Pro/Forums - View Single Post - Min. temp delta above room temp with water cooling

Cathar · 01-01-2004, 05:02 PM

I believe that we are getting pretty close now.

For a 100mm^2 CPU die, and a strong pump, the best waterblock's C/W is around 0.14-0.15 (my guess - I could be totally wrong - and I know a number of people who would probably object to that figure - but anyway).

Of that 0.14-0.15:

~0.07-0.10 is due to the thermal paste layer
~0.020-0.025 is due to the thermal resistance of having the (thin) metal base-plate on the block (non-direct die)
~0.005 is due to the thermal capacitance of the water (the water does unavoidably warm up when in contact with the heat)

Which leaves an estimate of around 0.02-0.06 C/W being due to the block's design.

I don't believe that a figure of 0.02 is a realistic value. I'd say that the real figure is more like 0.04-0.06 C/W is what we have left to gain.

Mid you, that's for 100mm^2 heat sources. For larger sized heat sources the figure is lower.

For a 100W CPU, that represents a 4-6C gain still to be realised, and as little as a 3-4C gain for larger CPU dies.

I believe that there is the potential to eke out about another 0.01-0.02C/W (1-2C), but doing so will be quite difficult, but beyond that I don't foresee any gains being realistically made unless the thermal interface layer is improved dramatically (it too being at a near standstill for improvements of late).

It must be noted that these are just my personal thoughts on the matter from my understanding of what I've observed, not necessarily based in hard fact.

01-01-2004, 05:02 PM	#3
Cathar Thermophile Join Date: Sep 2002 Location: Melbourne, Australia Posts: 2,538	I believe that we are getting pretty close now. For a 100mm^2 CPU die, and a strong pump, the best waterblock's C/W is around 0.14-0.15 (my guess - I could be totally wrong - and I know a number of people who would probably object to that figure - but anyway). Of that 0.14-0.15: ~0.07-0.10 is due to the thermal paste layer ~0.020-0.025 is due to the thermal resistance of having the (thin) metal base-plate on the block (non-direct die) ~0.005 is due to the thermal capacitance of the water (the water does unavoidably warm up when in contact with the heat) Which leaves an estimate of around 0.02-0.06 C/W being due to the block's design. I don't believe that a figure of 0.02 is a realistic value. I'd say that the real figure is more like 0.04-0.06 C/W is what we have left to gain. Mid you, that's for 100mm^2 heat sources. For larger sized heat sources the figure is lower. For a 100W CPU, that represents a 4-6C gain still to be realised, and as little as a 3-4C gain for larger CPU dies. I believe that there is the potential to eke out about another 0.01-0.02C/W (1-2C), but doing so will be quite difficult, but beyond that I don't foresee any gains being realistically made unless the thermal interface layer is improved dramatically (it too being at a near standstill for improvements of late). It must be noted that these are just my personal thoughts on the matter from my understanding of what I've observed, not necessarily based in hard fact.