The voltage VID may be a fixed quantity from the factory. It can be difficult to sort different voltage and temperature readouts because different reporting software will use different terminology. Most of us would recommend HWMonitor64.exe over HWMonitor. It gives a lot more information.
The fact that you took the CPU Core/Cache Voltage off "Auto" and manually assigned it a value is what I was really getting at. Uncle "Auto" will typically assign more voltage to the cores/cache than may be necessary to keep the system stable and so drive up temps. What you need to do now is experiment with that voltage and stress test until you find the lowest safe voltage that will keep the processor stable at the highest possible frequency without generating excessive temps, which I would limit to 95c. Higher than 95c will put you into throttle range anyway and you should leave some room for changes in ambient temps related to weather if necessary. Keep in mind that every degree of ambient temp change will cause the same amount of core temp change since cooling, even with a water cooler, is still accomplished by the delta between radiator temp and air temp flowing across the fins.
I would suggest using Realtemp for the stress test and set the amount of memory used to half of what your system has. Realtemp uses both the GPU and the CPU for rendering ans so simulates real world computing. It will generate higher internal case temps than most other stress tests because of that. Test for a couple of hours.