How can temps be lower than ambient?

[davidm71]

Hi,

My idle temps for my E6600 are usually in the high teens or low twenties especially when ambient is just about the same. The processor is OC to 2700 with only one .01v increase at 1.335v. Under load it goes to 36-40 degrees after 10min priming. Coretemp, realtemp, everest all agree. The only thing is that the main cpu temp is always a few degrees warmer, when I thought it was suppose to be cooler than the core temps...

I have an Abit In9-32X. Anyone have experience with this motherboard?

Thank you.

 

[Mithent]

It's not possible to be cooler than ambient without some kind of refrigeration (e.g. Peltier cooling), whether you're air or water cooled - thermodynamics means that heat won't transfer from a cooler to a hotter place unless you put energy in. They could, of course, be equal to ambient, if the cooling is exceptional.

I'm not sure about what the issue could be though.. rather strange that the CPU temp is higher than the core temps. It rather sounds like the core temps are being reported as low by something along the line, but I don't know anything about that board.

 

[leetskeet]

i have an ac7 and coretemp always reports the same temps as my ac thermostat. i think the antec 900s big ass fan on top contributes to the low temps though more than the hsf.

 

[nomad8u]

You've got a pretty moderate OC and virtually no additional vcore, so it isn't going to generate much excess heat at idle, but it's basically impossible to be lower than ambient...?

Idle temps aren't nearly as important as loaded.

 

[BillParrish]

They can't with just fans.

Remember you are measuring temps a different locations with different methods and different software none of it calibrated with the exception of core temps which have a +- 1C error tolerance if the software is correctly coded.

so the question is


"main CPU temp"

Where is this coming from/what sensor and where is it, what software is used to read it and blah blah blah. If it is a motherboard sensor available in the bios and read by speedfan etc it is probably so wildly inaccurate as to be worthless for other than a general indication of internal case temps and if your cooling fails.

As you noted Coretemp Everest etc. all agree, thats because they are all reading the same information from a software register on the CPU that contains the data from the embedded Digital Temp Senor circuitry inside the CPU. Trust that, it is very accurate and is what the CPU itself uses to determine if it is too hot.

The best way to determine the amount of "thermal headroom" you have is to set the option in Coretemp to "show the Deg C to Tjunction". This will act like a countdown to "too hot" as the temps go up, the Deg C to Tjunction will go down, when Deg C to Tjunction reaches 0 the processor considers itself too hot and will issue an interrupt to spin all the fans up to 100% . Several deg C, no one knows for sure how many, I have seen Intel docs that indicate it is at least 10C, above the temp where CPU temp = Tjunction (Deg C to Tjunction = 0) the CPU will shut itself down to prevent damage.

 

[aethelwulf]

It's not possible to be cooler than ambient without some kind of refrigeration (e.g. Peltier cooling), whether you're air or water cooled - thermodynamics means that heat won't transfer from a cooler to a hotter place unless you put energy in. They could, of course, be equal to ambient, if the cooling is exceptional.

Peltier cooling is NOT a type of refrigeration. It is a "heat pump", and cannot be used on its own to cool, thus requiring either water/air or some other method to remove the heat.

They could, of course, be equal to ambient, if the cooling is exceptional.

That would be thermodynamically impossible. Heat can only be transferred when there is a temperature differential.

 

[Mithent]

Peltier cooling is NOT a type of refrigeration. It is a "heat pump", and cannot be used on its own to cool, thus requiring either water/air or some other method to remove the heat.

That's true, but while it's not itself a type of refrigeration, a heat pump is required to refrigerate, and Peltier cooling can be used to accomplish that heat pumping. I've not really looked into it myself though, I'm probably a bit inaccurate in my terminology.

That would be thermodynamically impossible. Heat can only be transferred when there is a temperature differential.

That was getting a bit too far into theoretical, really; that was kinda the idea I was going for, that the best you could possibly get would be for the CPU to be at the ambient temperature because heat wouldn't transfer if it were any cooler than that. It would have to be in equilibrium with the ambient temperature though, and that's not going to happen.