Some Considerations on Thermal Transfer: Variations on a Theme

Good mounting bracket design could prove to have a noticeable impact on thermal transfer. The conclusion of my previous article led me to consider some further considerations for heatsink and waterblock designers. A designer needs to consider two factors:

  • First, mounting pressure is not an inconsequential amount of force;
  • Second, certain parts of the CPU will produce more heat than others.

In this article I will be expanding on some themes first brought up by Bill Adams in his article from last year – I would recommend reading it first for background.

A concave baseplate is the worst of all possible designs. Because thermal compound is the most restrictive leg of thermal transfer from the transistors to the air, it is in the designer’s best interest to minimize its thickness. This much should be obvious. A concave baseplate will have the greatest thickness in the center, the area of greatest thermal output.

All of the best water blocks I have tested in previous articles have had 0.5 mm thick baseplates. For sake of argument I will assume they were perfectly flat. The mounting bracket will need to apply the weight of the cooler plus the amount of force needed to keep in contact with the die when mounted vertically. This will be enough force to bend 0.5 mm of copper. A perfectly flat base will become concave.

Accounting for this is relatively simple. For the manufacturer, it is a matter of determining how much their chosen alloy will flex under the amount of force their bracket will exert. For the DIY modder, it will take a dial indicator and a weight. The weight needs to be equal to the amount of force the mounting bracket exerts. Apply the weight to the base and measure the amount of deflection with the dial indicator.

Once you know this you can then lap the base to achieve a curve with the opposite amount of deflection. There are two ways to do this: the precise way and the easy way.

The precise way is to grind two pieces of glass, or other hard material, together like a telescope mirror until you have the appropriate curve. Then grind the baseplate against this. The easy way is to use the traditional ‘sandpaper taped to a piece of glass’ lapping method with something soft under the sandpaper. One or two pieces of printer paper should do the trick.

Bill Adams’ article focused on how heatsinks could be tested more accurately. A good cooler should be designed specifically to pass this test. In short, the article points out that a CPU is not a uniform field of transistors. Certain parts of the chip will have a higher energy density than others. To operate most efficiently, a cooler should be designed to transfer the most heat at these points.

Far and away the most important factor in optimizing a cooler is the design of the metal to coolant transfer point. In plain English that means the fins. Forgive me for stating the obvious. A discussion of how this can be optimized is well beyond the scope of this article.

Another often overlooked factor is the mounting bracket. A mounting bracket determines exactly where the center of force will be on a cooler. The center of force determines which point will have the best contact on the die. Thus the mounting bracket can count for as much as surface figure.

There is only really one detail to be concerned with and that is whether the board is mounted vertically or horizontally. If the board is mounted horizontally, then the force of gravity lines up straight through the chip. In this case the bracket only needs to be designed to keep the center of gravity exactly where it needs to be.

If the board is mounted vertically, the force of gravity wants to pull the cooler off the CPU. Because the cooler is firmly mounted to the motherboard, this translates into a torsional force on the cooler. The mounting bracket needs to be adjusted to account for this.

Calculating the force is just a matter of high school physics, so I won’t go into it here. The implication for overclockers is how soon will manufacturers recognize this issue and how many will be willing to spend the extra money to design and include two mounting brackets for each type of processor. I am not about to guess dates. I can tell you that a cooler that does include extra brackets for this purpose is likely well designed.

Ian Anderson

Be the first to comment

Leave a Reply