Cosmos 2 extreme water cooling system

iB CoMpL3XiTy · 11 Dec 2012 at 11:31

OliverS said:
Cheers for the pointer
I've only encountered that error once in ~19 years (~3-4 months ago ) twhen i ********ed up a few 2011socket pins.

Still. I suppose if i have broken the ram I cold get the new 8b sticks. My ramdisk always needs more space..

Ouch! that must have been a expensive mistake :/

My point about checking the RAM first is just because I think its more likely that the RAM would be damaged or dislodged from the slot and it may be slightly easier to check the RAM.

RamDisks look Epic from what I've seen and read, may go down that route with my new build.

OliverS · 11 Dec 2012 at 11:33

Yeah. Cost me £300 for a new msi big bang xpower ii.
Still need to contact msi to see if the sockets repairable. Doubt it but :/

If this new board is shot too then

OliverS · 11 Dec 2012 at 17:58

After a day laying on its side, the PC is now posting.
I have no idea what the problem is, that last bit is somewhat concerning.

Ordered £200 worth new kit to redo the layout with, primary goal is more flexible tube, lighter weight and same peformance. Still 1/2" id & 3/4" OD, reason being OC didn't do 1/2" ID & 5/8" OD when I ordered, and the negoiated discount was worth it to use 1/2" & 3/4".

There's a small air bubble in both my resesviors, nothing that would represent a leak on any pipes/connectors (one of the joys of sparkly fluid

) .

New fluid & parts incoming.

TLDNR!
tubing iteration #3 in soon.

OliverS · 12 Dec 2012 at 22:48

This is really bugging me.
It was working again earlier, booted up, loaded windows, and a game, then BAM!
turned off.
Now "Late CPU initialization"

BWoolley · 19 Dec 2012 at 00:13

Keep the updates coming

I'm fascinated with the idea of a custom water cooled system and I love seeing the solutions and ideas people come up with that makes it amazing.
Hopefully you can pinpoint the issue soon, this is an awesome big build that shows the process of creating your first custom loop.

Tattysnuc · 19 Dec 2012 at 07:40

Sounds like you've got issues with the LGA CPU contacts to me....

OliverS · 19 Dec 2012 at 22:41

Update:
Taken the cpu out the socket, checked it, cleaned it. Looks like a small amount of thermal paste slipped under the cpu. Not enough to cause any persistant problems, so I'm still boggled. (I pretty much just caked on that IC diamond while cackling manically, with lightning going off in the background). However, never had that problem before with an AMD socket, but then again, pins are on the chip there.

Worth noting that the IC diamond isnt electrically conductive.

Checked the socket.

Sockets fine, 0 bent pins.
CPU looks good. All gold contact points are unscathed and intact.

Tightened the cooler, hoping its

Flushed the socket with de-ionised water. Sockets drying.
New piping is all water tight. For the moment I've cut out one of the ram loops so it just daisy chains for debugging purposes. Easier to shift one side now.

Tattysnuc · 19 Dec 2012 at 23:14

I'm intrigued when you say flushing the socket - pray tell...

OliverS · 20 Dec 2012 at 11:00

Take mobo out the case.

Get some thermal grease remover, load it into a syringe.

Get some de-ionised water, as in water which has not only had the metal particals/salt removed but also CO2 etc molecules, giving an electrical conductivity of ~10^-12. (10^-8 or so for distilled water).
Load it into another syringe.

Squirt small amount of thermal paste remover at the paste in the socket, follow up with half a dozen syringes full of de-ionised water.
Contact pins are made of gold, so bugger all reaction from that.

Allow mobo socket to dry for ~3-4 days.

All PCB's get washed in de-ionised water after coming off the production line. This is just more local to remove paste from the socket.

OliverS · 22 Dec 2012 at 11:37

Just on my way back to cambs for Christmas.
Here's an update. so for its posting with just the cpu&ram in.
(well posting far enough to error about no gfx.)

Something Else · 22 Dec 2012 at 13:20

that looks very complicated!

how long do you think it will take before you complete it?

*Rich* · 22 Dec 2012 at 13:37

Could i ask politely, why is there so many bends and joints with numerous 'dead ends' ?

It does seem to look over complicated to me. (Not trying to sound nasty)

OliverS · 22 Dec 2012 at 15:18

There's 2 dead ends at the start and 2 at the end.
2 spare loops; except to add to these loops all I have to do is remove the sealant plug, screw the new connectors to the components and then open the ball valves. You can't just have one ball valve or how do you connect the exit without opening the main loop/spilling fluid?

It means I don't have to drain the system before adding any more loops in.

As for the curves? Snake connectors with 2 degrees of freedom. Great for reducing torsion in the pipes &thus torque. That's what broke the first iteration. The lower the torque the more stable the pc is while being taken to lans.

Most tubing is shipped and delivered in coils. okay I can see the logistical reason but when you try to straighten it, it has a tendency to curve.

Also now most the weight is anchored to the case (admittedly with cable ties at the top for now) which reduces strain in the mobo from holding all the weight.

1 loop is for the graphics, and the other is spare.

In terms of when I finish? That's dependant on the thermoelectric cooler and the graphics cards. At the moment the ends have no thread and are just cut pipes. Connecting to that is a pain. Solveable just busy at work.

Still not decided what graphics card I want. (1x 690 vs 2x680 vs waiting for 2x690.... Given the excessive nature of this so far, a bit of restraint doesn't seem in keeping).

Also if anyone noticed that blue tube? its a 1/2" id 5/8"od pipe.
not very flexible but great with angle connectors. its a different material to the 1/2" id & 3/4" od pvc pipe.

Tattysnuc · 22 Dec 2012 at 15:55

so, did the flushing the CPU socket work then? Sorry - I'm presuming that you've not tried witht he GPU's back in place looking at the last photograph.. is that right?

OliverS · 22 Dec 2012 at 16:01

seems to. before it'd error without graphics too.
Been forced to go Christmas shopping so im having to be patient
it's killing me.

Edit:
Got back, checked it, got a small leak in one of the snake connectors in the actual middle joint of the connector (i.e. not a part I've screwed in). Not too pleased. Had to replace it with a back up.

Still, new part is a 60' koolance snake M/F connector & a male/male adapter, old part was a generic male/male 60' snake.

****** generic parts.

OliverS · 24 Dec 2012 at 13:48

All up and running now. just running super pi to check that the OC is stable again, currently at ~4.6ghz

OliverS · 5 Jan 2013 at 16:15

bsod'ing again.

Ram's screwed or a dimm slot.

Ripping out the ram coolers for now, that loop is a massive pain in the arse and always the one which causes a problem/leaks.

For anyone which is considering ram water cooling.
Just dont.
Its a massive pain in the arse, unless your coolers are sold as 1 phyiscal block then its a pain.

I have 8 Koolance double sided coolers, and thats 16 connections for daisy chaining them together which can leak, plus one in the middle of the bridge. so ~20 connections.

Its a massive pain.
Those bloody things have litterly cost me hundreds. You think "yeah, that'd be awesome to do, lets do it!" but you dont think "oh what happens if i need to get to the ram? hmmmmmmmmm".
Answer? You flood the damn board as there's 20 stupid connectors to disconnect and no easy way of draining those blocks.

Just cool the northbridge instead.

Still, now I can reduce my pump speed and just have more cpu cooling.