Quote:
Originally posted by utabintarbo
I would think that the conductivity of the block medium would be less of a factor than the "heat-holding capacity" (insert appropriate technical term please) of the medium. This would alter the propagation pattern from one medium to another, no?
Bob
|
This is a common misconception and a large part of the "aluminum better at getting rid of heat" myth. For the most part heat transfer from our CPUs is a steady-state phenomenon. This means that once everything has been running a while, the temperature at any given point within the system no longer changes. Once you reach this point, specific heat has absolutely zero bearing on heat transfer.
All specific heat does is determine how long it takes to heat up a material (or cool down) when a heat source is added (or removed).
Some folks think that since the water flows through the block you couldn't possibly have steady-state. It's all a matter of definition. Yes, the water undergoes a warming and cooling cycle as it circulates through the loop. At any given position within the system, however, the water there will be a steady temperature after the initial warm-up period ends. To an engineer, this is still steady-state.
My take on your posts is that you would like to find the "optimum" aluminum shape for a block and then begin making it in copper. I guarantee you will not have the ideal copper shape if you follow this method.
Honestly, however, we're really splitting hairs when we start talking about "optimum". Any block will have different results depending on the amount of heat from the processor, the size/shape of the die, and the flow rate of the fluid. This means there is no single optimum that applies universally to all situations. We're also talking about small differences between the "best" blocks that take different philosophies to cooling (direct impingement, spirals, multiple cross channels, etc.).
What would be kinda neat would be to see a manufacturer employ investment casting to produce blocks. You'd have a near-final geometry that would require nothing more than mounting a cap and final grinding of the base. It would also have a low cost per part (but pretty high tooling cost for the die). I guess there just isn't enough demand to justify that initial investment.