There is no need for the -190 range of liquid nitrogen (which barely last a couple hours by the time the liquid evaporates).
You are incorrect there Lehpron, Ivy Bridge scales very very well with cold and voltage. Ivy for the most part has no cold bug, so it is really opened up when below 100C.
In that bold, I wasn't talking about Ivy Bridge, I was talking about the cryogenic liquid's ability to last.
That said, if JacobF can approach 6GHz within phase, then once someone reaches similar temperature range, that frequency range is within reach.
As for how Ivy scales at extreme low temps, I'm not too convinced we're looking at a capability of the process lithography or architecture. I detailed it out a while ago, but 32nm Westmere Clarkdales
have reached 7GHz before. Therefore, 32nm Sandy Bridge quads should have also been able to, but castrated due to Intel's 57x multiplier maximum and sensitive 100MHz strap, most had a hard time approaching 6GHz. All record runs of Ivy Bridge happen to have higher straps of around 110MHz (since clock generator is in the southbridge chipset) and Intel bothered to raise the max multi to 63x, therefore 110 x 63 = 6.9GHz-- making 7GHz a matter of raising the sensitive strap frequency.
In otherwords the supposed scalability is artificial (it's fake); we're only looking at Intel relaxing restrictions with Ivy versus Sandy.
I'm sure the Ivy Bridge processors are capable of 8GHz, but Intel won't let it happen.
It is odd that while LGA2011 can boot with the 125MHz strap at relatively lower frequencies that air/water can reach, not at much higher frequencies. At least all record runs keep the range of 100-105Mhz
while running into the max multiplier restriction. Of course sometimes I wonder, since the clock generator is in the southbridge, why not chill it too with the CPU? Then LGA2011 may finally reach 7GHz too if 125Mhz was stable (125 x 57 = 7.1GHz). Gulftown did it, so it isn't a TDP issue or lithography issue.
It's the southbridge; at least I'm certain every record run depends on its ability to raise beyond 100MHz and I have no doubt in my mind the processor can take the maximum multiplier.
<message edited by lehpron on Friday, June 29, 2012 5:11 PM>