Project: Omega [56k warning]

Note. This is done entirely off my own initive and is in a few places clearly quite stupid. In particular no one should ever put water inside an atx power supply.

Thanks to PhillyDee for donating a 50mm square piece of "Koolpads", a tim pad rated for 1kV, without which this project would be more dangerous and less successful.

It's been a long time coming this one, but I finally have some free time to spend on it so should make progress. That and a motherboard rma forced me to make some changes, starting point for what I suppose is a rebuild is this mess.



Most of what follows will have been planned somewhere on these boards, I'll link the source of the idea where possible. I hope you enjoy the read.


First off, a flycut ek supreme. I don't know if this was a good idea or not yet, concept [thread=18028104]here[/thread]. Flycutting is a far, far better idea than lapping where possible, as it takes about three minutes and leaves you with a perfectly flat surface.



The grey stuff stuck to the insides of my case is bitumen flashing tape, details [post=14198118]here[/post]. The idea is mass loading to make things quieter, I think it's fairly successful. The following photo is an internal shot of my psu from that thread. The big copper heatsink turns out to be live when the psu is running, something like 90V. Don't touch it to see if it's hot.



And here is the psu as it stands now. Spot the difference




Psu is causing various problems so has been ignored for today, I think the next step is enough pvc tape to cut mains ac down to <0.1V and then ground the tubes either side of the psu, test to see if heat is still removed through the pvc or not.

Today has been spent learning to solder and sticking things together with jbweld.

Here's my "new" 8800gt, with a d-tek heatsink attached and a maze 4 gpu block beside it. The threads on the gpu block won't play nicely with my fittings so I've epoxied a couple of compression fittings in place. As I don't know what shear forces epoxy will take before it leaks I've included short pieces of 11/8mm tubing with 90 degree bends on the end. The mounting holes on the maze are some strange imperial thread, so I've drilled them out to 2.5mm and will bolt it on using M2 screws and nuts, the Dtek heatsink conveniently using M2 threads for its mounting holes.



Next up are the 120mm feser radiators which will still be mounted at the front. I want to turn one of them through 90 degrees, at which point the bleed hole fouls the case. So here's the screw removed and the hole full of epoxy. Fingers crossed. Look for the grey circle. The fan arrangement is a 38mm san ace with gutted 120mm fans either side, the tubing wraps around and connects to the second 120mm rad. If I manage to get the psu to behave itself I'm hoping to run the entire computer off this one fan, undervolted. The tubing is a 190mm length of 11/8mm tygon with a 200mm spring wrapped around it. There's a 90 degree barb at the other end which I think is unavoidable.



I've put a fan controller of sorts together. 12V from molex is applied directly to the fan, and a locking rocker switch changes the fans "ground" between 0V, very large, and 5V. This has been tested and will indeed run a couple of fans at 12V or at 7V. It's ugly as sin though, the holes in the blanking plate were made with a very hot M4 screw (my new gas soldering iron has a blowtorch attachment ) and I wasn't coordinated enough to get them to line up neatly. Only one switch is wired up so far as I need to buy another molex plug.




Nasty feeling the next thing on the list is lapping the processor.

Processor has been lapped, mounted with "liquid metal", a gallium based tim which alloys with the waterblock and ihs surfaces, hopefully leading to lower temperatures. 120mm radiator is now sealed. Continuing to work with the psu, i.e.



I think I'm going to solder the connectors downstream of the psu to ground, testing with the wire visible. PSU barbs seem to spend most of their time at about 0.4V, some at 0V, and brief stretches at mains AC. I suspect the result of directly tying mains ac to ground will be a blown fuse, so I'm hoping that the water poses sufficient resistance for nothing to go badly wrong. As long as the connectors taking water to the motherboard are definitely grounded, I shouldn't kill any hardware. I'm using quick disconnects to keep the motherboard removable, so I think I'm going to solder these to the motherboard tray.

Reserved 1

I've now bought a sheet of copper 200 x 100 x 2mm, and one of acetal 150 x 100 x 20. Total cost of 20 quid, gambling on being able to make motherboard blocks for less money than ek are selling them for. If this works out badly there are some enzotech products which I really like the look of. Plans finalized, added 20 quid on enzotech copper heatsinks.

Northbridge block. The X58 has a TDP of 25W. It's quite possible that a flat piece of copper with water running over it would be fine for this, but for the first iteration I've decided to overspecify it. The first photo is of the enzotech CNB-S1L, this is a single piece of copper. I don't fancy buying a thick piece of copper and trying to mill it into shape when enzotech have managed to craft this and bring it to market for a tenner.



So the plan is to use one of these and a block of acetal, cut the tabs off the enzotech, mill out the centre in such a way that the enzotech sink will fit, and epoxy it in place. Initial sketch of this plan below. Drilling it will be trivial, milling should be easy enough. Assuming this works I'm looking at £12 in raw materials for this one, and probably an hours labour. A planned revision 2 would cost £3.50 in materials and has the potential to be much smaller, the idea being that only the die in the centre has to be cooled.



Discovered rotary tables, which I think opens up the range of things I can mill considerably. O rings are now possible.

Reserved 2

water cooled psu I like!"!

An interesting idea, not sure If I would be as brave as you though you.

First electric shock of the night. The big copper heatsink I thought I'd put a waterblock on turns out to be live when the psu is running, multimeter clocked it at 90V, after I touched it to see if it was hot. Nasty. Need to find out if a tube full of water can be at 90V at one end while safe for a processor at the other.

Suppose I should add a disclaimer.

Nice bet you are wide awake now

I like the idea going for some silence are we ?

Will follow this one, if only because you're mad

Crazy

Cheers all. With no water in the system and the psu running, all the connectors are at ground. If I pump water through it with the psu running, each piece of steel reads between 6V and 3V depending on distance from the psu. It's better insulated than before, I think electricity is only carried through the water at present. Next step is thermal pads I think, since they're probably electrical insulators. Can't see them coping that well with 240V ac somehow.

They didn't, turns out the 6V was due to poor physical contact. Adding thermal pads just moved the compression fittings from 6V up to 240V. Need to find a good thermal conductor which is able to stop 240V ac, might try electrical tape.

I genuinely do not understand why anyone would pump water through a transformer/mains supply. Not only is it daft it's also quite pointless since you're psu is noisy because it's cheaply designed/inefficient. A good (quiet) psu will have a grilled back, 120mm fan on the bottom and rate 80+ efficiency. Seasonic is the first stop (and only dedicated psu manufacturer) for such psus which incidentally can't be heard over the fans you would use on the radiator without electrocuting someone/burning do the house.

240V straight down the water from PSU to the CPU. Hrm.. might be a bad idea

Problem is that most materials that will conduct heat well will conduct electricity well.

Azuse05 said:

I genuinely do not understand why anyone would pump water through a transformer/mains supply. Not only is it daft it's also quite pointless since you're psu is noisy because it's cheaply designed/inefficient. A good (quiet) psu will have a grilled back, 120mm fan on the bottom and rate 80+ efficiency. Seasonic is the first stop (and only dedicated psu manufacturer) for such psus which incidentally can't be heard over the fans you would use on the radiator without electrocuting someone/burning do the house.

Click to expand...

I'm hugely entertained by your reply. The words turbo-cool 860 are clearly visible in the photos of the psu. A swift google should inform you that prices for this start at £120 and that it will output 64A at 12V, every hour of every day, while running at an ambient of 50 degrees centigrade. The manufacturer is pc power and cooling. What exactly would qualify as a well designed power supply in your evidently critical eyes?

Suffice to say that I'm doing this because I want to know if I can or not. My psu is already very quiet.

Trial using pvc tape between block and psu didn't go well, I'm registering my second electric shock of the night. Second attempt has taken the voltage at the compression fittings down to 0.4V.

JonJ678 said:

Trial using pvc tape between block and psu didn't go well, I'm registering my second electric shock of the night.

Click to expand...

It's not so much of a shock if you are expecting it, well that's the way i play with electrics

double post, somehow

Jon you're crazy mate!!!

Holy god....good on ya for have a stab dude.

But if there's a gap where you haven't posted for a while I'll check the obituaries.

ManCuBuS said:

Holy god....good on ya for have a stab dude.

But if there's a gap where you haven't posted for a while I'll check the obituaries.

Click to expand...

LoL!

He has gone very quiet come to think of it...

Hats off to the guy for trying stuff like this, I personally wouldn't want to be doing this type of thing myself, as I am extremely accident prone, and would end up killing myself.

I'm not entirely sure what the benefits of doing something like this would be, but it's interesting to follow regardless.

The base of that Supreme looks awesome though Jon, how exactly was that done?

Hey guys, Not dead, just learning to solder. Very, very slowly.

I'm a bit surprised the psu has generated so much fuss. I think it's very unlucky that the heatsink I chose happened to be at mains voltage, if that wasn't the case it would have been fairly trivial. As it is I'm going to have to do some testing on using pvc tape as a tim and hope for the best. Also of note is that I've been pumping diluted vinegar around the tubing which probably conducts rather better than water, so it may be less of an issue than I think.

Anyone know what the consequences of grounding the compression fittings would be? There would be a continual flow of current down the water to the compression which is probably ruinous for efficiency, and may encourage electrochemistry, but at least the cpu would be safe.

Thanks for noticing DavyBoy, you're the first to pick up on what I thought was the more exciting photo. A flycutter is a cutting piece mounted a couple of inches away from the vertical axis, which is spun quite swiftly in a milling machine. It'll only cut by a quarter of a mm or so each pass, but it leaves the surface very flat and doesn't take very long. Took four passes to get the ek looking like that. A thread on xs showed much worse results as a result of lapping a bowed block, rjkoneill has an ek which had this done to it and reported good results, so we'll see. Main benefit was getting all the corrosion off the surface so that "liquid metal" tim would wet the surface nicely.

So far I've glued some barbs into a waterblock which wouldn't accept the threads, that'll be set in about four hours and ready to use by tomorrow, decided on component placement and started making a basic fan controller. All good fun, will have photos up later.

Photos in the OP.

Bit of good news in that a ddc which WoZZeR donated me turns out to be a 10W version 3.3, therefore moddable. It's a bit noisy, Bubo's best guess is that the bearing is suffering from old age and I'm inclined to agree. The hope is two in series will outperform one alone, even if one of the two is a bit shagged. Will test this.

May have lost my chance at this through poor soldering skills. Any ideas how to tell if a DDC is running at 18W or at 10W? I have a multimeter, so guessing put it in series somehow. Mod unsuccessful, two loops looking more likely.

Looking forward to seeing more of this jon, (if you dont shock yourself to death) in the process that is. Good job on the home made fan controller, if id known you were needing one, you could have got my old akasa junior 3 channel unit, dont need it anymore myself. Fly cutting sounds like a great idea, done a bit of googling and watched this vid.

http://www.youtube.com/watch?v=koYGLVGkRsI

Very precise, and a much less laborious task than lapping by hand, plus a lot better finish.