I know it's hard to resist not trying to keep going faster and faster but at some point you will hardly notice any increase in performance. Squeezing out that extra 400mhz of speed comes at the expensive of higher voltage and possibly shorter CPU life. Are you trying to use the same voltage @ 4.4 that you are using at 4.0? You are going to need a significant increase in voltage to be stable on 4.4; may also be a bit risky on air cooling. Intels site states that max voltage for these chips is 1.3625v; on the box mine came in says max is 1.25v. If you find it absolutely necessary to run 4.4ghz, find a stable voltage for your 4.4 and compare benchmark results for 4.0 vs 4.4. You can then determine if the jump in voltage is worth the little increase in performance.
Agree, I ran mine at 3.8Ghz and benchmarks did not really indicate much in the way of a performance increase.
I had to crank the voltage a little to get to 3.8GHz whereas 3.6Ghz runs pretty cool, and with low voltages. For me, this seemed to be the sweet spot.
But to answer your question, the answer is: more voltage, more cooling.
3.6 is a good OC as it required no additional voltage (for me and for others I've read as well). Since I was able to achieve 4.0ghz and still be well below max voltage I figured to go with it. My cores idle @ 31C:31C, max temps under small fft load 55C:56C. Temps are with a tjmax of 95 (realtemp)
xwinx: You should use realtemp for your temp readings because it uses (by default) a 95 tjmax where as coretemp uses 105; Coretemp will give you temp readings 10C higher than actual. There is controversy over the actual TJmax for 45 nm chips but the writer of real temp has done much research to determine it is closer to 95 and not 105.
As for voltage readings, I would trust cpu-z more than coretemp.