cpu waterblocks thermal performance

[laubman]

Howdy everyone,

I've been meaning to ask this question for a long time, but didn't know where to ask.

I have a question regarding the behavior of cpu water blocks thermal performance.

A typical thermal resistance vs flow rate graph is here:

http://www.swiftnets.com/assets/images/products/apogee/Single-Die-TR_VS_FR-SMALL.PNG

My question is why is the relationship between flow and thermal resistance not linear, but pseudo inverse square?

 

[laubman]

sorry, forgot to add, here is the product page for a typical cpu water block

http://www.swiftnets.com/products/APOGEEGT.asp

 

[IRstuff]

It's hardly "psuedo" inverse square, doubling the flow rate doesn't even begin to halve the thermal resistance, so it's just barely inverse.

Is this for school?

TTFN

FAQ731-376

 

[laubman]

I wish I was still in school, the old days with not a worry in the world. Sadly it appears I should head back to maths class.

So let me rephrase my question, why does the thermal performance trend taper off once above a few GPM and appear to be headed for a flat line?

 

[25362]

The convection heat transfer coefficient (htc) in plate heat exchangers is generally assumed to be proportional to G^0.38 in laminar flow, and to G^0.65 in turbulent flow, assuming the geometries and the thermophysical properties of the fluid don't change. Where G is the mass flow rate, generally incorporated in what is known as Re number.

In fact fluid properties change with temperatures (as a result of changing flow rates). For example, for water a drop in temperature increases the density, the viscosity and the Pr number, while decreasing the thermal conductivity. All of these factors also affect the value of the htc.

I wonder whether these considerations can be applied in the case of "cpu water blocks".

 

[IRstuff]

OK, the simple answer is that even if the thermal conductivity of the liquid interface were infinite, there's still the thermal conductivity of the metal parts of the heat exchanger, which remain constant. The thermal connections between the die and the heat sink are likewise constant.

Therefore, you'd expect asymptotic behavior.

TTFN

FAQ731-376