swifty gpu blocks how efficant becasue water directly cools hot side??

[qwerty57]

yeah swifty blocks for gpu are

water
hotside
coldside
coldplate

is this bad?? or would i gain anything from butting the tec on the bottom of the block and add a new coldplate so both surfaces have copper contact??

[winewood]

You dont need a tec plate and tec to cool a gpu.. yet. There isn't enough heat put out by them that watercooling cannot take care of. You would be wasting your money.

[qwerty57]

what a waste of an answer come on you are tring to tell me that a gpu at -20 will run as fast as a gpu at 15c??? thats but no thanks for your responce..

[winewood]

Be that as you may, I know of no program that cannot be run on a current card that needs watercooling, that is not limited by the memory bus speed. Therefore if you gpu already hits that limit, how would it reaching that limit faster help you?
Let me ask this. What do you need that needs a processor at -15?

[BrianH]

winewood

You dont need a tec plate and tec to cool a gpu.. yet

What are you doing in this forum? Without these people experimenting with this stuff, we would not have the experience to use the tec when the time comes. Oh, and I disagree with you, -15C GPU overclocked in an overclocked CPU with a high PCI speed can use the bandwidth.

Take a look at the 3dmark scores... as the GPU speed increases, so does the score. Why would that happen I wonder ....

[winewood]

You guys should test your questions yourself so you will appreciate and understand the results. I'll edit my posts. Sometimes hard earned data isn't taken with appreciation, because it wasn't their work that derived it.

[qwerty57]

so

block
water
copper base
pelt
gpu

pelt cold side right on the gpu

would be more efficant than

block
pelt
coldplate
gpu

here as here water cools down the hot side.

also this si how the blocks are manuf. with no buffer on the hotsdie at all the water cools it... my question was what if i modded the gpu block to add a buffer to the hotside.

[joemac]

so

block
water
copper base
pelt
gpu

pelt cold side right on the gpu

would be more efficant than

block
pelt
coldplate
gpu

No thats not what being said?? go with:

block
water
copper base
pelt
T.E.C plate (copper)
gpu

This will allow you to apply the high pressure a pelt needs.

[BrianH]

OK guys ... from Swiftech

Module hot side is directly submersed into the coolant for optimal heat dissipation.

This means that there is no "hot plate" between the water and the tec. Therefore the mounting pressure for the tec is not valid here. Whether this is effective or not I do not know but others have reported good results.

qwerty57

is this bad??

I do not think so. The reason up to now to use a "hot plate" was to distribute the heat over a larger copper area. Hence the need for clamping pressure that has been documented. With Swiftech's approach, the size of the block matches the tec size. Also, because there are no thermal layers, efficiency should be much better. They are a pretty reliable company.

Therefore, it will be Water --> Tec --> base --> GPU

winewood
I did not attack you (well maybe a little ). That was only because you asked why he wanted to do this, this would not help for anything and never answered his question. Overclocking a 2.66 CPU to 3.2gHz using a chilled system does nothing in most programs also ... but it is fun and creative to get there. Most people think I am nuts for water cooling let alone running -10C coolant!

Oh.. and pci bus isnt a factor on the new mobo's and graphics

Oh, can we run the PCI and AGP speeds independently now? I do not think so. AGP is always 2x PCI. That is what I was refering to.

Every one who tweaks knows

The AGP bus has its own access to the processor outside the pci bus.

Hope this helps qwerty57 and joemac
and sorry winewood

[qwerty57]

well last night I did some testing and found that I had -19c on the block out of the case running 10-15c chilled water through it. this is my 172w (24v)tec at 12.6v.. so results are good but Im going to do this now.

block
mcw40 non tec copper bottom
pelt hot side
pelt cold side
mcw40-t stock copper bottom
gpu

since both plates are exactly the same size all i figure I have to do is get longer bolts that will fit though both plates and sandwich the pelt between two MCW40 plates the hot side will have dimmples (stock mcw40 non-tec plate)for added turbulence and Ill use the stock coldplate.. this way I can get superior pressue that I hear is what needs for optimal performance and at the same time give the hot side a base to disperse the heat away more evenly.. will this help? I hope so.

[msv]

Practical question:
Does all this hardware on the video card affect the AGP-slot?
It sounds like a lot of weight hanging in the slot. Does the water tubing carry any of this weight? The video card "hanging" in the tubes?
Oh, yes, there can hardly be any *need* for this performance, not to mention the money it costs, or the time, but sure it´s fun.
regards
Mikael S.

[qwerty57]

Practical question:
Does all this hardware on the video card affect the AGP-slot?
It sounds like a lot of weight hanging in the slot. Does the water tubing carry any of this weight? The video card "hanging" in the tubes?
Oh, yes, there can hardly be any *need* for this performance, not to mention the money it costs, or the time, but sure it´s fun.
regards
Mikael S.

the block is alum&copper and doesn't really weigh much at all. and no it doesn't effect the agp slot at all..

Oh, yes, there can hardly be any *need* for this performance

why are you in this websites forum if you talk like this?? this is a forum of people electro cooling there pc's to the "XTREME" we do this for fun and hobbie as many play basketball we overclock pc's there no reason to do anything in life we do this as others do other things...

so far alot of people responding that shouldn't be in these forums at all

[winewood]

Frankly, any forum worth its weight will have people who are new or don't understand why or what you are trying to accomplish. Having users who cannot tolerate, explain, nor reasonably voice why or the value added to a fuction are of no use to a logical, and professional site. Adversity makes you wiser as you learn to think, voice your opinion, and rationalize why you are there in the first place. Not being tolerant, nor accepting the fact of what is in life will not lead to a rich understanding of this topic. It just so happens the guy that doesnt belong "in your opinion" happens to know quite a bit about TEC cooling, compression. However, if you desire a blind leading the blindly accepting, that is your choice.

oh.. and agp is not always 2Xpci. Explain AGP 2, 4, and 8X. I think an understanding of the NForce and Nforce 2 architecture will help, as to independent graphics and pci access.


edited to remove my testing data: I feel that people would appreciate results if derived from experience since many aren't listening nor appreciate it.

[Since87]

Originally posted by qwerty57
well last night I did some testing and found that I had -19c on the block out of the case running 10-15c chilled water through it. this is my 172w (24v)tec at 12.6v.. so results are good but Im going to do this now.

block
mcw40 non tec copper bottom
pelt hot side
pelt cold side
mcw40-t stock copper bottom
gpu

I've been curious about the clamping pressure issue with the Swiftech 'wet TEC' blocks.

I'd suggest you try it both ways:
Hotside -> copper -> water
Hotside -> water

I'd be very interested in knowing the results of the comparison.

Do you have a torque driver? How are you going to control the amount of clamping pressure?

[winewood]

edited to remove my testing data: I feel that people would appreciate results if derived from experience since many aren't listening nor appreciate it.
Thanks since87 and joemac for being open.
Hope your tests work!

[qwerty57]

Do you have a torque driver? How are you going to control the amount of clamping pressure?

yeah I have one and its going to be

block
hotplate
pelt
coldplate
gpu

im using two stock mcw40 plates so i can use the stock mounting holes for awsome presure 160lb's or such

[BrianH]

Winewood....winewood ... winewood...

Ok here it goes, I tried to lighten it up but ....

Frankly, any forum worth its weight will have people who are new or don't understand why or what you are trying to accomplish. Having users who cannot tolerate, explain, nor reasonably voice why or the value added to a fuction are of no use to a logical, and professional site.

Agreed. But in your original post of

You dont need a tec plate and tec to cool a gpu.. yet. There isn't enough heat put out by them that watercooling cannot take care of. You would be wasting your money.

You did not answer his question. You simply gave your opinion that he is wasting time and money. He may be, but I do not agree and neither does he. I assumed that qwerty57 knew what he was doing and why ... unlike you who thought that with 11 post, he was a noobie. Your first post was ... are you stupid, that will achieve nothing but waste money...

qwerty57's question was cooling the hot side with water directly instead of a hot plate ... good or bad? Not whether chilling the GPU is good or bad. Two separate theorical discussions, but your post basically says waste of time. I think qwerty57 also agreed with

what a waste of an answer come on you are tring to tell me that a gpu at -20 will run as fast as a gpu at 15c??? thats but no thanks for your responce..

So, instead of trying to put down someone's discussion, try providing useful info as why direct water cooling if the hot side of a tec is good or bad. You may be a tec guru but from you answer, I have no way of knowing.

Oh and

oh.. and agp is not always 2Xpci. Explain AGP 2, 4, and 8X. I think an understanding of the NForce and Nforce 2 architecture will help, as to independent graphics and pci access.

And the NForce not withstanding, the PCI Mhz bus speed is ALWAYS 2x the PCI (ie 33Mhz vs 66Mhz) The 4x/8x AGP is adding additional transfer points on the rising and falling signals being provided as 66mhz. Yes the thruput is increased but the frequency is still 66 mhz. By boosting the PCI bus speed thru OC'ing, you can increase the bandwidth thru the AGP port ... of course with the AGP multiplier.

[8-Ball]

Winewood,

I have to agree with Brian here.

While it is your opinion that peltier cooling a graphics card may appear to be a waste of money, the original poster has already decided that he will try it anyway.

He was asking a question about the mode of operation of this TEC/waterblock, which you chose not to comment on, instead voicing your opinion on a matter not originally questioned.

As for your comments about the "quality of the forum", it is brought down by people like yourself. If you want to throw your opinions about, particularly if it does not provide answers to the question at hand, then perhaps you should have a look at OCforums.

I'm sure there are threads around here somewhere discussing the benefits of peltier cooling a gpu. THIS thread is asking whether a hot plate is needed between the peltier and the coolant.

If you have no useful experience or knowledge to impart, or find a constructive way to contribute to the thread then don't post.

[/rant]

[back on topic]

There are a number of questions that need asking.

1. What is the heat load from the hot side?

2. What is the convective heat transfer coefficient for the flat surface of the block?

3. Would adding a copper hot plate with furniture of some kind, whther this be designed to increase turbulence or surface area, or both, provide a sufficient increase in convective heat transfer coefficient at the interface with the coolant and the hot plate, to more than compensate for the additional temperature gradient required across the copper heat plate to transfer the heat. The thickness of the copper plate will play a large part here.

4. Clamping pressure. Is it the case that clamping pressure is required to reduce the thermal resistance of the interfaces between TEC and copper, or does it also affect the performance of the TEC itself, by compressing the peltier. If this is the case, then having the hot side cooled directly might reduce the efficiency of the TEC directly by reducing the pressure on the TEC.



I suspect that if you use a more powerful TEC, then it may become necessary to use a hot plate with an intricate furniture to provide the highest possible convective heat transfer coefficient, however, for a low power TEC, such as that provided as standard with the MCW50T, will be sufficiently cooled by direct contact with the coolant.

I would also suspect that in this case, the performance would be LARGELY dependent on the flow rate, whereby if the flow rate were too low, the convective heat transfer coefficient would be too low, requiring a large temperature delta between the coolant and the TEC hotside but since the geometry factor of the convection is poor, then turbulence will have to be introduced by means of high flow rates.

As for whether the clamping pressure affects the TEC directly as opposed to just affecting the thermal interfaces, I don't know the answer to this, but I suspect there will be someone here who does.

8-ball

[BrianH]

1. What is the heat load from the hot side?

Using Kryotherm with resonable value, the hot side would be around 120w.

2. What is the convective heat transfer coefficient for the flat surface of the block?

Not sure whay you mean here. If you are talking about the MCW50-T block, then direct TEC cooling would eliminate the heat transfers between the hot side/AS3/Copper base. Should be much more efficient overall???

3. Would adding a copper hot plate with furniture of some kind .......

I do believe in turbulance. I believe the design of MCW50-T would cause a lot of turbulance within the block. If not, perhaps a "turbulance generator" could be designed into the to of the block.

4. Clamping pressure

With diect TEC cooling, the clamping pressure on the hot side is no longer needed. The cold plate and the TEC with still need a large clamping force. Since the hot side would only be clamped along the edges (I assume??), then great clamping force on the cold side may be difficault to achieve.

I suspect that if you use a more powerful TEC, then it may become necessary to use a hot plate

Agreed!

I would also suspect that in this case, the performance would be LARGELY dependent on the flow rate,

As most cooling systems are. May be of more importance with the MCW50-T.

qwerty57

Are you still going to use the MCW40 block with 2 plates? If you have not purchased yet, I would consider the MCW50-T block. Others have reported good results, I believe. I think the direct TEC cooling will be the way of the future.

[8-Ball]

Originally posted by BrianH
Not sure whay you mean here. If you are talking about the MCW50-T block, then direct TEC cooling would eliminate the heat transfers between the hot side/AS3/Copper base. Should be much more efficient overall???

The mechanism by which heat is tranferrd from a solid body into an adjacent fluid, whether stationary or flowing, is by convective heat transfer.

It is "possible" to work out the convective heat transfer coefficient for a heat exchanger, though not really necessary. It is however beneficial to understand the factors which come into play.

The Nusselt number (non-dimensional convective heat transfer coefficient) is as follows.

Nu = C x Re^m x Pr^n

Pr is the Prandtl number, which is the the kinematic viscosity divided by the thermal diffusivity of the fluid, and is dependent on properties of the coolant fluid alone, so does not change with the design of the block.

Re is the reynolds number, and is a measure of the flow characteristic of the coolant. The higher the Reynolds number, the more turbulent the flow. There is a geometric factor in here, which depends on the furniture of the block as well as the flow rate and dynamic viscosity of the fluid and it's density.

C, m and n are constants which are dependent on the geometry of the heat exchanger and the range of the reynolds number, ie the characteristic of the flow.

Ultimately, we want to maximise this, which means changing the geometry and the reynolds number to suit.

The point I am getting at is that the convective heat transfer coefficient of the flat surface of a TEC is not likely to be very high unless the flow can be made extremely turbulent, ie by having very high flow, or with the addition of "turbulators". So would the increased convective heat transfer coefficient of an elaborately machined hot plate be sufficiently high to more than compensate for the additional heat transfer "stage" introduced, ie, getting from the surface of the TEC to the fins/pins or whatever has been machined into the hotplate.

Originally posted by BrianH
With diect TEC cooling, the clamping pressure on the hot side is no longer needed. The cold plate and the TEC with still need a large clamping force. Since the hot side would only be clamped along the edges (I assume??), then great clamping force on the cold side may be difficault to achieve.

I agree that high clamping pressure is required to reduce the thermal resistance of an interface, but I was wondering if there was anything else whic would very with pressure.

Would the operational efficiency of the TEC ITSELF, not the interfaces, vary if depending on how much it is being clamped.

ie; might the semiconductor jucntions behave slightly differently if they are clamped.

Also note that the thermal resistance between each of the little "pellets" in the TEC and the white ceramic sides will also change with clamping pressure.

So it may not be just the interface between the hot side and the hot plate or the cold side and the cold plate which is affected by clamping pressure.

I don't know the answer to these questions, though I suspect the junction properties will probably not change, but there are still interfaces between the individual peltier elements and the ceramic mounting plates. Could this interface be affected by removing the clamping pressure.

8-ball