Topic: Submersing a rig - page 2. (Read 5772 times)

Progress report #1.

Pur foam dried out and I started cutting :



As you can see im very good at measuring and sawing straight lines



I also underestimated the size of these cards, so this will be a single motherboard, 4 card box.

The PUR foam will need some sort of coating, it will always keep giving off particles, that cant be good. 

Good thinking. Would need a lot of nails but something like that should work. Or just rocks or bricks. Create a bottom layer of pu foam , then put whatever weight on it, then apply the next layers so its firmly embedded. Ill try that for V0.2 Now Ill just make some "bridges" in the PU that stick out above the PVC bowl and press it down with rubbers. Or something.

I like the idea of nails or maybe model weights over glue.  Who knows how well plumbers glue holds up over course of a year in hot oil.

Maybe stick nails into the foam to weigh it down?
Have you tried plumbers glue?

With my setup, a pump failure wont cause the card to run dry. The only real chance of running dry is when the oil leaks out in to my pond. I intend to install a floatswitch of sorts to shut the system off, possibly even close a valve, more to protect my pond than my shed . Even so, assuming oil cooling could manage to keep these cards around 50C, you could just define a shutdown point at 70C or whatever.  If there isnt a big difference between working oil cooling and no oil or static oil, then the whole exercise would be proven futile.

BTW, I also already have smoke detectors near my rigs. Of course that only helps if Im home. Still not a bad idea to grab one, they cost next to nothing.

People have nothing but problems w/ aircooling too.  Check mining troubleshooting thread.  I have a quad 5970 workstation water cooled for 2 years now.  Cold & Quiet 24/7.

This is about oil submersion.

Don't do it. I have a sli watercooled 6990s and its been nothing but problems. Running water and electronics non stop is not a good combination!

Im bumping in to my first problem. PU is lighter than oil. It doesnt stick well enough to the PVC container, so it will float. I guess that can be worked around by mechanically forcing the block down, but its not a good start lol. On the bright side, its a joy to dremel canals and stuff in PU foam. If I didnt have 2 left hands, Id start sculpting!

I was not concerned about the oil directly in fact due to the high flash point of oil I would say it is likely to put out a fire.

Consider this, you have intermediate fuels in close proximity to the card, PU foam, plastic wire insulation, plastic moldings etc all nicely heated up by the card. If you lose the oil flow all those materials are now lightly coated in oil and with the splash method you might even have created an air-fuel mixture. Your ignition source would be the card, it has a lot of current running through it, a fatigued solder joint, poor connector or faulty component and there is your initial flame. Normally in an air cooled pc you would only expect get a brown spot on the PCB or a blackened connector, maybe not even that much with the oil running, without the oil running the risk of fire is considerably higher.


Anyhow good luck with the project and don't forget to post pictures when you get it up and running

I hear ya. Among others, I still got an old (oil cooled!) GSX750F in my garage. Its no longer in driving condition, been thinking about lending that oil cooler.  But  Im worried it might be corroded and contain lots of metal particles. Think its safer to buy something new.

I'm a bit of a motorcycle gearhead, so I'm really curious about repurposing a wrecked oil cooling system for laughs.

All oil is flammable, so Im not sure what risk the PU (or fibreglass) would add to that. If it catches fire, my shed will burn down anyway. But in case the pump fails, the cards will just shut down, and long before the oil reaches flash point.

Anyway, while I appreciate the thoughts and drawings, Ive pretty much made up my mind. Unless experimentation throws up some unexpected results, I will be using a single PVC container that will hold the cards (im building the first to hold 6 with 2 motherboards), pump, oil and PU foam (if needed, coated with something, perhaps just some oil resistant primer paint). The oil will be circulated to the radiator in my pond filter.  There is no risk in it running dry unless I have a leak, or overflowing, and its a fair bit easier to set up, not too mention more compact and easier to move.

Ive just filled a tiny plastic cup with PU foam and vegetable oil that I will keep an eye on to see if it breaks down. I suspect it will, but hopefully only in the very long run. Once hardened, PU is supposed to be pretty inert.

The tray only has to be big enough to accommodate the cards and a channel big enough to deliver oil to the containers and overflow. Even then you can eliminate most of the surface area by raising what parts you don't need above the oil level.
                                                               
The pump control is easy just put in a bypass, place a valved connection between inlet and outlet, adjust valve to regulate flow.

I don't have any probs making fiberglass waterproof, try thinning the resin with acetone.

With this design I would be wary of the fire hazard, it would be bad enough with fiberglass I wouldn't want to go near PU foam. You will need a fail-safe shutoff switch in case the oil pump stops, you can make a relativity bullet proof one out of a mirco switch and a syringe.

I will. But dont hold your breath. Im really slow (and not even very thorough) when it comes to stuff like this

My current plan of action; make a mold of sorts in a PVC tub, that will keep me busy for a while; then fill it with vegetable oil, insert a single spare old ~100W nvidia gpu with no oil circulation or cooling (other than the gpu fan). See what happens. I want to see the difference between oil temperature and gpu temps. I want to see how fast the oil heats up, and if that is local heeat, close to the card or if the entire bath slowly heats up.

 If the oil gets hot but without causing the card to really overheat, that would be perfect and Ill keep it running for some time to see the effects on the oil, the card, and measure temperatures in various places in the oil bath. With low ambient temps and a fan if needed, I hope to get there. 70C or so oil temp would be perfect as a worst case test scenario.

If you do it, please post a picture here as well:
https://bitcointalksearch.org/topic/--7216

Me interested as well. One friend of mine tried submerging a power supply in oil, ended up heating the whole  bath because of no radiator.

That more or less boils down to what I intend to make, except, I dont like so much air. Too much chance of dirt getting in the oil. So fill most of the air in your drawing with PU foam or similar, and replace the freeflowing oil with some inserted tubing and its roughly the same idea. Only I dont have to worry about the pump capacity to maintain levels.

Im okay with fibreglass btw. Its messy, but I love the stuff. But its not as easy as youd think to make it water (well, oil) tight. Ill experiment a bit tonight.

DnT has pretty much nailed it with what I was doing. Cutting and welding metal for me is easier to do than any other material such as wood and even cardboard for that mater. It is perfect for heat exchange, in fact most of the difficulty working with it is due to heat exchange. But as you don't have the skills or equipment.......being such a small simple job it might be worth getting the local workshop to do it but that would prob exceed your budget.


Anyhow back to the idea you and DnT were working on. How are you with fiberglass? I don't like it much because it is messy and makes me itch but it is probably ideal for individual covers and it is cheap. You can make a mold out of a block of wood and waxed paper.

Expanding or simplifying the idea DnT suggested you could make the containers and fiberglass them into a tray made from ply wood. Have small holes in the base of each container and flood the tray with oil. The oil will fill the containers and drain out of the holes. Adding a large hole on the tray at a higher level will allow you to crank the pump up until it goes down this overflow thus giving you a uniform flow of oil through each container.

The real beauty of this idea is that if you ever give up mining, with some seeds, hydroponic solution and a grow light you have a ready made indoor hydroponic system and a brand new hobby, lol

I don't think Defkin intended to use pond water in the "water loop" because his post indicated using a pump and radiator he has. 
So it would be oil box -> water loop -> radiator in pond

Also you likely don't need as much surface area as you think.  Remember air is an utterly horrible heat transfer mechanism.  Radiators have massive surface area because air is so friggin bad. 

Thermal conductivity:
Air 0.025 W/(m*K)
Mineral Oil 0.138 W/(m*K) <- 5x higher
Water 0.600 W/(m*K) <- 24x higher

The effective difference is actually much higher because air density is so low it takes a much larger fan/pump to move the same amount of material.

So radiators have such high surface area due to two properties of air:
a) moving a lot of air is difficult
b) air is such a poor conductor of heat

So for an equal amount of working fluid you could get away w/ 1/5th the surface area relative to an air-radiator design.  In reality you could get away with maybe 1/20th because you could simply pump working fluid faster.   As a practical example in those links I had above for heat exchanger are flat plate exchangers and they have much surface area than an air-radiator design.

@Defkin
I dont have the metal working skills to make something like that work. I dont have any hand working skills really, so thats no surprise . I also think you have a lot of work to achieve relatively small heat exchange surface. And Id still be hesitant to pump pond water through it. All kinds of things are going to live in there and accumulate. Ever seen a string algae blossom?

@DnT and everyone
DnT gave me another idea with his "single gpu container" approach. Its too much work as he put it, but there is an easier alternative that achieves almost the same. Im thinking of making slightly oversized mockups of my videocards with some room for oil flow, position them in the PVC container, and fill it with polyurethane foam.  It will help keep the cards in place, drastically reduce the amount of oil I need, and can help directing the oil flow. I could also embed some tubing to deliver the cool oil right at the fan intakes and remove the oil at the exhausts. Its a bit of a hybrid approach between the video I linked above (splashing oil on the hot components) and DnTs suggestion.

Now Im not entirely sure if PUR foam will resist being submerged in oil for extended period of time. I fear it may not. I know water eats it away slowly (we are talking years, but still), so I may have to coat the hardened foam with something. I have some fiberglass and epoxy raisin I could use, though perhaps just coating it with some kind of oil resistant paint or sealant might be enough. Suggestions?

Im also wondering how to make the mockups and be able to remove them without tearing up the PUR foam. Or maybe I should just fill it with foam and carve out the areas with a dremel or something. Im not much of a handy man so suggestions are welcome