Quote:
|
Originally Posted by unregistered
well, I'll bite since jd no longer has your 'ear'
- an awful lot of blah blah here with no specifics
1) define flatness as it relates to cold plates
2) identify your source of data that reveals sub-potential performance
3) what 'should' the flatness be ? - and why ?
4) and what performance gains will result from using cold plates to your spec ?
|
First of all the whole thing was a misunderstanding caused by lack of precisssion in expressing myself.
1.Contact surface between CPUs die and cold plate should be as flat as possible/feasible.
2.First amendment please
3. See point 1, reason as many molecules of each of materials are gonna be in cotact with ever flatter surfaces, flatness 'level' is gonna be governed by costs only.
4. One is quite certian, cold plate will be lighter
dT and flux spikingg capacity is goning to be increased potentially allowing for higher overclocks (yeah, it is lots of blah, blah, blah without numbers support - so take it as you wish everybody).
Now, theoretically perfect water block of classical pin design resembles a hedgehog. Imagine swiffy's lates but base is a hlaf sphere with pins 'bit' curved. inlet at the top and outlet as a gutter on the perimeter (get a glass ball and run a stream of water down on it)
Reason? Equal distance from each point on the thermal exchange surface from energy source resulting in perfect isotherms layout and superior surface are to any of the current designs.
This is why Bill's new block performs so well compared to Cathars designs. Swiftech's block has much lower energy dissipation per mm^2 (jetted area) compared to Cathars design but makes iit up with MUCH larger surface on which such desnity is present making it extremely good low flow (read universal) design - overall winner.
Trully elegant engineering solution, no fireworks but good solid conceptual work transpiring into extremely low production costs (pleae, do not read any brown nosing into this, it is sincere and I have no business saying one way or the other).
Anyone want to attempt spherical block with 'gutter' return?