[Cathar]

The cups are there to "manage" the water-flow, more than they are there to increase surface area, however I think that you will find that the cups offer more surface area where it counts over a pin design.

Jet arrays have been around for quite a while but only ever applied to flat or dimpled bases, but never completely "cupped". One of the biggest issues with getting them to realise their full potential are down to two things:

1) Adjacent jets interfering with each other - after the immediate impingement action the impingement "circles" clash into a turblent region that is basically doing nothing much at all as only a minute part of that turbulence is coming into contact with the base. Due to the clashing of these regions, this also serves to diminish the radius of the primary jet impingement region under the jet. The only real solution is to widen the separation between the jets to allow the impingement region for each jet to develop properly, but this then carries the drawback that as you move further away from the central impingement area the cooling effect of the jet is diminished. This is what jaydee116 was experiencing with his first multiple jet block.

2) All the "stale" water has to go somewhere. This typical means taking it past the path of other surrounding jets, interfering with their power and disrupting their direction, which creates extra boundary layers because the impingement regions don't form properly.

"Cupping" with the tubes partially inserted into each cups solves these problems. It allows a "private" area for each jet to form properly. The jet forms the primary impingement area. The cups width is such that the walls are positioned just outside of the main stagnation region of the jet. This is where the flow pattern is at its "thinnest" and just where the boundary layer starts to form outside of the primary stagnation region. Here the thin high velocity water flow strikes the cup walls. The water as this stage is travelling at close to the same velocity as at the jet nozzle. So we now get a "secondary" impingement effect that occurs as a ring around the base of the cup wall. This, couples with the main jet region, concentrates two highly efficient impingement zones both at the cup base and within 1mm of the cup base walls. It's important to note that the cup width is rather critical for this to all balance properly. Too wide and the water from the primary jet starts to slow down, and boundary layers start to form before the cup wall. Too narrow and the pimary stagnation zone cannot form properly, loses power, and instead of achieving a full effective impingement, results in more of a super-turbulent mash of flow, which while still effective at cooling, is not as efficient.

The cup height is dictated more by the optimal jet height, rather than anything else. The insertion of the jet tubes into the cups helps to establish a flow path out of the cup to provide less interference with the incoming jet stream, and the insertion also protects the jets from the collective outwashes of nearby jets as the outwash flows towards the outlet(s).

From initial appearance, the apparant drudgery of a few holes drilled into the copper plate seems very mundane, but only until it is understood what their exact purpose is, and that the actual height, depth and width of each cup represents a very exacting balance in conjunction with the jet tubes to ensure a maximising of the efficiency of the impingement principles used.

The separation of the cups is also important. We want enough copper to both maintain structural integrity when used with extremely thin base-plates (substantially less than 1mm), while also offering just enough copper to conduct the heat up the cup walls far enough to take full advantage of the secondary impingement effect on the cup walls.

The whole point of the design is that a vast majority of the heat is soaked up within 1mm from the CPU core, or even less. The White Water did the same within about 2mm of the CPU core on average when the fins are taken into account.

The cupped design allows for the simultaneous mix of ultra thin base-plates, high native structural integrity, and the ability to dissipate the heat even closer to the CPU than even the White Water could manage. IMO, it's about as close to direct die cooling as you can get, but gives you about twice the effective surface area to trade off against the thermal interface layer penalty. It would take about a sustained 80000 W/m^2K thermal transfer region applied to a direct die cooling scenario to exceed what the cupped/jetted design can offer even after taking into account the TIM layer, and I have trouble seeing how that high value could be achieved, not to mention the inherent dangers with direct-die cooling itself.