Any Advantage at 10C vs Ambient

Does this belong in the extreme thread?

I am thinking of going to an external water cooler - external to the case - with a Peltier cooler on the input to the case. I can put the radiator in the basement, etc.

I have the H80 and am at 64C package/core temperature at 100% load, with 30C
ambient. So far, I have not detected any clock modulation, by reading the MSR's.
The LGA 2011 BP 11D2B1 socket has a shiny 3/16" backing plate on the back side of the
motherboard. Looking at the back side of the MB, at 64C package temperature, the upper right corner of this plate is uncomfortable to touch. The other three corners are cooler. The right side is warmer than the left side. My IR thermometer gives an
absurdly low reading on this shiny metal plate!

If I reduce the coolant temperature to 10C at the inlet to the heatsink, will I see any
improvement in the ability to OC? I know I will have to insulate the inlet coolant pipe to prevent condensation.

I can not find a specification for the thermal resistance of the Core i7 3930K. Intel
only spec's this as a function of power dissipation, 0.18*power+43.4C. I see the core
temperatures and the package temperatures approximately the same, within 2 or 3
degrees C

What is the temperature drop across the headsink-case junction with good thermal
compound?

Tom Dean

If you reduce the ambient temp your load temps will go down by about the same amount.
On SB (and likely SB-E) that means more stability at given clocks, but not higher maximum clocks.



You will need a spectacularely large pelt module to do what you're talking about, and every watt the pelt draws is added to the amount of heat that the cooling system has to remove from the pelt.

Example: 125w CPU, you're going to need at least a 230w pelt at stock clocks and volts to get it decently cool. Now your cooling system has to get rid of 355w of heat.
Start overclocking that SB-E and you'll need a 350w-400w pelt, and be trying to remove over 600w of heat from the peltier module.


Short version: pelts are not practical on modern high end CPUs or GPUs.

Two things.

I can disable the clock modulation of the CPU and get mult x base clock. I can adjust the base clock.

When I say external cooling system, I mean all external. I don't need nice looking,
just good cooling.

The loop will be
from <reservoir> to <pump> to <heat exchanger> enter <case>
to <cpu heatsink> exit <case> to <radiator> back to <reservoir>.

I had an ascii sketch, but, the formatting of the forum jumbles it.

The radiator will be something with 4x120mm fans and be located in a sub-floor area
with ambient temperature of 25C.

The peltier will mount on the <heat exchanger> outside the case and has its own cooling system outside the case. I need to remove the heat of dissipating 20 to 30 watts from
the coolant. I am thinking of using the H80 I already have.

The intent of the peltier is to drop the coolant inlet to the case down to 10 degrees C.

Tom Dean

an h80 can't handle the pelt. problems with wattage capacity and surface area for heat exchange.

you have the right idea with using pelts remotely. but to cool them you would want a custom waterblock with large surface area. the pelts would be sandwiched between waterblocks in 2 loops. one for CPU, one for radiator.

I would use waterblocks big enough to sandwhich 3 200watt pelts. 2 pumps, one big rad on the hotside of the pelts, big fans. we are talking serious Rad and fans. would be a pretty sweet setup.

I am thinking of the peltier cooler as an addition to a water loop, not the primary
coolant. A large primary radiator to extract most of the heat produced by the CPU.
Get the coolant outlet from the radiator down to ambient + 10C. Large radiator,
and fans. Then, the peltier to reduce the coolant temperature down to ambient - 15C.
This would be a 25C improvement in coolant to the CPU heatsink.

Look at http://en.wikipedia.org/wiki/H2Ceramic_cooling and follow the link at the
bottom to the technical paper. Technical Paper on H2C Technology (PDF)

This would reduce the power needed for the peltier to something like 50W or less. At this power level, the H80 running full out, would handle the heat load, I think. The H80 handles my CPU at 4.16GHz and 64C.

Measuring the temperature of the metal backing plate on the back side of my motherboard with a type-k thermocouple and direct contact, I see 44 to 45 degrees C. Looking at the construction, 4 large screws to the corners of the LGA2011, I think the case is not very different. So, I have a 20 degree C temperature rise from case to Tcore. Does this seem reasonable?

If I needed more power, then, a compressor based cooler would be more efficient and
maybe about the same $ overall. (?)

Tom Dean

It would be easier and more practical to build a chiller unit.

The problem with a mix of radiator and peltier, is that its a bad idea.

if you put a radiator in "to help" on the cold side of the peltier, then the ambient air is working against the peltier as its temperature is warmer than the cold side of the pelt. then the peltier is trying to cool against the CPU heat and against the ambient heat.

better to go chiller or dual loop, you will get optimal temps by removing heat from the hotside of the peltiers.

if you use a radiator to help on the coldside, you get very inefficient, for a very marginal cooling improvement.

don't forget the power requirements input to the pelt, they are significant - you will need something special to power the pelt.

Please read all the thread.

I plan to use ALL the cooling I am talking about outside the case, powered independently from the case/motherboard,etc.

The radiator will not be on the cold side of the peltier. The peltier will be on the output side of the radiator.

Going around the loop,
First, the radiator removes as much heat as it can from the coolant, hopefully, down to ambient + 10C. Then, the pelitier removes additional heat from the coolant, hopefully, down to ambient - 15C. then, the coolant goes back into the cpu heatsink and back to the radiator. The pump and reservoir are between the radiator and the peltier.

The peltier has its own, independent cooling system on its hot side.

I am thinking of having two loops. The radiator will not be removing any heat generated by the peltier, that heat will be removed by the peltier cooling system, at first, an H80.

The radiator will not be on the cold side of the peltier. The peltier will be on the output side of the radiator.

Click to expand...

That is the cold side of the pelt. Will it improve temps? possibly a very small amount. You will not however remove 20c worth of heat from the liquid temp. As was pointed out earlier bringing the temps of your coolant below ambient you are now fighting the radiator rather than enjoying its benefits. You have to understand that water has a very high specific heat and while flowing the temp variance across your entire loop will be minimal.
There really is only 2 configurations in which pelts are even moderately effective. The first is direct die IE coldplate>>pelt>>waterblock. The second configuration is a dual loop setup, the first loop is cooling the pelts and uses radiators to dissipate the heat, the second loop is cooled entirely by pelts.
The first method provides the best CPU temp Delta but provides the greatest risk in the case of a failure. The second reduces that risk somewhat at the cost of overall cooling capacity and alot of extra electricity.

With the money you would sink into such an impractical unit you buy a air conditioner and build a nice chiller that would drop your temps significantly more than what your describing.

ssjwizard said:

That is the cold side of the pelt. Will it improve temps? possibly a very small amount. You will not however remove 20c worth of heat from the liquid temp.

Click to expand...

If I read this correctly, I should not expect little or no change in coolant temp across the peltier?

ssjwizard said:

As was pointed out earlier bringing the temps of your coolant below ambient you are now fighting the radiator rather than enjoying its benefits. You have to understand that water has a very high specific heat and while flowing the temp variance across your entire loop will be minimal.

Click to expand...

I expect the sum of temperature changes around the loop to be zero. Some are negative and some are positive.

I have some questions about measuring - will start a new thread.

thanks for the discussion.

you have some basics mixed up, so just continue thinking it thru before you start. we aren't trying to put down the idea, just trying to help you go about it in the best way.

putting the pelt in the same loop as the radiator and CPU will not work very well.

I have read the whole thread. one problem is the peltier wont drop loop temps by more than a few degrees if the cold side is on the same side as the radiator.

tomdean said:

Please read all the thread.

I plan to use ALL the cooling I am talking about outside the case, powered independently from the case/motherboard,etc.

The radiator will not be on the cold side of the peltier. The peltier will be on the output side of the radiator.

Click to expand...

It doesn't work that way.

Generally, the coolant temperature variation within a loop is VERY small. A few deg C, tops. In order to accomplish what you propose, you'd need to slow the coolant flow to an extremely slow speed, a level at which heat would not be removed from the CPU fast enough for water cooling to work effectively.

In short, your proposal in its current form doesn't work. Pelts are neat devices though, so just because design proposal 1 is a no go, doesn't mean the whole project is out of the question.

unless you have a series of pelts running @ around 5 to 6 volts through water heat exchanger plates connected in a series, then use a lower flow pump so the cooling effect is maximised, it is a delicate balance....high flows destroy peltiers cooling efficiency, and the fluilds have to be cooled enough so that the lowwer flows will still remove the heat....it can be a chemistry experiment to control...

I am doing it, but it is a highly modded marriage on two independent systems, with custom waterblock and pump, pump is on a 30w fan controller.

Not going to encourage you one way or the other, mine evolved over time, this is probably the fouth itineration of my cooling setup, I have been using it for around 4+ years already....it works great, just an electricity hog....

Go for it...

laterzzzzz......