Our CPU overclocking section will focus on the performance of the i9-12900K. We have fat last found a simpler way to get dry ice. So we can finally offer you the tutorial on how to use it in overclocking. In the next few days, we will be testing the i9-12900K, i5-12400 and i3-12100F under dry ice. One of the pots used will be the Copper ECO Pot from Bartxstore.
Before starting, it is necessary to have one or more reference scores with our i9-12900K at stock frequencies, i.e., without having made any modifications other than having activated the XMP profile. It is also very important to have a sheet of paper at hand that will allow you to note all the tests you are going to perform as well as the results or failures. Personally, I have hundreds of sheets of my tests on which I sometimes go back to find out how the processor behaved on a particular motherboard or in a particular test condition. My own little bible. For ‘daily’ overclocking, it will of course be necessary to take more time to fine-tune the voltages in order to provide it with the minimum necessary, but also to adjust each core independently if you want to take advantage of maximum performance.
Load-Line Calibration :
A quick tour through the « Load-Line Calibration » option to fine-tune this parameter to keep the voltage as constant as possible. The idea is not to have too much difference between the chosen voltage, the voltage measured at idle and under load. In the case of our Z690 Unify-X, we opted for mode 5.
For information, mode 3, when our voltage is set in the BIOS to 1.20 volts, gives us in reality a frequency of 1.22 volts in bench. Mode 5 allows us to keep 1.20 volts and mode 7 drops it to 1.12 volts.
Overclocking under watercooling in different steps:
As usual, I will start the tests by choosing, via the BIOS, a starting frequency applied to the 8 cores of the i9-12900K. I have chosen a starting frequency of 5000 MHz with a multiplier of 50, a BCLK of 100, a voltage of 1.20 volts and LLC mode 5. All other options are set to AUTO except for XMP enabled.
The idea is then, in the OS, to test the stability on several runs of Cinebench R15 multithread. If it’s stable, I increase the frequency by 100 MHz while rechecking the stability. If the benchmark crashes, I increase the voltage in 0.05 volt increments in order to regain stability allowing the benchmark to complete. The objective is to get a « quick » idea of the maximum benchmarkable frequency.
It is also very important to have a benchmark score under Cinebench R15, taking care to note the voltage (vcore), the maximum temperature during the benchmark, the power consumption and of course the score obtained. During the benchmark, the Core Temp window is opened which has a slight negative impact on the score. Here are our different results.
The limit for Cinebench is around 5.3 GHz on all eight cores. It will be possible to go higher in frequency on benchmarks that make less demands on the CPU. As you can see, it’s pretty hot. We go from 4160 to 4398 points under R15. The score under R23 is 29362 points.
A quick word about the cooling system that we changed during the tests. We took the opportunity to test the new EK-Quantum Velocity² waterblock. This one allowed us to lose/gain almost 7 degrees! If you want to know more about it, I invite you to read this mini test.