I am new to pc building and overclocking, but I will share my experiences so far.
Software
CPU Stress Tests: Prime95, Intel Burn Test, OCCT, Cinebench, 3dmark
Temperature Monitoring: OCCT (built-in), Real Temp, CPU-Z (for voltages, etc)
Stability Categories:
CPU Overclocking
I'm running an i5 3570k on Asrock Z77 Extreme-4 using a CM Hyper 212 EVO with MX-2 paste, inside a Rosewill Thor V2 case.
The highest clock I pushed my chip to that was able to boot Windows normally was 4.9Ghz, but it required a lot of increased voltage and the temperatures were hitting the TJmax of the chip under load. I have been able to get 4.8 and 4.7 Ghz to run stable under some tests, but temps were too high, so I'm going to keep backing it down and experiment to get it to what I think will be a safe overclock.
For me the process has been to primarily change three things: the multiplier, the Vcore, and the LLC level. Since Ivy Bridge is highly dependent on voltage for temperatures, I started by pushing my clock rate up 100Mhz at a time from stock until Windows wouldn't boot/BSOD, then I increased the voltage (using offset so that my chip could still idle at lower clocks) until I could boot to Windows. At each new clock rate (4.5-4.9) I tested using a 5 minute prime95 run to get an idea of temps, but I found out a problem with this later on. Read the last part of the post for an explanation. Since I started looking for stability, I have been running longer stress tests using prime95 and IBT to make sure the chip is stable. When it isn't, I increase the voltage by .005 volts at a time until it is. At some point I increased the LLC from Level 5 to Level 2 to add stability. I'm trying to keep temperatures below 85 C.
Precautions Before Stability Testing:
1. Double check your fans are working properly, especially your CPU fan. The CPU Fan controls of my bios only allow limited control, and take a few seconds to ramp up the fan speed when the cpu gets hot, which happens instantly when testing. So when I do stability testing I temporarily set the CPU fan to run on max for optimal cooling. I have also turned my chassis fans on max, but those don't seem to make a difference.
2. Different programs seem to stress the chip differently. My cpu has passed an Intel Burn Test (which is supposed to be really taxing) but failed a Prime95 test on the same settings.
3. Prime95 doesn't seem to really kick into high gear until ~16 minutes into the test, at least for me. Even though load was already 100% on the CPU for the first 15 minutes, at the 16 minute mark temperatures start skyrocketing--sometimes by almost 20 C. This makes makes me reconsider clocks that I thought were reasonably stable, temperature-wise.