[Gooserider]

Note that Starman97's project is a SERVER, not a personal machine. Judging by the reference to a 1U chassis, I'm assuming it's a rack mount, and therefore probably one of several machines. (or possibly driving a big drive farm)

I'm not going to ask him to divulge anything he doesn't feel comfortable with, but it sounds like a corporate application, that involves a big database and lots of users. Something like that, response time becomes a big issue, and there are lots of advantages to keeping as much of the data in RAM as possible. Another possibility that occurs is a graphics 'renderfarm' for movie special effects.

So yes, it's extreme for most apps, but not for some of the industrial strength computing that is so important to modern life even if those not directly involved in it seldom hear of it.

Assuming he can sell a WC solution (not sure it would be possible) I would think that it would be possible to make a series of the copper 'clamp' design I mentioned earlier, probably in some sort of series / parallel array which would deal with the heat issue quite nicely.

IMHO the big challenge in such an application would be to come up with a 'fool proof' method of changing out either the entire drawer or some part of it in a quick way that didn't allow for ANY leakage of coolant. The method would have to allow for very rapid service (such systems often have 'Five 9's' (99.999%) uptime requirements, which translates to about 90 MINUTES PER YEAR when the overall system can be down, this includes routine maintainance, upgrades, and so forth as well as repairs) by field service people of unknown (but must be presumed low)competence in the presence of nervous client management. Keep in mind that this kind of machine is likely to have a cost in the millions of dollar range, and be in a computer room with even more expensive hardware, and you understand why leakage is not an acceptable risk...

Gooserider