Watercooled PSU
 

I modded my PSU bladerunner-style and i already did a test-run.
I think one of the wires, i extended the mosfets with is to thin because i saw something glowing in there and it smelt a bit burned. The fan i connected to it to see if it works started spinning. What kind of current Ã*s going to the mosfets and what kind of wires would be needed ?

Yikes! Turn it off!

You probably want a fairly large gauge of copper wiring, something similar to a lamp cord.

On the mobo connector, each pin can carry 7 (or is it 9?) amps, and you can see that there are a few pins for the +5 volt supply.

In the PSU, the wire must carry everything!

mines fine with 16awg wire, never had a problem with it.. what size are u using? (oh and if its glowing, i'd keep a bucket of sand nearby and not turn it on again)

.. buckets for your next wise idea btw :P

neway, largest device i spotted in my q-tec 550 was 2x15A rectifier (i believe it was a rec.)

I used 6 gauge solid copper, for the power bus of my PSU array (3 * 300W). ;)

The current capacity of solid versus stranded cabling is different, from what I've read (not confirmed).

Mine work perfectly well :)

:D

I used a std issue PC power cord and striped it, and use that cable too make the Mosfet feet 10cm long :)

and here testing...

Hmmm - I have alot of worries about this method of watercooling PSUs. I mean - PSUs are engineered very carefully - and that extra impedance could present troubles, methinks. I'm not that much of a electronics guru though - more of a wannabe :p Anyways - has anyone ever tried making an actual replacement for the stock heatsinks? I've noticed those are normally grounded - would that be a problem for the waterblock to be grounded? Anyways - touch the wires (lightly - after licking finger) to see if it is warm. If it is - then replace it with a thicker wire. At least that would be my strategy. :)

yeah i turned it off immediately of course !
I used a cable of another PSU, the kind that goes to the ATX connector, so its probably a lot too thin. Can you guys talk in milimeters please ?

u guys got balls. Water on the cpu one thing, but water on the psu takes skill. Kodos for bravery. HOWEVER i would love to see a fan free system. Is there a single pump that could handle a psu, gpu, cpu, northbridge system? Perhaps 2 pumps in series? I have not considered that, but it would make a great picture and VERY cool!

The system is inn the works..

I´ll be using a 1250 eheim pump inn the system, and going for a parallel system innsted of the std serial stuff...

Kinda like the system that blade runner has... linky

Working on the blocks now... i´ll be trying something similar too leejsmith and his aksa based block. and using the base plate from a Athlon MP HSF, they have a copper plate that can be unscrewed from the base that will work perfect for a GPU and NB :)

-P-

Ok... here is an idea of mine I'm going to try out...

solder the mosfets off from the PSU. Turn the PSU around and solder them back on the other side of the circuite board... Maybe you will have to cut a hole in the PSU shell but as an advantage you don't have to enlongate the wires and therefore ther shouldnt be any affects on the stability of the PSU...

u´ll have too make WB that, will goes between the Mosftets then and fasten then too the sides :) hmmm..... sound like å good idea...

Only thing is you will not be able too fit the power supply again inn it´s intended place :)

hmm its kinda weird...
i took everything apart and all the wires to the mosfets look Ok.
No molten insulation or something like that.

Obasau:

Burnin :)

All the wires are placed right on the PCb right?

:)

Usually, there's some kind of visible trace.

The plastic insulation starts to deform under the heat, and eventually exposes a bare wire. If it's hot enough, it'll burn the insulation.

On the other hand, if the cables sit perfectly straight, the insulation will only shrink a little bit around them. The problem is that over time, it will become hard and brittle, and eventually flake off. The bigger issue though, is that it can melt the solder right off of the PCB, which can in turn create an open circuit.

You don't want to know what happens when all the power lines are ok, but one fails.

The Enermax 550 watt PSU I modified is now about a year and a half old since being water-cooled, (it's about three years old overall). In that 1½ years its been run almost every day all day without any fans or forced air-cooling. In that time I've had different people voice concerns about the issues extending the mosfet wires will cause, a Bang & Olfsen engineer was very worried and told me it just wouldn't work :p

Others have said it will make no real difference as long as the wires are kept short and the PSU is not run anywhere near it's full rated output. The beauty of being a modder at heart over an engineer, is we are willing to try things accepting the consequences of negative results. I was fully prepaired for it to not work when I started, but like I said it's running fine and has been for 1½ years water-cooled.

I have done some further mods, one to cool a coil that was getting hotter than I liked, the other to the case to help the general airflow and for aesthetics as well.

Those that have copied my solution, may find it wont always work, because PSU's all differ by make. Some designs may rely on the fans airflow to cool other components that are not so obvious as mosfets etc. Basically ALL current PC PSU's are poor by design layout and build whatever the make, (at least all that I've come across), and a clean sheet design is required imo..... something that is designed from the outset to be cool, powerful and has some quality in the componentry, layout and construction. I did think my 5v rail was a problem as it was dipping to a reported 4.62v in gaming, although system stability was fine. recently I upgraded to a new motherboard (from a Via kt333, Abit KX7 333-R to a Nforce 2 Abit NF7-S rev2), and I now have a 5v rail that reports no lower than 5.00v underload in windows software (MBM5) . This just goes to show you can't rely on windows software reported voltages, best to test it with a quality multimeter.

I have an update to the PSU build detailed on my site, but not the time to update my site atm. Contains some findings and solutions, if you want to read it you can download the update file HERE

The pics are not included in the file but the image file names are mentioned in the text and on my sites server running from:-

http://www.zerofanzone.co.uk/picture...su/psu_024.jpg

to:-

http://www.zerofanzone.co.uk/picture...su/psu_034.jpg

in numbered order.

Here is what the PSU looks like now, (except to unify the system with my new CPU splitter block, I've changed the PSU water-block to 8mm festo from it's previous 6mm).

http://www.zerofanzone.co.uk/pictures/jpg/psucase5m.jpg

BladeRunner: Nice work as always... :)

What have you done to the molex cables? They don't look like the use to...

an excerpt from the update download:-

Quote:

"over time a coil was getting quite hot and warming up associated components. This was in the area where all the ATX and device power wiring is soldered to the PCB. To help with this I decided to thin out this wiring in two ways. One was to replace all the multiple +5v, +12v and Ground wiring with much thicker single wires, (40 amp rated wire design for nicad power packs used radio controlled model cars). It was then spit to the normal wiring outside the PSU. I also removed the other unused power connectors, like the P4 one, and the old AT type".

I've since put the p4 connector back as the nforce 2 mobo has a P4 type connector ;)

I've got a fanless system, not pretty, but silent, and as they say, silence is golden :)

I basically screwed a waterblock to the top of the heatsinks in a generic PSU. Yeah, it got hot, really hot, but was stable for a 2 month uptime, then I had to reboot for a driver update.

In the end, I got worried about the heat and installed a 5V papst.... I was ashamed of myself ;)

One thing to remember :

PSU's can run at very high temps, I figure, without measuring, but my mosfets were at at least 70 degrees C. And STILL stable for a LONG time.

However, I had to watercool the RAM on my GFUltra, or I was getting crashes from that, because of the zero airflow, even the little bit from your current PSU fan helps a bit.

Worthless basterd! :p

I got a rant! I am looking at Pritorian's second pic with his system on the desk. What the heck is that about? Where are all the excess bits of hose and wire? Where are the TOOLS, where are all the empty coke cans or dirty coffee cups, where are the candy wrappers, how about all the misc. connectors that didn't get used or were wrong the first 3 times, screws, access sheet metal, scraps of block making material, METAL SHAVINGS? BAH, What kind of crap is that.... :p:D I know someone just like you, he happens to be a damn good engineer! :) Wish I had the patients to to pick up everything right after I used it.

Nice work! Same to Bladrunner. You guys give us wantabe's a reason to carry on! Now I am going to have dreams about PS water cooling.

Bladerunner!, The NF7s use the +12v rail to power the CPU, not the 'traditional' +5v rail, hence your boosted up +5v rail :D ...

Check out your +12v rail now, mines actualy better than when it was on the KX7, but I use an Antec truepower that has an independant 12v rail, not 'tied in' with the +5v. It has voltage sensing/correction on the +12v rail as well, which is a godsend for a +12volt based mobo :cool: ...

Ummm I still think it's more down to inaccurate measuring by the mobo(s) either software or moitoring IC. I had no stability issues with the previous mobo reporting 4.62v lows, and these lows were never reported by the multimeter. Funnily enough the +12v rail is now reported lower than it was on the KX7 333R at 11.98v low, never dipped below 12.5v on the KX7. I think all these measured parameters should be taken with a large bucket of salt, especially as they, (the motherboard makers), have such a job getting something as elementry as CPU temp reporting correct.