Scratch Build – In Progress Project RabidCicada (Mineral Oil Computer) - 20th October 2010

I've searched but can't seem to find the specific one your were talking about. But I did find this: http://www.maximumpc.com/article/features/hardcorepc_reactor?page=0,2

It looks like they use the same idea of direct pump liquid over the hottest components.

 

Yea same company.

 

I remember looking at those/that company a few years ago. I found a video that they posted of them doing a shotgun test on their tank, literally. It was pretty sweet!


This case looks pretty sweet, I wanted to do a mineral cooled pc back in the day as well, but never had the funds

 

What i meant was 'horrible' for going full min-oil. You could get better temps on water with that rad. Obviously might be more expensive due to water blocks and not be as fun as saying you have a submerged PC or the fun of building it, but if youre running stock at those temps you wont have much headroom to OC.

I think a simple plenum system with tubing and brackets would suffice so that the flow get directed over the components or heatsinks. You could even plumb in some ball valves to control flow to the main pieces (CPU/GPU) and have a 'main circuit' for the ret of the tank.

 

Hehe.

Yeah, I forgot about that reason. I wanted a system where I didn't have to buy new waterblocks for each new component I added to the system.

It'll cost way more up front as it already has....But after that I get to say "I just buy off-the-shelf components with no additional kit required!"

And indeed a large part of my motivation is also "Look--->. It's kool."

Can't wait for PAX East. I'm taking this thing there...I hope they'll let me set it up. Paid for the Bring your own computer (BYOC) fee already. I didn't tell them I may trip breakers though , or that I'm bringing in 20 gallons of mineral oil . I will probably add an internal power strip because you're limited to two plugs in their LAN section. Need pump to plug in somewhere so internal strip fixes that.

 

you really, really need to do some research before dumping on someones project like this. its quite clear you have no idea what your talking about.

first, those temps are actually quite respectable for an oil submerged system. mineral oil has HORRIBLE thermal properties, and so these systems dont achieve impressive temps.

comparing a mineral oil system to a full water cooled system is absurd. they are not even in the same class. water cooling will always out-perform a mineral oil system by a long ways.

and the liquid used in that system that was running no heatsinks was not mineral oil. it was a special and expensive fluid called novec 7000 which is electrically non-conductive and boils at something like 34c. see below:

 

That is awesome.

I like that stuff but I'll finish with the mineral oil first.

Soon as I receive the LED strips I'll test. I might also go get some smaller pumps to try the pump-blower approach instead of the big fans (but that is lower priority right now)

 

A quick read on someone elses forum says it sells for about $250/gallon

$5000 to fill my case with that. I like it and all....but not that much. Especially considering I'd still need to get more stuff like an exchanger and make the case air tight for proper operation.

In addition I have a fairly large thermal load compared to the Novec system. He had an I3 and no GFX card. You'd need a LOT of exchanger capability to properly condense the Novec in my system with 3 gfx cards, large PSU, and hotter CPU.

 

i told you it was expensive

ya its neat stuff, but completely unreasonable for the home user. its intended for things more akin to supercomputers and the like. IIRC, NASA used it aboard one project or another that they launched a while back.

 

Just a quick update:

Got my LED strips today from Xoxide. One strip was DOA. Someone broke the plastic covreing and a little circuitry on the strip. Got an RMA email out.

I took the Orange/Amber LED strip that was working connectorized it and put it on my fan controller. And tested around a little bit.

The lights are as bright as I expected but their light projection is narrower than expected. I can't really get a glow from behind the motherboard unless I have the LED's facing directly to the side (literally pointing sideways from behind the motherboard).

As a user of the computer thats not a problem but anyone who's near it from the Fan controller side or the opposite side will get an eyeful.

Here's my connectorized tester strip:


Here's the strip lit up on the lid:


I have a video I'll put up later of dimming and lighting it. I have more testing to do and more positions to try...but it looks like getting good lighting is going to take a little more "engineering" than I planned.

 

some sort of diffuser in front of the LEDs should get the light to spread out more and give more of a glow instead of focused beams. frosted plexi or the like ought to do.

nice keyboard BTW

 

Let me know if you have any trouble with that i know the general manager well.

 

No problems whatsoever that was the easiest RMA I've ever done.

Them:Would you like a refund or a replacement?
Me: Replacement please. Trying to test color and placement etc. Thanks.
Email:*Shipping notice*

It probably helps that I included an actual picture of the broken one. But that was quick!

and perhaps Furball Zen?

 

Well i brought the mineral oil computer upstairs to drop in my 2 gtx 580s. Found out I'm going to have to remove the baseplate so I can remove the fan. I need to get rid of the fan so it doesn't block oil flow from the afb1212ghe.

Got a t6 torx screwdriver, fujipoly 1mm thermal pads, and some courage on order.

Already have ceramique from earlier.

Also have longer dvi cables because the case is to tall to fit under the desk and too big/awkward to fit on it.

 

GFX card in. Running Fulltime. More to Do...

Quick Summary:
Modded 1 gfx card and put in oil computer.
Running full time on oil computer now.
Did LED placement testing with buddy and 2x12in strips. Ordered 24in strips 2xGreen, 2xAmber.
Temps not where I want (guessing and testing below).

Gfx Card Mod

Alright. I got the thermal pads and modded one of the gfx cards to go into the oil. I realised that if I was going to use the gfx card in the oil I'd need to removed the onboard fan to avoid obstructing oil flow from the afb1212ghe 120mm fan. After prepping both cards to remove the fan I found out that I have to remove the Heatsink bracket and undo 3 screws....I was only willing to do this on one card and test in oil for a while since I've never modded gfx cards before. Turns out it's extremely easy with nothing to worry about.

Gfx cards shroud off:


Gfx card workspace prepped:


Getting to work:

HS off bracket still on:




HS and bracket off:






Fan removed:


In tank (Sorry for the horrible picture):


LED Testing:
After I modded the GFX card I got to the LED testing. I don't have all of the testing because I forgot to take pictures but I did get the best placement and one other. I'm going to go with 2 green and 2 amber led strips slid down inside the channeling of the aluminum bars and pointing into the tank just like in the pic but with a longer strip (24inch). I was surprised that green and amber went as well as they did. The idea is that the 24inch will fill the full height of the case unlike these 12inchers.






Small video:


Temperature Testing and stuff:
I'm not happy with the temperatures. I think the ultimate reason is oil/heat exchange rate. My temps are now hitting 55c for processor and 58 for gfx card (Folding@Home) with only one gfx card. This leaves no room for another gfx card let alone 2. I thought my problem had to do with air flow through radiator or oil flow through radiator. So I modded 3 more fans to add to the radiator and put them on.

Rad with 6 fans:


Whole testing setup:


Adding those 3 fans didn't noticeably help so I decided to finally do a real flow test.

I took the radiator output and had the pump actively pump out the oil through the radiator into a jug, then without the radiator in the loop. The results are abysmal for both .

With radiator inline: 5 gallons in ~8.5 minutes.
Without radiator inline: 5 gallons in ~2 minutes.

The 'With Radiator Inline' test looked like a heavy drool coming out of the tube when I lifted it out of the jug.


This means that with my radiator inline I am pumping less than 1 gallon a minute. If I estimate I have 17.5 gallons (excluding volume for computer parts etc) then it takes 29.75 minutes to exchange all the oil in the tank. My goal, maybe naively, was to exchange the whole tank in about 2 minutes....

I really think that heat exchange due to oil flow through radiator is my problem with the whole temperature setup right now. I just gotta figure out what to do about it.

 

How did Legoman666 get such good flow out his stuff? Or performance if not flow? My guess is that my Mora radiator and quick disconnects are generating a ton of effective head pressure for that pump.

Gonna have to work around it. I think the radiator has plenty of cooling power with 9 fans but I'm just not circulating enough oil through it.

 

I never actually measured the oil flow with the radiators in place. I might have a video showing it though.... *looking*

Nope, no video. But I do have 2 pictures showing the force at which the oil is expelled from the pump
http://i.imgur.com/dW5uN.jpg
http://i.imgur.com/g3D4E.jpg

The oil is going through both radiators in this test (normal operating conditions). Looks like it's being shot up about 1 or 1.5". Knowing the diameter of the tube is 0.5", I'm sure it's possible to calculate the velocity at which it's expelled and from there a flow rate.

Or, here's test data from earlier:

 

calculation the flow from a few known variables and a picture....i wish i could math that hard

rabid, a suggestion for the LEDs: if they would fade slowly on-off by themselves i think it would add an awesome ambiance to the whole case. im sure someone here could whip you up a simple PWM control circuit to get them to perform.

how large are the passages inside the radiator? depending on the style of the radiator it could be the rad itself thats presenting the restriction.

 

Talked with a Varna Products engineer over lunch. I really need to instrument my radiator and certain spots in the tank with temperature sensors.

I was hoping I could use their XD4 pump: http://www.varnaproducts.com/XDPump.php
I found out that it can't be submerged and I suspect, after talking with the Tom from Varna, that it would be too loud. They actually tried to take sound measurements but they were doing it outside and traffic noise from 100ft away was overwhelming the sound meter.

I plan on putting temp sensor in the inlet and outlet of the radiator and probably one on the "output" side of the cpu heatsink.

I'm requesting samples of the temp probes used by crystal fontz...

 

Dallas one-wires. They're cheap, once you know what they are.