Water blocks for amd 64

[Since87]

Originally posted by Brians256
Is any of this making sense?

Just trying to learn.

It's not making much sense.

Try thinking through it again with a fixed quantity of heat being transferred. That heat being equal to the electrical energy dissipated in the die. (In actuality there will be heat drawn out of the motherboard, air, etc, but it's probably best to neglect these for the first pass at trying to understand what's going on.)

Instead of the amount of heat changing, think in terms of how temperatures need to change in order to move that fixed amount of heat.

[Brians256]

I actually was assuming a fixed heat load. I was trying to work with a fixed amount of heat being dumped into the cold side of the TEC and being moved out the hot side.

When you change the mass flow rate (and ignore things like laminar flow and such) the only thing that changes is the dT at both TEC and CPU.

The question is why a TEC would move more heat when the dT between hot and cold sides of the TEC is larger. I was under the impression that a TEC is moves fewer BTUs when dT(TEC) is larger.

[cristoff]

You know I was thinking.... hehe happens sometimes ....

You have two options...

High Flow which will create the copper to become cold by the water giving up its cold to the copper... In this you are overcoming the energy that the cpu is giving, very little movement of the energy through the water (I think at least)...

Low Flow (impingement), which will create the copper to give up its heat to the water, in making the water warm... In this you are more or less moving the energy, absorbing and moving (block doesnt become as cold)....

Unfortunately my theory makes no sense now that I think about it... The longer the water is in if it is cold enough to overcome the heat by a certain amount of X and then stays to absorb it for on a certain amount of time Y.... hmmm.....

Now in each you have energy exchanging within itself always creates heat, which is very, very minimal here...

Hmmm..... just thinking outloud...

Its interesting though from what I have read about the High Flow vs Low Flow (Impingement) that High flow doesnt need a very good radiator and a Low Flow does.... (correct me if I am wrong please)

Well guys tear it up hehehe... we need to get to the bottom of this and I am sure we can... we just need to find the formula's to use...

Someone help me out with figuring heat exchange for the elements... copper and water....paweese! Thanks!

later guys..

[superart]

My theory is based upon steady state situation, where the goal is to move as much heat energy as possible out of the water. The goal is not to get the lowest temperature water.

Aren't lower temps an indication of removing heat energy? Are you saying you can remove heat energy without making the water colder? Am I reading you right?

[Brians256]

Aren't lower temps an indication of removing heat energy? Are you saying you can remove heat energy without making the water colder? Am I reading you right?

Absolutely not what I mean! I must have explained poorly.

Removing the same number of BTUs/Calories from different amounts of water will yield a different temperature.

If 10g of water are moving through the chiller per second, then moving 10 calories of heat per second will lower the outgoing water temperature by 0.001C (check my figures on this calculation please!). Moving 1g of water through the same theoretical chiller would lower the outgoing water temp by 0.01C.

Same amount of heat moved per second. Different amount of water being chilled per second.

[Cathar]

I must admit that I'm slightly perplexed about people talking about using ultra-low flow rates through a TEC chiller.

It doesn't make sense to me.

TEC's are more efficient when their delta T between Tc and Th (cold and hot sides of the TEC) is low.

Now, as many waterblock tests have shown, block performance (measured as C/W) improves as the flow rate goes up.

This means that the dT within the TEC is lower, because the cold side is warmer, thereby raising its effiency. Even though it may seem somewhat counter-intuitive at first glance, the cold side being warmer is actually drawing more heat out of the water due to the increase in TEC efficiency.

When one wants to build a TEC based chiller, we want to follow exactly the same principles as for cooling the hot side of the TEC. We want the cold side to be kept warmer than with a less efficient block. TEC dT will therefore drop, the TEC efficiency will go up, and more watts of heat will be pumped from the cold side to the hot side. Now that wattage is coming out of the water, hence the water will get colder, faster.

Remember too, we're talking about closed loop systems, so it matter little what the entry/exit temperatures of the TEC chiller is. Over time, more watts are being drawn out of a fixed quantity of water than with a less efficient cold-side setup utilising ultra-low flow rates.

[Since87]

Bill,

By any chance, can you tell us the dT between; chiller copper at the inlet barb, and chiller copper at the outlet barb?

(And the conditions under which that dT was measured)

Edit:

By "conditions" I mean heatload and chilled water flowrate specifically.

[gone_fishin]

Originally posted by Cathar
I must admit that I'm slightly perplexed about people talking about using ultra-low flow rates through a TEC chiller.

It doesn't make sense to me.

TEC's are more efficient when their delta T between Tc and Th (cold and hot sides of the TEC) is low.

Now, as many waterblock tests have shown, block performance (measured as C/W) improves as the flow rate goes up.

This means that the dT within the TEC is lower, because the cold side is warmer, thereby raising its effiency. Even though it may seem somewhat counter-intuitive at first glance, the cold side being warmer is actually drawing more heat out of the water due to the increase in TEC efficiency.

When one wants to build a TEC based chiller, we want to follow exactly the same principles as for cooling the hot side of the TEC. We want the cold side to be kept warmer than with a less efficient block. TEC dT will therefore drop, the TEC efficiency will go up, and more watts of heat will be pumped from the cold side to the hot side. Now that wattage is coming out of the water, hence the water will get colder, faster.

Remember too, we're talking about closed loop systems, so it matter little what the entry/exit temperatures of the TEC chiller is. Over time, more watts are being drawn out of a fixed quantity of water than with a less efficient cold-side setup utilising ultra-low flow rates.

A tec will be operating at X efficiency using the hot side cooling, insulation, voltage applied as adjustable variables. This is the tec itself, not the efficiency of the cooling chamber. The cooling chamber is actually the applied load to the tec cold side (heat stored in the cooling chamber block). Slowing the flowrate will increase the load seen by the tec.

Is this not the same thing you said quote "the cold side being warmer is actually drawing more heat out of the water due to the increase in TEC efficiency.
" end quote

But the tec is not drawing heat out of the water, it is drawing it out of the heat stored in the coldplate. Slowing the flowrate increases the stored heat in the coldplate. In a die waterblock you want this stored heat to be minimal for a low die temp but in a chiller you want this stored heat to be high in the coldplate.

This is my logical understanding, please explain why I am off or on track.

[Cathar]

gone_fishin, allow me to dissect your post if you will:

A tec will be operating at X efficiency using the hot side cooling, insulation, voltage applied as adjustable variables.

You also forgot to add that the efficiency depends on Th - Tc, being the temperature difference between the hot and cold sides of the TEC.

The greater the temp differental, the more the heat wants to flow back from the hot side to the cold side by natural forces of conduction as the TEC elements are essentially made out of metal.

You are right when you say that the TEC's efficiency is affected by hot-side cooling, but to say so is only stating half of the issue. The efficiency, being watts moved, is really affected by Th - Tc. The minimum temperature achievable by the cold side is a side-effect of keeping the hot side cooler. This is a desirable thing, but really what's going on is that we've lowered Th, and Th-Tc remains essentially the same (given perfect insulation and identical heat loads).

Slowing the flowrate increases the stored heat in the coldplate.

This one point hinges upon most of the others you made. The cold-plate is just a thermal buffer, in equilibrium the "stored heat" of it is essentially irrelevant. It does provide a thermal gradient though between the water/chamber and the TEC cold side.

Through increasing the water flow rate on the cold-side/cold-plate, we are increasing the TEC's cold-side temperature (because the water is always warmer than the TEC cold-side) due to an increase the thermal convectional transfer rate. In doing so, we raise Tc, and hence Th-Tc gets smaller, thereby improving the heat pumping efficiency of the TEC. Heat pumping, measured in watts, goes up, therefore there is an increased flow of heat energy moving from the water, to the cold side of the TEC (via whatever is sitting in between).

Now I can't be sure if this isn't exactly what you were already thinking or not. It could very well be, just that I disagree with some of the terminology used.

[gone_fishin]

Originally posted by Cathar
gone_fishin, allow me to dissect your post if you will:



You also forgot to add that the efficiency depends on Th - Tc, being the temperature difference between the hot and cold sides of the TEC.

The greater the temp differental, the more the heat wants to flow back from the hot side to the cold side by natural forces of conduction as the TEC elements are essentially made out of metal.

You are right when you say that the TEC's efficiency is affected by hot-side cooling, but to say so is only stating half of the issue. The efficiency, being watts moved, is really affected by Th - Tc. The minimum temperature achievable by the cold side is a side-effect of keeping the hot side cooler. This is a desirable thing, but really what's going on is that we've lowered Th, and Th-Tc remains essentially the same (given perfect insulation and identical heat loads).



This one point hinges upon most of the others you made. The cold-plate is just a thermal buffer, in equilibrium the "stored heat" of it is essentially irrelevant. It does provide a thermal gradient though between the water/chamber and the TEC cold side.

Through increasing the water flow rate on the cold-side/cold-plate, we are increasing the TEC's cold-side temperature (because the water is always warmer than the TEC cold-side) due to an increase the thermal convectional transfer rate. In doing so, we raise Tc, and hence Th-Tc gets smaller, thereby improving the heat pumping efficiency of the TEC. Heat pumping, measured in watts, goes up, therefore there is an increased flow of heat energy moving from the water, to the cold side of the TEC (via whatever is sitting in between).

Now I can't be sure if this isn't exactly what you were already thinking or not. It could very well be, just that I disagree with some of the terminology used.

How can you say this stored heat is irrelevent? In simplistic terms it is the local heatload that the tec is seeing and moving. By saying it is moving the heat directly from the water is the same as moving further along the source path and saying the tec is removing the heat from the die. Also the heat being added by the die does not change, but the amount passing through the coldplate per unit of time can. The hotter coldplate would satisfy your argument of Th-Tc increasing heatpumping efficiency. Now how do you get a hotter coldplate? By lowering the flowrate.

[Gulp35]

Through increasing the water flow rate on the cold-side/cold-plate, we are increasing the TEC's cold-side temperature (because the water is always warmer than the TEC cold-side) due to an increase the thermal convectional transfer rate.

This statement changed my mind about this issue.

Originally, when a low flow block was mentioned i thought of some thing like the maze 4 ( confusing low flow and low resistance), so in a block like this there isn't as much turbulent flow as laminar and convective heat transfer. Therefore raising the flow would not cause such a great change in performance ( steeper (sp?) initial d(c/w)/d(flow) ).

But in a block like the cascade or WW performance is better with higher velocity,sacrificing of higher possible flow. So If you used something like the Cascade Peltier sized to warm the cold plate wouldn't that work better?

So in the end isn't this just another Flowrate disscussion ( note: that argument was ended saying more flow is always better ( as mentioned earlier)).

[HMB]

The only thing that i can think of that is good with the low flow of the chilled water is if you gain more by the colder water than you gain with a higher flow that will lead to a more efficient heat removal from the die. But then the difference of the water entering/exiting the chilled loop of water has to be pretty high unless the WB is extremely well suited for lowflow/high pressure? :shrug: I think im way out of track here.

Edit: Compare situation 1 and 2:

#1 Heatload 50W, HS+Fan 0.5c/w, ambient temp 20 degrees C.
Heatload temp = 45 degrees celsius.

#2 Heatload 50W, HS+Fan 0.6c/w, ambient temp 10 degrees C.
Heatload temp = 40 degrees celsius.

Thats just a crude example of what im thinking atm, im probably completly lost.

[gone_fishin]

Originally posted by Gulp35
This statement changed my mind about this issue.

Originally, when a low flow block was mentioned i thought of some thing like the maze 4 ( confusing low flow and low resistance), so in a block like this there isn't as much turbulent flow as laminar and convective heat transfer. Therefore raising the flow would not cause such a great change in performance ( steeper (sp?) initial d(c/w)/d(flow) ).

But in a block like the cascade or WW performance is better with higher velocity,sacrificing of higher possible flow. So If you used something like the Cascade Peltier sized to warm the cold plate wouldn't that work better?

So in the end isn't this just another Flowrate disscussion ( note: that argument was ended saying more flow is always better ( as mentioned earlier)).

This is a different situation altogether. In a straight waterblock loop, the stored energy in the bp at equalibrium will determine die temp. Higher flow lowers this stored energy at equalibrium, effecting die temp accordingly.
Here we are arguing the flowrate for best chiller performance. The reverse is true in this situation. You want higher stored thermal energy in the cold plate at equalibrium for the detaT differential Cathar pointed out to raise efficiency of the tec. Lower flowrate does this.

How this effects the whole system loop is another discussion. To me it seems clear that this approach will enable extremely low flows with the use of a tec to equal or better the use of a high flow pump without a tec. BillA's showing was of a unit that has dual tecs and dual coldplates and has shown data that looks promising (below ambient die temps).

[Since87]

Originally posted by gone_fishin
This is a different situation altogether. In a straight waterblock loop, the stored energy in the bp at equalibrium will determine die temp. Higher flow lowers this stored energy at equalibrium, effecting die temp accordingly.

This is certainly true. However, there is little, if any, benefit to being concerned with the quantity of the stored energy. As Cathar pointed out, the relevant information is inherent in knowing the temperature gradient. Limiting the discussion to the amount of energy transfer per unit time (power) and temperature gradients covers the relevant issues and IMO reduces confusion.

Originally posted by gone_fishin
You want higher stored thermal energy in the cold plate at equalibrium for the detaT differential Cathar pointed out to raise efficiency of the tec. Lower flowrate does this.

The "efficiency" of the TEC is only relevant with a context of hotside temperature and desired coldside temperature.

Why use a TEC and two TIM joints to acheive 'max TEC efficiency' and dT of zero? Discussing "efficiency" of a TEC without reference to specific hot side and coldside temperatures is fairly meaningless IMO.

[Since87]

Originally posted by Cathar
You are right when you say that the TEC's efficiency is affected by hot-side cooling, but to say so is only stating half of the issue. The efficiency, being watts moved, is really affected by Th - Tc. The minimum temperature achievable by the cold side is a side-effect of keeping the hot side cooler. This is a desirable thing, but really what's going on is that we've lowered Th, and Th-Tc remains essentially the same (given perfect insulation and identical heat loads).

Actually Th-Tc increases as hotside temperature increases.

Some Kryotherm calcs with nothing changing aside from the ambient temperature input, yielded: (172 Watt TEC, 50 Watt heatload.)

Ambient = 23C, Th-Tc = 42.6C
Ambient = 33C, Th-Tc = 45.0C
Ambient = 43C, Th-Tc = 47.8C

TEC's are somewhat 'forgiving' of hotside temperature.

These Kryotherm results are also consistent with other literature on TEC operation.

Edit: Changed "53C" to "43"C.

[Gulp35]

We are trying to transfer heat to the baseplate to warm the tec( cold side). Using jet impingement rasies the Heat transfer value between the water and copper. So, since the water will always be warmer than the TEC's cold side ( as stated) the heat will try to go to the area that is colder (right?). So by increaseing the heat transfer you also increase the temp of the bp.

Also as an explanation of why it is always harder to cool something: Because of the laws of thermodynamics ( specificlly the 2nd i think) you will always make more heat than you remove causing more heat for you to remove. ( I think this statement is correct, my Chem teacher said some thing like this during our unit on thermodynamics I just exterpolated a little.
)

[pHaestus]

I am going to change the topic here somewhat. If flow rate is reduced below a certain point, would water not freeze at the pins of the chiller? That seems pretty obviously a bad thing. But what if we are near that point and with laminar flow overall? Does that imply that there is a substantial amount of water that is staying in the chiller for longer times due to the velocity profile?

->
-->
--->
------->
--------->
------->
--->
-->
->

That water near the pins would get colder and would then serve as another buffer between the cold chiller and the warmed inlet water? And I suppose if residence time near the pins of the chiller was long enough then you could effectively decrease the volume of water that is actually circulating.

Could be irrelevant or even incorrect but it is a thought I just had.

I am still having an issue with how the change in peltier efficiency would offset dropping from turbulent to laminar flow regimes inside the CPU block anyway...

[gone_fishin]

There is an important DeltaT that is not being considered here. That is the differential between coldplate/heatexchanger temp and temp of the water in contact with the heat exchanger. Greater efficiency in thermal transfer would coincide with a higher DeltaT here would it not?

This would mean that tec power would have to increase with increased flowrate to maintain an equal deltaT here when changing flowrate. If flowrate were reduced then the opposite would be true - tec power would have to be reduced to maintain an equal DeltaT.

Does this shed light on the benefit of a lower flowrate in a tec chiller?

Remember the coldness of the coldplate is not cooling the water, the heat is traveling through the coldplate and on up through the tec.

[Les]

Originally posted by unregistered

some results:
Watts . coolant . die
39.4 ..... -1.9 ..... 8.1
49.8 ..... -1.0 ... 10.9
59.6 ...... 0.4 .... 14.1
69.9 ...... 1.8 .... 17.6
79.6 ...... 3.3 .... 21.2

steady state assumed, but may not be absolutely correct

Said I would play.
Not got far, but for what it's worth:-


and an unencouraging comparison of "Effect of Coolant Temp on C/W" with some of my predictions :-


EDIT Using info(Cold-side Flow rate=2.417 lpm) from Switech Site* and raw MCW5002 test data have up-dated 2nd graph to include 25c result.
* http://www.swiftnets.com/products/mc...sp#description

[Althornin]

Originally posted by gone_fishin
Now how do you get a hotter coldplate? By lowering the flowrate.

not at all!
Lowered flowrate results in a COOLER coldplate, as far as i can see.

[BillA]

aiiii
I think the words on this page are cavitating
sorry but I've not time to catch up

"Lowered flowrate results in a COOLER coldplate", are we getting closer ?

Les, help me here
give me a (frothy ?) verbal description of that second graph
(the first I would hazard represents 'reaL world' efficiency, as my uninsulated system was setup)

[gone_fishin]

Originally posted by Althornin
not at all!
Lowered flowrate results in a COOLER coldplate, as far as i can see.

Please explain "the way you see it". I look forward to changing the "way I see things" if I am reasonably proven wrong in my thinking.

[Les]

Originally posted by unregistered
aiiii

Les, help me here
1) give me a (frothy ?) verbal description of that second graph

2)(the first I would hazard represents 'reaL world' efficiency, as my uninsulated system was setup)

1)They are wb only C/W s .
The predicted C/Ws are a composite of the calculations posted here:
http://forums.procooling.com/vbb/sho...?threadid=6347 .
The joined "calculated data points" are at the same flow rate.
In that the Temp dependance of the Flomeric and Kryotherm "h" appear to be similar no differentiation is made betweeen wb type..

Your C/Ws simply calculated from "coolant temp", "die temp" and Watts..
Dunno whether your flow rate is constant but would conveniently explain the large discrepancy if was decreasing with coolant Temp.

2) Something like that.

[BillA]

ah so
#2 then is VERY enlightening
pHaestus ?

BTW, a 25% soln of antifreeze was used

[Cathar]

Althornin, I know you understand the following, but I just wanted to expand upon the point you raised, I am not contradicting you.

Originally posted by Althornin
not at all!
Lowered flowrate results in a COOLER coldplate, as far as i can see.

Yes, this is true, but this is a bad thing given everything else being fixed except for flowrate.

The non-intuitive flip here is that just because the cold-plate/water-block is cooler, this does not necessarily mean that the water will be colder. If the amount of heat being moved is the same, then what happens is just a large temperature differential between the cold-plate and the water.

If we keep the cold-plate warmer, the TEC's efficiency picks up, and more heat gets moved from the cold to the hot side. Also, and VERY importantly, for a high flow rate the differential between the water and the cold side is MUCH less than with a low flow rate. Yes, the cold plate is warmer than with a low flow rate, but the water is also THAT much closer to the cold-plate temperature.

Rather than mince words like the above description, this is why I tried to keep referring to the wattage moved. Don't think of things as temperatures, think of things as watts moved and then it makes sense when you keep in mind that you can't create/destroy energy.

It is true that improving the hot-side efficiency tends to have a greater effect though, but the cold-side efficiency is still important.

If you don't believe me, go to Kryotherm and type in some values for the cold-side C/W. The lower the C/W (more efficienct representing a higher flow rate), then the lower the equilibrium cold-chamber temperature, but the higher the cold-plate temperature.