CPU voltage questions

Jonathan Lemmens · June 14, 2016

The past few days I've been busy overclocking and stress testing an old "hobby-PC" and I've noticed something a little odd. The specs:

Xeon E5450 CPU (771 to 775 mod)

ASUS P5Q Motherboard

Brandless towercooler with one 120mm fan

4x2GB 800MHz Hynix DDR2 memory

80GB Intel SSD

550W Coolermaster V PSU

I'm using ROG Realbench for stress testing.

Before I started I disabled CPU spread spectrum, speedstep and C-state and set all voltages at their default manually. When I reached the max OC of 4.0GHz I had to push the voltage to 1.418V to pass an 8 hour stress test, temps where still well below 80C. Setting CPU voltage to 1.412V would fail after ~6 hours. By this time I also increased NB voltage to 1.28V and DRAM voltage to 2.02V. I proceeded to tune the RAM timings to 5-5-5-15, this was still stable for 8 hours. Thinking that I couldn't get CPU voltage any lower, I enabled CPU spread spectrum, speedstep and the C-states. This passed another 8 hour test. To be 100% sure I set the voltage back to 1.412V and tested again, this turned out to be stable, so did 1.406V and 1.400V, I got bold and tuned it back to 1.375V which passed a full 24 hour stress test. Right now I'm testing at 1.362V. All the time temps at load remained nearly identical at 67-72C, regardless of the voltage setting. Lowering the NB- or DRAM voltage will result in system instability, so does setting CPU voltage to 'auto'.

This left me with a few questions:

1. Is it because I enabled spread spectrum, etc. (or tuning the RAM) that I can suddenly be stable at lower voltages than before?

2. Temps aren't getting lower as I decrease CPU voltage, how can I be sure something isn't boosting the voltage to 1.418V (or beyond)? I know CPU-Z isn't very accurate but its Core Voltage reading is at least consistently ~0.012 off from my setting in the BIOS at every stage.

3. If I keep decreasing CPU voltage for as long as it is stable, could that somehow be harmful or be problematic in the future? Or is a lower voltage always better?