Waterblock C/W and contact surface area

[RoboTech]

Am I correct in assuming that the thermal contact area (such as the core surface area) will have a big impact on a blocks C/W?

For example - say I mounted a waterblock on a CPU core, which dissipates 70 watts into the bp and thru careful measurements came up with a C/W value (say 0.24 C/W).

Now if I take this same setup and insert a 226 watt (Qmax) TEC/coldplate in between the CPU and waterblock, the input power to the Pelt will get added to the 70 watts of CPU heat producing somewhat over 300 total watts now going into the base of the waterblock.

If - I used the same 0.24 C/W as before and ran the numbers with the Pelt (0.24 C/W x 310 watts = 74.4 C + 25 C water temp) that would give an estimated hot side Pelt temp of 99.4 deg C! Not good...

Now I'm assuming this isn't really the case and that the much larger contact area between the Pelt hot side and waterblock will greatly decrease the waterblocks overall thermal resistance, resulting in a much lower C/W, right???

[myv65]

Originally posted by RoboTech
Am I correct in assuming that the thermal contact area (such as the core surface area) will have a big impact on a blocks C/W?

Variable depending on design, but yes it is significant.

Much of the resistance is contact resistance. This contact resistance still applies to the die-pelt interface, but is much less severe across the pelt-block interface due to larger contact area. Second, a big part of C/W is due to the heat flux (load/area) rather than overall load. Sort of a double-bubble for increasing the contact area.

Note that blocks will react differently. ie, direct impingement does not benefit much from spreading the heat over the total baseplate area. Maze-style blocks would probably see the biggest bump.

[RoboTech]

Thank you sir...

Now would it also be correct to make the assumption that since Q is directly proportional to the surface area in conduction that the relationship between contact surface area and the block's C/W might be somewhat linear?

[myv65]

Linear is likely too generous an estimate, but that's only a hunch on my part. Hmmm, sounds like a good item for some accurate testing. . . Any volunteers?