3M Novec 7000 Group

This is awesome! Is there any evidence that the HFE interacts with any of the insulators on the board? This reminds me of some of the oil immersion experiments and they always had trouble with the electrolytic capacitors seals being degraded by the mineral oil.

Keep up the good work!

I've used the original 3M solution.
The end result is the cost and need to keep a sealed hermetic environment does not outweigh the benefits. Ultimately air cooled servers are more efficient and use less space in 1U rackmount configurations without the need for a condenser. Increasingly, for overclockers, it serves less purpose as it is not a good working fluid, and is prone to gelling. So you get water results at many times the price, difficulty, and encumberance.

It's fun and a cool idea. But also, insulation for sub zero methodologies has progressed incredibly in the last five years. Insulation is simple and does not place permanent marks upon motherboards and rarely results in damage with minimal preparation, something far from this configuration.

Sure is the power recoup nice? If you can produce high enough pressures for a regenerator, yes. Is it viable? Not really.
Also the fluid is often used with masks, as the effects of the fluoroether long term on lung tissue is not yet known or tested.

I get your pushing your product, but you might not want to be so callous. You might get input from the throw arounds (throw arounds; I mean, are the people in this section who post "can I put my computer in my fridge" topics five times a week), but the people with thermo degrees and years of experience might chime in if your more hospitable. As to your first claim, your system does not appear to be smaller then a 1U rackmount.

So let's hit your post shall we?

First
http://multimedia.3m.com/mws/mediawebserver?mwsId=66666UuZjcFSLXTtlXftMxMVEVuQEcuZgVs6EVs6E666666--
Novec is based on 3M's fluorinert, with a slightly lower boiling point to serve as an R123 replacement in autocascades and "blast injection chillers", mainly due to it's lack of oil pocketing that is seen with r123. So it's pretty close to the same thing. Almost has the same visocosity pattern. The Novec is a little less so. But following its chart in the above manufacturing PDF, by -40C it wouldn't be doing any good.

So, 40psi to power a generator?
Not at anything more then theoretical gains, no. Most of the power you could harness would be in it's heat. A stirling engine and flywheel or static pendulum. That'd help you harness. But, neither of those systems are going to keep removing enough heat that you can simply say you aren't passing heat off to the environment to recondense the working fluid. So it's not replacing AC units. It's simply a separate medium. Same heat load is being applied to the environment. Yes some could be harnessed as energy, but for the most part, no. This isn't the land of theoretical no friction physics.

Also, 40psi is already more pressure then you'd want in one large standing tank, much less a server room of thousands of racks.

Back to the post.
You are fully cooling a server. Every little piece. Which is great for server life, if servers were prone to failure due to heat. But they aren't. Servers are designed and use the highest binned parts, and often run on forced air cooling alone. They're are no actual gains from cooling the mosfets to ~30C. There are no gains from cooling the ram additionally. There are no gains from cooling the power supply. These systems as is, meet their specifications, as they are produced; with air and fans and for the most part, simple aluminum heatsinks.

I'm not to worried about the health risk. So what the NAVY does isn't on this forum. Isn't being pitched, and isn't what we're discussing.

Servers are replaced before they fail, at their rated hour limits, sold and passed down the supply chain or retrofitted, and replaced. Your system does not actually provide a tangible positive effect as compared to the standard of the industry. Yes you added a working fluid that doesn't actually harm the chain. But it's simply another working fluid. If your point is that its easier to vent the heat from the building, it's probably not when you realize your going to have to refit all those buildings. And the cost of your fluid, watercooling systems could easily be employed if a medium is needed.

Additionally in working on your "generator" theory behind the pressure. You have no ability to create a pressure gradient and restriction. So you will end up needing to employ pumps in large installations, to and from the condenser.

I am approaching this in a friendly fashion to see if you've worked around the design limitations. Please respond as such. Ad hominem will get you no where. Mostly because I'm sitting beside a tank of R123 that Novec is designed to replace. And have worked with immersion systems for many years now.
Edited by NoL - 2/25/12 at 10:10am

Just to keep working on this then...
Servers will not be overclocked. Even at our "stable overclocks" CPU's and RAM inherently produce errors on the order of one cycle per hundred thousand. This is the main reason for ECC registered RAM. When they need more power, it's more lateral power, and they add more units.
The 40W peltier on your unit is not cooling the server on it's own. That's a bit of a silly thought. The heat has to leave the system. It's doing so through normal heat transfer through the peltier and other areas. Additionally, capacitors do not enjoy pressure.
While the CPU heatsink would reduce height, you still have RAM and other not so thin items. 1U is as thin as it is going to get.

And if I were you I'd pop a heatsink on there, little reason not to improve the results if it's within the standard form factor.

The point is though, you have created an elaborate system to do something that server could already do for it's entire lifetime. Server hardware doesn't overclock. Doesn't want to overclock. It just wants to live as long as possible and with as few issues. It's designed under high precision with high bin componentry. You have reinvented the wheel so to speak. Ultimately all heat produced needs to leave the system, the environment, and that means out through the change of gas back to liquid. The same loads, if not MORE, are going to be applied due to heat removal from areas that did not need active cooling.

There's no reason for it is the end effect.

wow. good discussion and interesting results. I'll never be able to use such a system, but I enjoyed the video