Les, you may wish to rework those graphs somewhat. You seem to be listing 8 channels over a 22x22mm area, when really it's 8 channels (9 walls) over a 30mmx17mm area.
I agree that to cool a 1mm^2 heat source that a thicker base-plate would definitely be needed, but we're not cooling 1mm^2 heat sources.
Most of the graphs you showed had the 1mm base-plate (actually I'm using 1.25mm) being ahead for the larger die sizes for flow rates that are quite achievable.
However, I would suggest re-running with the finer and closer packed together channels and walls as I'm using. This will have a significant impact (it does for me) on the effect of the "desired" base-plate thickness. I can't stress this enough. It's the density of the walls and channels that makes the super-thin base-plate feasible. I agree that as we shrink down to well under 50mm^2 sized dies, that a thicker base-plate would be required, however as all future CPU's will be using integrated heat spreaders, the work will already be done.
|