I wondered if an electrical spark from the mains could set it off. Easy to try in a bowl, I suppose. Not a problem under normal operations.
I wondered if an electrical spark from the mains could set it off. Easy to try in a bowl, I suppose. Not a problem under normal operations.
Can we get an update on the status of the original setup? Is that MB still functioning?
I'm extremely interested in building a hybrid, system. Where everything but the hard drives are submersed. Its too cost prohibitive to use SSDs for large storage needs.
I have 4 1U, 1 3U, and 1 4U servers running in the spared bedroom. The systems are running at around 36C, but the room is very loud.
I'm very interested in this technology for sound purposes, but I don't want to submerge 10K of servers in oil if the MB are going to fail.
Update please!
The original setup is still functioning and we are now running three setups at our office. None have failed at any time, and no hardware has ever failed. Pretty impressive, really. However, all of our original disclaimers still apply. We don't want any responsibility should your experiences be different than ours!
6 months back there was a debate about whether the cpu heatsink helped or hurt and someone mentioned they wish they had an engineer on staff. Although it's been 15 years since I took Heat Transfer in my undergrad (and I'm a metallurgical/mechanical engineer), I do remember enough that my horse-sense is based on math and knowledge of metals, and I can give you an educated reply:
Relatively speaking, when compared with motionless oil, the thermal efficiency of a heat sink is high enough that the surface of a 30 fin heatsink with 4 in^2/fin will be within 5-20 degrees of the cpu temperature (depending on a ton of factors, from the heatsink design to the cpu conditions). So you can nearly consider the surface area of the heatsink as an extension of the CPU itself if you plan to be 20 degrees or more below the max operating temp for the CPU.
What this means is that you can do an EXTREMELY rough estimate of how much an aluminum heatsink is helping by taking a ratio of the areas of a heatsink vs. no heatsink. This works out to 30x4in^2:1in^2, or 120:1. In otherwords, THE MATH SAYS YOU'RE TOTALLY GOING TO FRY YOUR CPU if you try no-heatsink, unless you blast the cpu with a stream of the oil (like a water-cooled setup, but with oil).
I'm guessing that the guy who aggressively advocated disregarding the math to just try this in the real world fried his cpu this way. It's been 6 months since he last posted. That's what he gets for thinking heatsinks insulate.
If however you plan to pump a good velocity stream of oil directly onto the cpu it should work fine. Like I said, that's basically a watercooled system but you're using oil instead of water. However, please note that it should be LESS effective than a watercooled system because 1) mineral oil will hold onto a lot more heat than water does, and 2) mineral oil won't be going straight to the radiator.
So if you're going to the effort to replace the heatsink with a pump blasting oil on the cpu it seems you should forget the oil computer and just use a watercooled system, but that's just me. To me the coolness of this is that all you have to do is pull out drives, and drop the rest in oil. When you start adding pumps and all it looses it's coolness to me.
The more I think about it I think it would be kind of cool to be able to see the CPU as it's operation, which as I said, you could do if you blasted it with oil. I'm not sure why it's cool because it's not like it's doing anything ... just the idea that you're looking at it while it's operating ... would be good as a demonstration for kids because otherwise they don't know what you mean when you point to this big heatsink and say: "that's the cpu".
I found a $12 submersible pump here:http://www.snakeestate.com/reptile-s...fall-pump.html
It only has 20" of head though. Not sure that'd be enough, but I bet it would (keep the tube short and aimed at a 45 degree angle).
This pump may be enough to feed an oil-fall (like a waterfall) as a cooling mechanism instead of a radiator. Sure would be cheaper, and a lot cooler if you ask me!
I guess you could make your own pump with a cpu fan and a funnel too. Hmmmm ... something to think about.
Okay, now I'm excited about using the cheapo submersible pump ($12) as an oil-fall pump instead of using a radiator ($200+). Compared to those indoor zen-like waterfalls, mineral oil is odorless, shouldn't evaporate like water, won't leave water deposits, and won't get moldy or slimy.
At just 30GPH for a 12 inch oil-fall length the Puget Systems 6 Gallon tank it would completely turn over the oil every 12 minutes. That should be plenty to keep the oil cool, if you use the right waterfall design. I thought I'd make one of pebbles embedded in cement, but now I'm thinking about a stack of homemade narrow-cut aluminum strips. Should be a piece of cake to do with my bandsaw.
Also, instead of a fish-tank I'm thinking of sinking the computer into an open aluminum-pan at the bottom of the waterfalls (better heat Xfer, much less oil required, so faster turn-over in the oilfalls, cheaper, and much more robust than glass or plastic). You'd still see the components through the top (rippled though because of the oil-falls), plus it's a computer masquerading as an artsy zen-like waterfall: priceless.
I'm also liking the idea of coating drives in silicone and submerging them to muffle the noise. Of course I'd back them up daily but honestly I don't think it's risky. Pure silicone should be 100% impervious to mineral oil.
Hey if anyone else wants to use these ideas feel free to do so (it would be awesome if puget systems made a kit like this).
Ps - - Note amazon.com has some pumps 2X as powerful at the one I linked to previously, for just a couple more $
Last edited by davea0511; 10-12-2009 at 01:13 PM.
I think he's onto something! The only thing lacking is color. If only it could look like a Lava Lamp, with backlit blobs of colored oil descending those falls!
I'm now thinking of just crimping an aluminum flashing (super cheap) like an accordion, and drip down that. Of course, the idea is to maximize surface area over a 12" drop. I may go with a larger pump and do an 18" drop. I have a metal brake, but the flashing is so thin you could probably do it with just a pizza cutter/ruler/rubbermat setup.
I suddenly got burdened with a big project though that will take another couple months so it looks like I'll have to put this temporarily on hold.
But I have everything I need ... except time. Will keep you posted. Like I said, anyone else wants to try this, feel free. I like the dripping lava-lamp-goo idea, but you'll need to find 2 immiscible dielectric fluids with the dye in only one of the two fluids - good luck with that.
I think this should be doable with mineral oil and an opaque Castor Oil (won't mix with mineral oil). To make opaque castor oil, mix CastorOil (32oz from luckyvitamin.com, $8) with TiO2 powder. Add a little Fe2O3 powder to make it reddish-orange. These are both electrically inert ceramic dyes (should stay that way in oil). Castor oil is very dielectric - used in capacitors and transformers
If all the pigment is first fully wetted by the Castor Oil hopefully it won't mix with the mineral oil. Then all you should need is a blacklight to make it glow (the TiO2). Castor oil should sink, which is good because then it won't cloud up with your mineral oil, just come out globy in the falls, plus, then all you need is a couple quarts if you put your pump in the bottom of the container to get the effect.
You might need to slow down all submerged fans to keep the fluids from intermixing too much.
Last edited by davea0511; 10-14-2009 at 12:33 PM. Reason: Oops, silicone is miscible with mineral oil (bad), castor oil isn't (good)
There are currently 1 users browsing this thread. (0 members and 1 guests)