Overclocking Experiments

Keeping in mind that Gigabyte GA-EX58-UD4P mainboard boasts very similar PCB layout and identical BIOS functionality as the higher-end models, I assumed that it would be enough to use the same settings that worked perfectly for Gigabyte GA-EX58-UD5 and GA-EX58-Extreme and we would achieve the desired results. However, our attempt to overclock the CPU to 3.95GHz ended with a blue screen of death after the first few seconds of stability tests in LinX and Prime95. The same happened when we attempted to overclock the CPU to a slightly lower frequency of 3.8GHz without increasing the processor Vcore. I understood that I wouldn’t be able to succeed that easily and would have to act in a traditional step-by-step manner.

Let’s leave the CPU in the nominal mode for now and try increasing the memory frequency to the nominal value for our Kingston HyperX KHX14900D3T1K3/3GX: 1867MHz. After another round of failures we discovered that the problem was the insufficient “QPI/VTT Voltage”. In case of higher-end mainboard models we had to increase it only to 1.335-1.355V, while Gigabyte GA-EX58-UD4P required an increase to 1.395V. Only after that our memory worked stably at high frequencies and we could easily overclock our processor to 3.8GHz frequency after corresponding increase in the memory frequency and lowering of its timing settings.

We overclocked without touching the processor core voltage. We only enabled “Load-Line Calibration” function that prevents the core voltage from dropping under heavy load and hence keeps all Intel processor power-saving technologies up and running in idle mode.

We assumed that then it would be enough to simply increase the processor Vcore in order to overclock it to the maximum frequency of 3.95GHz. However, all our attempts failed. I was subconsciously ready for this outcome, because I had to increase “QPI/VTT Voltage” more than before, which in its turn increases the CPU temperature. Well, looked like we might have to stop at 3.9GHz. But I was very surprised to find out that even this frequency remained unattainable and all my attempts end up in errors.

It took me a while to find out that the mainboard was setting the frequency of the North Bridge part integrated into the CPU way too high. I am talking about the parameter called “Uncore Frequency”. I have actually stopped paying attention to this frequency, because mainboards usually set it at twice the memory frequency most of the time. However, in our case it was always at 3720MHz for some reason, which corresponded to 1860MHz memory frequency no matter what the actual memory frequency was. We overclocked our processor to 3.95GHz by raising the base frequency to 188MHz. The memory in this case worked at 1504MHz with 7-7-7-20 timings. For this memory speed Uncore frequency should be set at only 3008MHz. Once we corrected this setting, our system worked stably with the CPU overclocked to 3.95GHz, which was confirmed by LinX and Prime95 stability tests.

However, we had to increase the processor Vcore to 1.3V to ensure stability at 3.95GHz. As a result, Intel processor power-saving technologies were partially off. Only the CPU clock frequency multiplier would go down in idle mode, while the voltage remained constant.

Well, despite certain difficulties our overclocking records obtained on top Gigabyte mainboards before have been successfully reproduced on Gigabyte GA-EX58-UD4P. Now all we need to do is find out how well the mainboard performs in these conditions and how much power it consumes in this case.