when you "shrink" architecture from 32nm to 22nm things get smaller. resistance
tends to increase. the more resistance, the more heat. same voltage/amperage
through a sized wire and decrease that same wire and keep the electrical load the
same, the smaller will always have a hotter temperature than its counterpart.
add the missed trick of Intel on the IHS thermal compound and now you have
more heat at the same voltage from sandy to ivey bridge. so a 1.45-1.5v on SB
is IB 1.3-1.35 to keep temperature levels in a tolerance.
But due to efficiency at lower clock speeds, and Ivy not needing as many volts as Sandy, at the lower clocks Ivy funs quite a lot cooler.
People make this out to be a huge problem however, when I really don't think it is.
On air cooling, or even H100i cooling, you can only really get to around 4.5-4.6ghz overclock anyway, at which, Ivy is cooler than Sandy, and performs better than Sandy, even if Sandy is running 4.7/4.8ghz. Therefore, to me, if you're running standard cooling, it really doesn't make a difference.
The only time I've properly noticed a problem with it was when trying to push more on water. I'm putting it mainly down to the thermal paste for this, but for me, the CPU block seemed incapable of absorbing the heat fast enough, so despite there being more than enough rad space available, I still couldn't really get above 4.8ghz, and at that clock, after half an hour of testing, the radiator itself didn't even get warm - regardless of whatever speed I set the fans to. That was the reason for me de-lidding the processor, but despite saving 5-10 degrees throughout, that still isn't really enough to push another 100mhz out of it.
I will try it with the liquid metal at some point to see if that helps any more.