MCW50-T pelt cooling enhancement

[mikead_99]

It looks like my Maze4-1 isn't going to fit on the DFI LP UT with a decent chipset cooler. I took the time last night to swap out the 172W pelt I had in it to my Swifty pelt block (no results yet, lots of other work going on in the box :D ) and was left with a question. With the water in the Swifty hitting the hot side directly, I was wondering if anyone has tried to enhance this a bit? What I was thinking could work is Artic Alumina'ing a small chipset heatsink to the pelt hot side. I see the pros and cons like this:

Pros:
- Increased cooling surface area
- Increased turbulance

Cons:
- Decrease in the volume of an already small block
- Possible increase in flow restriction
- Whatever inefficiency the TIM introduces

So since I didn't do this, I'm left second guessing myself. Anyone out there try it?

[saaya]

for an 80W tec the surface area of the tec is bigger than needed to dissipate all the heat i think, so directly cooling it results in the best temps as theres nothing between it and the water that would buffer the heat and increase the temps slightly.

for a 172W tec a bigger surface might pay off and work better even though the heat has to travel through a lyer of metal first.

has anybody ever compared the mcw50-t with direct tec cooling and the old ones with a copper base plate above the tecs hot side?

[mdzcpa]

Leave the direct hot side of the pelt in contact with the water. The more exposed surface of the hot side to the water the better. By placing anything onto the hot side of the pelt (ei heatsink), you will loose more cooling capability from reduced water contact to the hot side of the pelt than you will gain from increasing total contact surface area. The thermal transference between the hotside and any heatsink is MUCH less than the hotside direct to water tranference is.

I speak from personal experience on this. Let the water contact as much of the actual pelt hotside as possible.

[mikead_99]

I speak from personal experience on this. Let the water contact as much of the actual pelt hotside as possible.

That's what I was looking for. Now I feel better for leaving it as is. :toast:

[saaya]

well it depends on the heat/surface we are talking about, think about direct die cooling cpus!

ive not seen any direct die cpu cooler that reached the same or even lower temps as the good blocks out there. i think the heat/surface on a cpu is too big, the surface area too small for water to remove it efficiently.

an oced athlonxp has around 120W of heat on a surface of 1cm^2.
that seems to be already too much to cool with a direct die block, as waterblocks with a higher contact surface area perform better than direct die blocks.

so lets say above 80W per cm^2 a direct die solution is less efficient at removing the heat by adding a layer of metal wich increases the contact surface. this also creates a buffer wich means the temps will actually be higher, but lets say at 80W per cm^2 the additional surface area removes enough heat to compensate for that.

we are talking about 40mm tecs here, not the entire surface is getting equally hot as the alloy blocks between the cermaic plates are not making contact with the entire ceramic surface. but lets assume the heat is spread equally.

not the entire 40x40mm are making contact with the water as there is a rubber insulation string on it to make sure its watertight, right? so lets say its 35x35mm.

then we get a surface area of 12.25cm^2.

so if my guess that 80W per square centimeter or less can be removed more efficiently by directly contacting the ceramic layer, then the limit where direct die cooling a tec becomes less efficient is:

12cm^2x80W=960W :D

my calculation is probably quite a bit off as i havent included some factors, but i think you get the point now. even a 227W tec is best cooled with a direct die block.

so i guess i was wrong, even a 360W tec is probably best cooled with a direct die block as it dissipates a max of around 550W of heat on its hot side. :)

[saaya]

hmmm id like to see a mcw50/t with an one without direct die cooling the tec, does anybody have an old mcw50/t that doesnt directly cool the tec?

id like to see some real life results to have some more facts i can play with to make sure my calculation is more or less correct :)