We performed all our overclocking experiments on a system built around Gigabyte GA-EX58-UD5 mainboard. We used Noctua NH-U12P with Enermax Magma UCMA12 fan (at 1500RPM) for CPU cooling. We tested system stability under heavy load with LinX 0.5.8 utility.
The first CPU we experimented with was Core i7-975 XE. Since it belongs to Extreme Edition series, it allows not only raising its clock frequency multiplier above the nominal value, but also lowering it if necessary. We used this particular feature during our overclocking experiments.
We managed to easily set the multiplier three steps higher, to 28x, without touching the default Vcore setting of 1.2V. The system remained absolutely stable and passed all tests. It means that our Core i7-975 XE processor with the nominal frequency of 3.33GHz proved capable of running at 3.73GHz at the default Vcore.
It is obviously not the maximum: the CPU temperature during our stability tests remained below 80°C, which indicates clearly that we can shoot higher if we raise the core voltage. That is why we increase the processor Vcore to 1.35V for our next experiment (0.15V above the nominal). As a result, the frequency potential of our processor went up and we could raise its clock multiplier by two more points.
The resulting clock speed in this case was 4.13GHz. The CPU passed all stability tests, but its temperature under load reached 98°C. Therefore, it is not surprising that we couldn’t get our system to retain stability with higher multiplier settings and failed to hit 4.26GHz clock with air-cooling alone.
Nevertheless, even 4.13GHz is a very good result, especially taking into account that the previous Extreme Edition processor we worked with, Core i7-965, could only get as far as 3.87GHz with an air-cooler. Today we pushed the frequency bar farther back and I believe the new processor stepping is the one to thank for it. However, to make sure that we were not just plain lucky and that the new processor stepping indeed boasts improved frequency potential, we also overclocked the second new CPU, Core i7-950, that is also based on the same D0 semiconductor die.
Core i7-950 is a much cheaper processor than Core i7-975, but their clock speeds differ only by 133MHz. The biggest difference that makes Core i7-950 look “less attractive” than its elder brother is the locked clock frequency multiplier. That is why the only way to overclock this processor is by playing with the base frequency of the clock generator and lowering the multipliers for QPI and memory frequencies. You can consult our Intel Core i7-920 Overclocking Guide for more details on Core i7 overclocking techniques.
The results of our overclocking experiments are the following. The Core i7-950 processor we had at our disposal remained stable at 3.8GHz frequency and default Vcore setting.
As you can see from the screenshot below, we had to increase the base clock generator frequency by 166MHz to hit the above mentioned clock speed. The processor temperature taken under multi-threaded load didn’t exceed 80°C.
Increasing the processor core voltage to 1.325V had a significant influence on the overclocking potential of our CPU, just like with Core i7-975. With this Vcore setting it was working stably under heavy load at 4.1GHz clock frequency.
Although, we have to admit that the CPU core temperature in this case got scarily close to the critical level, but nevertheless, we experienced no overheating problems.
Core i7 processors announced today differ from their predecessors not only by increased clock speeds but also by the processor stepping. At this point we can conclude that they boast higher frequency potential as well. Our overclocking experiments proved that these CPUs can surpass 4GHz with air-cooling. However, in this case it is important to remember that there is one critical problem typical of all Core i7 processors. These CPUs dissipate a lot of heat and it gets worse during overclocking, especially if you increase their core voltage. That is why insufficient cooling becomes the bottleneck that prevents us from achieving even greater results. In other words, had we used a more efficient cooler or a liquid-cooling system, the results could have been way better.