Any downside from dropping the CPU voltages too much?

Alright so I have an i5 2500k CPU, a Asus P8Z68 V-PRO mobo, and a CM Hyper 212+ for cooling. To stress test I am using Prime95, CPU-Z, CPUID Hardware Monitor, and the sensor info on the AI suite too keep track of voltages and temperatures.

Now to overclock and adjust the voltages I am using the AI suite, that you can install from the asus website for their MOBO's. At first I used the auto tune feature on the program, and after 4.4GHz I would get temps above 70 degree celsius (like it would be a massive jump up to 70). Then I realized that the CPU voltage changed a lot, I noticed that past 4.4GHz the CPU was starting to pull in over 1.4 V which then kinda clued me in to what I was doing wrong.

So I've set my goal to overclock to 4.7 GHz, and it's pretty much all set. I've also successfully dropped my CPU voltage, so that if I stress test my CPU. It won't pull in more than 1.29V (rounded up number), which gives me maximum temperatures of ~63 degrees celsius.

But now I want to push the limits, and get the lowest possible stable voltages I can for this clock. But how will I know when I get there? and is there any risk from dropping the voltages too much?
3 answers Last reply
More about downside dropping voltages much
  1. If you use Prime95 to stress the CPU, it's easy to tell if you don't have enough CPU voltage -- your computer will reboot (with or without BSOD) or the program will generate errors. I recommend running Prime95 on the Blend setting for eight hours or more to test for a stable overclock. If it cannot run that, then it's not stable. I personally run Prime95 for 24 hours to test stability, but a lot of people don't have that kind of patience.

    Unless you got very very very lucky in the CPU lottery, 1.29v is unlikely to be stable at 4.7GHz overclock. I hope I'm wrong though.

    To keep a Sandy Bridge CPU happy for a long-term overclock:
    1. Keep the peak Prime95 core temps below 75ºC
    2. Keep the CPU voltage below 1.4v
  2. Most people do not have that kind of patience. I'm a little more fanatical about stability. I run P95 small fft's to test the CPU and the blend test's for the memory. For a K chip, I would recommend using the small fft's test.

    With an unlocked multiplier, the memory subsystem should not be a problem.
  3. Quote:
    I'm a little more fanatical about stability.

    Good to know I'm not the only one. I actually run 24 hours of Small FFTs, 24 hours of In-place large FFTs, and 24 hours of Blend to test final stability.
Ask a new question

Read More

Intel Stress Test CPUs Overclocking