Overclocking: The Rules Have Changed
First off, we need to discuss overclocking the Sandy Bridge chips before we move onto the actual results because things are very different now from previous generation processor.
As far as overclocking goes now, Intel has changed the rules, plain and simple. There are some significant differences, and we’ll try to break these down succintly in order to best understand the methods and implications of what can be done (or not) in terms of overclocking.
The first thing to understand is that the locked CPU models have limited overclocking: 2400, 2500, and 2600. The K-models such as 2500K and 2600K can be overclocked significantly, and for a modest price increase, we recommend the K models over the regular chips.
The second thing to understand here is that the BCLK of 100 can be increased only slightly. Unlike the old days where a 450 FSB was possible on Socket 775, and a 220 BCLK was achievable on Socket 1156 and 1366, these Socket 1155 Sandy Bridge boards will only hit about 109 BLCK at most. The bus cannot be changed much at all, which is why the locked chips cannot be overclocked much due to their limited multiplier and minimal BCLK movement. Essentially, non-K CPUs will therefore have very limited overclocking, end of story. Since overclocking Sandy Bridge is almost reliant on the multiplier, the unlocked K-series chips can be pushed much farther even though the BCLK has little wiggle room.
The next major difference is that Turbo now comes into play very heavily. Even the base 2400 has a Turbo Mode that will inherently act like an overclocked situation. The Turbo will boost the multi on the chip when under load, like an "insta-OC". Linked to all this is that C1E, EIST, and Speedstep must be left ON. Unlike all previous generation processors that overclocked better with these turned off, that has now changed and these "speed adjusting" settings must be enabled.
Another critical point in overclocking Sandy Bridge is that all K-series chips have a multiplier wall. Previous generation Intel CPUs responded to increased voltage and lower temperatures to achieve higher clock speeds. That is no longer the case. A chip has a maximum multiplier it will achieve; additional voltage or lower temperatures will not improve the results. As mentioned, there is currently D1 and D2 stepping, and the D1 is preferred. Each chip will have an inherently different multiplier wall. Essentially this means that voltage, temperatures, and motherboard model will not make a difference in the maximum overclock of a particular chip; it all depends on the chip now.
Therefore, this also means that sub-zero cooling with dry ice or liquid nitrogen will no longer help improve overclocking results. In fact, Sandy Bridge chips will downclock and will lose performance scaling when very low temperatures are applied. As a result, air cooling is all that’s really necessary to hit the maximum overclock on a particular chip. Water cooling will help keep load temperatures lower, but will not produce a higher multiplier. We’ve learned that probably most 2500K CPUs will overclock to approximately 4.3GHz or so, while 2600K CPUs will be able to hit about 4.8 – 5.2GHz, depending on the chip. We’ve been told that the D1 stepping is pre-production, while retail chips will come with the D2 Stepping.
So, despite having an unlocked K-series CPU, you may reach the maximum mutliplier regardless of the cooling used or voltage applied. There is a bit of wiggle room to tweak the BLCK to squeeze a bit more out of a chip to hit the magical 5GHz if you have a cherry chip to begin with, but the days of juicing up a cheap chip to insane speeds are pretty much over.
Let’s take a look at our overclocking results now.