fine on multiple dev machines with varying vintage but exhibits huge timing numbers on our production PCs

1. Dev1- Intel i72700K - 2-3 ms OK,
2. Dev2 - Intel i7-8700 OK,
3. Production - Intel i9-7940X - Massive overtiming 30-40 ms for same function

We use other filters without issues on the same PC and timing numbers usually correlate with age of Intel chip speed.

OS - Windows 7, Win 10, IPP 2017 Update 3, static linking

Any ideas what might be causing this