My "Lumpy Channel" TEC concept block.

[jaydee]

Originally Posted by JFettig

-edit- oh yeah, good luck trying to fit that large of a block WITH insulation on a motherboard. I couldnt fit mine with insulation on my 8k3a so I had to wait till I got a new board, even now its pretty close on one side.

Jon

It actually will not fit at all. Just measured 3 of my boards and the Caps get in the way on all of them. Have to cut it down to 2.5" which means ,as you said, removing the O-ring groove which isn't an option. Guess I can deal with a 172watt 40mm TEC.... I heard rumors of 226w 40mm comming but have yet to find any...

[8-Ball]

Personally, I would suggest getting a mobile barton core putting out only 45W at stock and then use a high powered 40mm TEC.

8-ball

[jaydee]

Originally Posted by 8-Ball

Personally, I would suggest getting a mobile barton core putting out only 45W at stock and then use a high powered 40mm TEC.

8-ball

My original plan was to get a Duron 1.8gig. How well does the Mobile's overclock?

[sandman]

Insanely well. 2.7gigs on aircooling isn't really uncommon, and 2.85 on water is attainable with a good stepping 2600-M

[jaydee]

Originally Posted by sandman

Insanely well. 2.7gigs on aircooling isn't really uncommon, and 2.85 on water is attainable with a good stepping 2600-M

Well maybe I will get both then. By the time I get this pelt block done they may not exist anymore though.

[8-Ball]

They're probably the best bang for buck overclockers you can get at the moment.

At least, that's the impression I'm getting.

8-ball

[jaydee]

Well I had a 4 hour flight this morening to Virgina. I spent what battery power I had drawing up the 172watt 40mm pelt "concept". I don't have Solid Works on this comp anymore as I just reloaded Windows for the first time in 3 years. I did get ACAD2000 on before I left though.
Note the holes will still be connected with the 1/16" endmill although it isn't shown in the drawing.

[kronchev]

Originally Posted by sandman

Insanely well. 2.7gigs on aircooling isn't really uncommon, and 2.85 on water is attainable with a good stepping 2600-M

2400-M can do the same one but it seems its rarer to get the good steppings

ive seen 3.0 on a prometia

[killernoodle]

Hey, there was a post on the previous page mentioning that the flow would be concentrated (somewhat) in the center. Wouldnt this be desireable? I mean, with the heat in the cold plate would be concentrated in the center over the core because that is the shortest distance of copper from the core. So, the peltier would be producing the most heat there because (I assume of course) the temperature over different the area of the peltier is subject to the intensity and position of the heat load and the nature of the cooling device on top of it.

What I'm trying to say is: wouldn't it be better to concentrate the water to the center of the block because that is where the most heat is being produced?

[8-Ball]

Originally Posted by killernoodle

Hey, there was a post on the previous page mentioning that the flow would be concentrated (somewhat) in the center. Wouldnt this be desireable? I mean, with the heat in the cold plate would be concentrated in the center over the core because that is the shortest distance of copper from the core. So, the peltier would be producing the most heat there because (I assume of course) the temperature over different the area of the peltier is subject to the intensity and position of the heat load and the nature of the cooling device on top of it.

What I'm trying to say is: wouldn't it be better to concentrate the water to the center of the block because that is where the most heat is being produced?

This seems to make sense, though I don't yet know enough about peltiers to be sure.

@jaydee

If this were the case, then I would imagine a block where you connected the holes to form a spiral with a number of parallel channels. This way, you could have a central inlet with an outlet towards the edge.

Seems a bit like a step backwards in waterblock design, but it may yet be the best optimisation of your design principle.

8-ball

[jaydee]

From what I understand of the TEC's the heat/cold is evenly spread across the entire TEC. maybe take some more looking into. The core of the CPU shouldn't effect the way the TEC operates.

[killernoodle]

Well, lemme explain this in a simpler manner. Remove the cold plate and put the tec right on the core. Wouldnt the center of the TEC recieve the most heat and therefore output this heat right to the other side? (as I understand it, thermocouples are vertically placed pins of dissimilar metals that move electrons vertically and therefore transfer heat directly vertical to the other side) With this reasoning and the knowlege that the heat from a CPU will radiate out-ward (see footnote) concentrically (hard to word) through the heatspreader or cold plate, wouldnt the heat from the core concentrate the shortest distance to the other side of the cold plate and therefore directly through the peltier and to the center of the waterblock?

It is tough to get my point across using this damn english.

EDIT: conversely, cooling the center of the TEC more would yield a cooler core temp... (even if slightly) correct?
EDIT2: would this be an appropriate application of the radius?

Hey, why does the forum **** out-war (without dash). A quick search of google tells me that this is some kind of virus program. Is there any way to remove the **** feature of the forums?

[jaydee]

Originally Posted by killernoodle

Well, lemme explain this in a simpler manner. Remove the cold plate and put the tec right on the core. Wouldnt the center of the TEC recieve the most heat and therefore output this heat right to the other side? (as I understand it, thermocouples are vertically placed pins of dissimilar metals that move electrons vertically and therefore transfer heat directly vertical to the other side) With this reasoning and the knowlege that the heat from a CPU will radiate ******d concentrically (hard to word) through the heatspreader or cold plate, wouldnt the heat from the core concentrate the shortest distance to the other side of the cold plate and therefore directly through the peltier and to the center of the waterblock?

It is tough to get my point across using this damn english.

EDIT: conversely, cooling the center of the TEC more would yield a cooler core temp... (even if slightly) correct?
EDIT2: would this be an appropriate application of the radius?

I Understand what your saying. I am just not certain it would make a substantial difference. Not even sure how to test that.

[killernoodle]

I guess it doesnt matter much with a thick cold plate.

[Butcher]

A lot of the heat is off the TEC itself rather than the CPU core anyway. Assuming you're running a 172W TEC you'e looking at 200+W off the TEC and 50-80W off a mobile barton with a high overclock.

[jaydee]

Originally Posted by killernoodle

I guess it doesnt matter much with a thick cold plate.

Well once equlaibrium sets in the hot spot should still be in the center of the cold plate so I am really not sure.... The block I just posted will still favore center channels somewhat so I guess it shouldn't hurt. :shrug:

[jaydee]

Originally Posted by Butcher

A lot of the heat is off the TEC itself rather than the CPU core anyway. Assuming you're running a 172W TEC you'e looking at 200+W off the TEC and 50-80W off a mobile barton with a high overclock.

True it very well might get mixed in with the TEC heat and be irrelevant.

[killernoodle]

A 172W tec cannot put out 200W of heat, but a TEC and a CPU can put out more than 200. This next section is mostly deep thought, so if my physics are terribly wrong please correct me: This makes sense because a TEC's heat is concentrated in the center. The middle of the TEC is surrounded on all sides by thermocouples producing heat. on the edge of the TEC, the heat of nearby thermocouples is only on one side. The closer to the center of the TEC, the more the thermocouples surround the point. I mean, you wouldnt want the TEC offset from the center of the core for this reason (only a certain amount of the cooling potential would reach the core). You cannot dismiss the 80W a CPU puts out if it is concentrated on a small area.

This is too hard to word. Lemme draw some pictures.

[killernoodle]

With this drawing, it seems like it would be best to cool the middle more than the outsides.

What we really need is some kind of thermal imaging system that can differentiate temps down to a degree or two and output this image using shades of grey or colors. That would be nice.

[8-Ball]

Originally Posted by killernoodle

A 172W tec cannot put out 200W of heat

Be careful. The Qmax rating of a TEC, ie the 172W in this case, is a rating of the maximum amount of thermal energy the TEC can dissipate FROM THE HEAT LOAD, and is NOT anything to do with the power drawn by the TEC from the power supply which must also be dissipated as heat. This can often be more than the Qmax of a TEC when run at Vmax.

8-ball

[killernoodle]

Oh, I just assumed the rating 172 was the amout of watts the TEC consumed. Guess I need to read up a little more on how they are rated.

[pHaestus]

JayDee you aren't making a TEC block based upon the lumpy channel design any more you're designing an all new TEC block. Here give me a moment to display my horrible Paint skills

[pHaestus]

pHear my skills!

The idea here is to equalize flow over the square by having shorter channels at the end. The 2 barb design and overall lumpy channel style is preserved by rotating the channels 45 degrees. Same block slightly modified in essence

[killernoodle]

WTF is that? Two jellyfishes fighting?
I see your concept, but I have a question or two.
pH, do you want to equalize flow or favor the center of the TEC? That has been the argument in the past couple posts. Would favoring the center bring lower temps to the core?

Another solution would be to use the original design of a bunch of parallel lumpy channels going straight across, but stagger the inlet and outlet: like one barb on the top right, one on the bottom left.

[Butcher]

Originally Posted by killernoodle

Oh, I just assumed the rating 172 was the amout of watts the TEC consumed. Guess I need to read up a little more on how they are rated.

Heat dissipated is given by P=VI. So a 24.6V TEC drawing 11.3A will produce 278W of heat itself. The heat load of a CPU or other device is additional to this - so for an 80W CPU you'd be looking at 360W of heat from the hotside.