The thickness that Cathar uses requires fins above that push against a cover. Without that, the baseplate would bow under the load resulting in zero contact over much of the die.
As for direct-die cooling, you'd find it far inferior to pretty much any water block. The die area is simply too small to allow enough convection under the most concentrated of flow streams. Let's also not forget that at some point even pure water will result in material erosion (not to mention being "unfriendly" to the CPU and motherboard). This isn't a big deal when you've got millimeters of metal to play with.
Yeah, you could make direct die cooling work with very high flow velocity (or very low coolant temperature), but for a given velocity/temperature you'll do better with a block.
Cathar's block works well because it has a very thin (low delta-T) baseplate that connects to sufficient surface area (the fins). Most blocks use a thicker baseplate to transfer heat both directly through the baseplate but also radially toward the edges. If you can connect enough surface area to the region directly over the die while controlling deflection, you'll get very good results. This is what Cathar has done and I personally find it pretty impressive.
|