Wait a minute -- Usually its the other way around ... cool off IC's/CPUs and they function better...
The antminer chips NEED heat??? Odd.
I know with the S2's, the hotter they ran, the more current they sucked. This summer the landlord at the office I was renting shut off the AC and the miners were running with 90+ degree F. intake air which made the chips cook at 75 C. to 80 C. I didn't notice any better hashrate with them.
What I'm about to say is only speculation:
I'm pretty sure the fans will draw more current as the temperature rises. When the fans turn at higher RPM's it results in more current draw.
As for higher hash rates and lower HW errors; this one blows my mind as well. I know in the wire-line telephone industry, the amount of resistance we have in the cable per foot changes with temperature. The cooler the temperature [Due to contraction - getting smaller in diameter] the wire would have more resistance per foot. The warmer the temperature [Due to expansion - getting larger in diameter] the wire would have less resistance per foot.
I'm sure it probably gets more deep than just the resistance changing with temperature. That's why I said someone with more experience in electronics might be able to elaborate on the subject.
The only thing I can "speculate" on is with the warmer temperatures and expansion of material and components in the blades possibly resulting in less resistance to impede transmission. However, there is a breaking point at how hot is too hot where the material and components can break down. I'm sure others with more experience would be able to elaborate on this with a more informative explanation than mine. I'm only speculating. That's it...
I know manufacturers of ASIC's have implemented circuits that are programed to throttle down an ASIC once it gets to a certain temperature and even shut the ASIC down once it gets to an even higher temperature to avoid potential meltdown and/or fire.
Edit: Once the rigs get above a certain temperature and potentially that creates even more expansion of materials that could be enough to affect the capacitance in the blades. For example, the two copper wires with a poly-etholine coating on them in the wire-line telephone industry are twisted together. These two wires are in essence a capacitor, which is at least 2 metal plates separated by an insulator(s). Three things that affect the value of a capacitor is the size of the metal plates, distance between the plates and the material between the plates. As the capacitance increases [Due to expansion - larger metal plates] higher frequencies are affected by this increase in capacitance. I will not elaborate on how and why [Taking up too much space in the forum] but it does affect them.
So, the only thing I can speculate on is the resistance in the materials used in the blades decreases with temperature rise because of expansion. This reduced resistance increases the hash rate and reduces the hardware errors. However, if the temperature continues to rise to a point of creating considerable expansion of those materials to change the capacitance, it can affect the impedance in the blades so much that it reduces the hash rate and increases the HWE's. Again, this was only speculation.
I know hybrids in carriers in the telephone industry have to be a certain impedance: 640 ohms for non-loaded pairs and 900 ohms for loaded pairs. If we ever have a customer's line not loaded properly, it can change the impedance on the line [Making it not balanced with the hybrid in the carrier] and can potentially create echo-return-loss or reflection in the signal in certain situations. That could be what is occurring when the temperature gets too radically hot above the high 60's or 70's; creating echo return loss or reflection of the transmission between components [because of a change in capacitance], which could increase errors and reduce hash rate.
That was my two cents. It was given to encourage discussion and thought on the subject. I promise, I will not curse out anyone this time if anyone was to chime in.
