Open a support ticket. There's a flaw in how certain IBM JDK's that report inflated CPU usage to PegaPlatform. In the Pega Management daemon, every 120 seconds it gets "total CPU" usage from the JVM and "number of CPU's". The management daemon takes ((CPU usage @ end) - (CPU usage at start)) / ((time elapsed [normall 120]) * number of CPUs reported by OS    and sends it to PDC . 
On certain IBM JDK's the API is reports in nano-seconds when pega is expecting timing ms

There is at least hotfix to platform to use a different API to get CPU when running IBM JDK