DO NOT TRUST ANY OF THESE MEASURED PERFORMANCE NUMBERS, my thermometry is now suspect
I have just started testing a new block I have come up with.
Comparing the results to those I have for my Crater block from My die sim and pHaestus' measurements, I think it is close to a Cascade/ MCW6000. Could even be significantly better with work.
There has been a lot of thought gone into this one. It is a departure from the thin baseplate idea, having not less than 11mm of copper between die and water.
Some comments from
Jabo (who seems to have moved away from this forum) were an impetus, regarding a spherical type of block.
This IS a hemispherical block. I'm seeing as low as 0.169C/W at 3.76lpm on my setup. This compares to .19 for the Crater block with the same pumps, before TIM settling.
Some logic to the design.
A thin parallel baseplate block has a fairly linear thermal gradient through the copper, a high "h" is possible because all the water can be concentrated on a small area. A Cascade.
A thick parallel baseplate has a less linear thermal gradient, the thermal resistance of the copper gets progressively lower the further from the die you are but it is offset by an increase in average delta L ; the "h" is normally (but not necessarily) lower due to the water being spread out over a larger area, area increase being approximately proportional to L^2, but the larger area itself is an advantage. MCW6000.
A
hemispherical baseplate has a
very nonlinear thermal gradient, the coppers thermal resistance decreases rather rapidly as delta L increases. L stays as in a parallel heat channel as cross sectional area increases proportional to 2pi*L^2. This is the key; relative to the available surface area, the distance between water and die is low. The problem is to maintain a high convection coefficient.
This is done in this case, as Jabo suggested, by reducing the height of the block structure clearance. I've done this progressively radiating out from the inlet. At the outer edge of the block the clearance between block and middle cap is 0.5mm.
This maintains the water velocity down the slope of the sphere and thereby a reasonable convection coefficient. The structure I have not optimised at all, the object was to test the concept.
Some pictures, sorry about the quality:
As a bonus you get to see my bodgy die sim.
I will post some better pictures of the internals later.
Incoherent