1
\$\begingroup\$

I'm designing a water block for a liquid cooling system. It will be acrylic on 5 sides and copper on the side that attaches to a CPU heatsink/fan. What I'm stumped on is the thickness of the copper side of the block.

A thicker sheet of copper can hold more total heat energy, but once the system reaches equilibrium that heat still needs to go somewhere, so I'm limited by the thermal interface between the water and copper, and then the copper and heatsink. So my gut instinct is that thickness doesn't matter. However, I often see cooling blocks advertised online as having extremely thick copper transfer plates, making me wonder if this is actually an important spec.

I understand that if my water block is larger than the contact area of the heatsink a thicker plate will help transfer more heat to the heatsink, but in this case assume that the water block, transfer plate, and heatsink interface are all the same size.

Does transfer plate thickness really matter when sandwiched between coolant and heatsink?

\$\endgroup\$
4
  • 2
    \$\begingroup\$ I think your gut instincts are listing out the considerations okay. Thicker copper will provide you more heat capacity, but once that capacity has been "charged" (in equilibrium), you still have to deal with thermal resistances to remove the heat to the "infinite capacity" storage of planet Earth (or your room.) The nice thing about added higher capacity (and copper which has a very low thermal resistivity) is that for momentary fluctuations, the copper capacity will quickly remove some of the heat energy and give time for the rest of the system to work. It also adds structural strength. \$\endgroup\$
    – jonk
    Commented Mar 24, 2017 at 19:31
  • \$\begingroup\$ the amount of heat energy removed Q=flow* heat_capacity* ΔT. I don't think there are any simple formula, latency, turbulence etc.. also loss of contact pressure from warp and surface roughness can be greater than expected. So determine thickness required for precision coplanarity and glass lapped finish. The surface area and water flow rate must be maximized to ensure cold water sink.. My friend invented his own copper sponge heatsink fabrication process for forced water built designed for Univ. researchers. \$\endgroup\$ Commented Mar 24, 2017 at 20:21
  • 1
    \$\begingroup\$ Its a nice discussion and all but, doesn't this belong in another SE? \$\endgroup\$
    – Wesley Lee
    Commented Mar 24, 2017 at 20:36
  • \$\begingroup\$ From a practical standpoint 0.062" (14 gauge) sheet copper is very common, readily available, and will do a fine job. \$\endgroup\$ Commented May 10, 2017 at 0:42

5 Answers 5

3
\$\begingroup\$

Let's consider what the thickness of the metal plate is doing

a) It's making it rigid, so the CPU-metal contact is good
b) It's spreading the heat sideways from the CPU, to areas in contact with the water
c) It's impeding the flow of heat from the CPU to the water

(a) is saying that it has to be thick enough mechanically, say at least 10% of its width.

(b) and (c) taken in combination say that if it's any thicker than the space between the edge of the CPU thermal contact and the edge of water contact, then it's not going to improve things making it any thicker. However, that's an upper limit, a thinner plate may still conduct an adequate amount of heat.

It's not a hard upper limit, as a thicker sheet will still impede the flow of heat very little, more an economic limit above which there's no performance gain. However, there may be an economic incentive for the manufacturer to use a very thick plate, as then he can advertise that it uses 'a very thick plate' (so it must be better, mustn't it!)

Where Ww is the width of the water contact, and Wc is the width of the CPU contact, the plate should be at least 0.1*Ww thick, with no thermal need for it to be thicker than 0.5*(Ww-Wc) thick.

\$\endgroup\$
0
\$\begingroup\$

A few factors to consider.

  1. The thermal resistance would be greator with a thicker place, everything else being equal.

  2. However, how well a plates contact with the heatsink matters. A thicker plate can maintain better contact over a wider area, reducing thermal resistance.

  3. Also a thicker plate can dissipate heat more evenly over a wider area within the plate, reducing thermal resistance.

....

The reality is likely a compromise of those factors, and money, space, weight,....

\$\endgroup\$
0
\$\begingroup\$

Thicker plate is more than heat capacity. Capacity is only important for initial conditions. In steady state it provides short path for energy between hot element on one side and each fin on other side. Sometimes it is what makes the difference between "all fins are effective" and "only 10%,the ones inside are effective, and others do nothing". If your plate becomes too thick, you can use heat pipes, they transfer heat much better than even copper.

\$\endgroup\$
0
\$\begingroup\$

Draw a grid of copper squares. Compute the thermal resistance of each square. Define the input heat region (presumably some 10% or 20% of the total width). Compute the #squares the heat must flow through, to get to fins or however else you will contact the cooling water. Experiment with thickness.

schematic

simulate this circuit – Schematic created using CircuitLab

\$\endgroup\$
-1
\$\begingroup\$

Yes it does.

Every material has a thermal resistance per unit thickness. So the thicker the copper plate the more temperature delta will be maintained across that plate at a given heat transfer rate.

Copper Has a thermal conductivity of 401 W/(m K)

Or to put it in more readable form, the temperature through the plate will rise 0.0025 Celcius per millimeter thickness for every watt you try to pass though 1 square meter of plate.

Whether that makes a hill of beans difference to your application I don't know.

As mentioned elsewhere, the thicker plate does help spread the heat more evenly across the device / heat-sink interface layer.

\$\endgroup\$
2
  • \$\begingroup\$ That .0025 deg C per Watt figure assumes the area of the plate is one square meter. \$\endgroup\$
    – user28910
    Commented Mar 24, 2017 at 20:05
  • \$\begingroup\$ @user28910, you are right, I forgot to add the how to calculate the heat watts part \$\endgroup\$
    – Trevor_G
    Commented Mar 24, 2017 at 20:14

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.