How to Repair a G5 Quad Dual Pump LCS

[mePy2]

Thank you @BadGoldEagle

Anyway I saw the screws have a "limit” (you cannot tighten them in infinite) and that’s the reason why I asked. I’m thinking maybe they were not tighten enough.

 

About the o-rings... yeah but the pressure is between the copper and the little aluminum plate. Which is "unusual”? Strange anyway. But I’m thinking LOL if they were put there for assembly only.

[360alaska]

8 hours ago, mePy2 said:

Mhm... yeah... are you sure about it? And then, why putting them at all?

(I found they are useless too. I mean, I cannot see their functionality)

 

Another thing, when putting the screws with the springs (the one that attach the CPU PCB to that metal plate behind the copper liquid thing) should I tighten them until the screws stop? 

I didn't say they had no use, I said it helps assembly when attaching the waterblock to the CPU.

[mePy2]

4 hours ago, 360alaska said:

I didn't say they had no use, I said it helps assembly when attaching the waterblock to the CPU.

Yeah, sure... but I expected something more "important"...

I’m going to replace them anyway though.

[mePy2]

Guys, another thing... what can I do for these?

 

Ps:

do you suggest me to change the thermal compound on the North Bridge too? How can I check its temperature?

[mePy2]

@kby You wrote about testing the LCS connecting it directly to its wire. What about those holes in the CPUs PCBs? I mean they are always power related... I do not want to make something wrong. Thanks

[mePy2]

Just sent the same post two times... apparently I cannot delete it...

 

Since I am here... would someone explain me “what I have done” in the picture above? It was not me, but the old owner. I do not like it though. I am worried since it is the CPU PCB itself.

On 11/12/2019 at 2:00 PM, mePy2 said:

Guys, another thing... what can I do for these?

 

Ps:

do you suggest me to change the thermal compound on the North Bridge too? How can I check its temperature?

Edited November 15, 2019 by mePy2

[mePy2]

Hi everyone, what is this blueish-green thing on the copper? (Oxidation?)

 

Also, what can I use to remove the  glue? And what kind of glue it is?

[mePy2]

Hello again... I wrote a long post until the page refreshed. I’m very frustrated now...
Anyway I made the disassembling.
What I have now is this:
• the liquid was like water, no color at all...
• there was little liquid inside
• no @BadGoldEagle orange little balls
• what kind of hoses they are? Seems very strong and made with two materials. The inside is something like plastic, and outside rubber.
• What kind of glue it is? I need the same glue.

That’s all...

[kby]

On 11/15/2019 at 2:41 PM, mePy2 said:

@kby You wrote about testing the LCS connecting it directly to its wire. What about those holes in the CPUs PCBs? I mean they are always power related... I do not want to make something wrong. Thanks

Sorry; not on here a lot. Presumably you have figured this out, but those are the connectors to power the CPU. They won't power the pump. You won't even have the CPU cards attached to the system to do the heat gun/hair dryer test.

The glue appears to be a thermal glue. I have not messed with the glue down by the blocks. I have had some on the pump connections that had to be replaced to removed the hoses; I replaced it with some organic-based (toluene solvent) flexible stuff; usually goes by the name "Goop"

(I know, really specific, but that is the trademarked name it is sold by). The old stuff will chip away fairly easily. It appears to be hot glue gun glue to me, but I don't know for sure. I have never pulled out the CPU blocks. If it ain't broke, don't fix it.

 

You can think of the assembled system as a sandwich for each CPU where the top piece of bread is the CPU and the bottom one is alumnium plate. The inside of the sandwich is the CPU cooling block. THe spring-loaded screws that go through hold two pieces of bread together tightly around the filling.

 

I don't prefer automotive coolants as they are toxic (not that we should be drinking them here anyway and shouldn't really be needed. The CPUs will shut down around 80ºC so the coolant doesn't get that close to the boiling point of 100ºC. Combustion temperatures in an automotive engine are much higher (~230 ºC) which is much higher than the normal boiling point of water. Such systems in addition run significantly pressurized. LCS for computers generally run pretty close to atmospheric.

 

The corrosion inhibitors in traditional antifreeze are probably more atuned to protecting iron rather than alumnium, although these days most include some aluminum corrosion inhibitors. Again, you're not dealing with all the ugly byproducts of closed combustion, so this is kind of a moot point for a computer LCS.

 

I would not worry about the look of the CPU board connects. They are bolted down with a star washer than bites into the surface, so they'll always look a bit ragged. Worry if there's no metal or the metal is coming off.

 

Commenting on the original poster's pictures, I have had bad luck with tygon (transparent vinyl) tubing. It looks good and I would normally think it would hold up better, but in practice it didn't. It leaks more as it gets soft at the upper end of the temperature range.

You can't clamp it down as tightly either, as it deforms and/or perferates.-kby

 

 

[mePy2]

Thank you for the reply.

 

I am pretty sure the connection on those holes anyway happen behind the board. The screws make it sure the pcb and power sources are well tight.

[Byrd]

Hi all,

 

I've a never serviced before Quad G5 - single pump revision 1 "Delphi" unit.  It's been a long time since I fired it up, I recall it ran OK but best with the G5 laying on it's side, which to me suggested low coolant.

 

I've pulled the water cooling unit, and can hear coolant sloshing around inside.  It is not leaking - there is no evidence of any wear or corrosion, it is spotless.  It may have been an Apple recalled unit?  To avoid complete disassembly I'm tempted just to clamp on some longer tubing and flush the coolant, or at least see what comes out.  Repasting the CPU might also be a good move during this process.  I've a spare Laing D5 pump (12V from PC watercooling days) that might help here.

 

- Are there any filters in the tubing to check - if so, where?

 

- Does the internal of the CPU block "maze" get significantly clogged - or do you reckon I just leave it be?

 

Thanks

 

JB

[beachycove]

There were two in mine. As I recall -- it's been quite a few years -- they were at points where liquid flows into the radiator. I have, however, removed those filters (they were pretty solidly clogged when I went at it) and replaced them with a single small in-line automotive fuel filter before the pump, which will be much easier to service when the time comes, and it seems to me much more forgiving just because of the physical size of the filtering. (To access those two small filters, btw, I had to cut away tubing with a utility knife, clean off sealant, etc.)

 

Yes, the mazes get clogged, noticeably. At one stage I went so far as to remove the mazes as an experiment because they are a failure point (did not go terribly well, the mazes matter, though the machine did run okay-ish without them). Once the thing is apart, however, the mazes are easily dealt with, as it is all just compression via screws around O rings, so I would not worry about that end of the matter. The most difficult thing is getting the radiator properly refilled. That can probably most easily be started once the mazes are exposed, through the open circuit at the heatsinks. All the same, it is impossible to get the air out at the initial stages of the refill, largely because so much gets trapped in the fins of the radiator, in the pump, and so on, and because you are dealing with airlocks in the system.

 

To deal with that problem, I installed a capped length of hose at the high point on my circuit, coming from a T junction, which allows me to get air out and more coolant in (I use a syringe) once the thing is running. On initial startup, you get a great deal of foaming in the system, but if the air bubbles have a way to get out, get out they will. My high-point T-junction solution does that job pretty well; the bubbles release air as they pass and it comes up the tube. It takes a couple of hours for the system to clear of all air after a refill in this way.

 

That sloshing in yours suggests strongly that the system is only partly full, but the greater problem is more likely to be sediments in the filters/maze and dated heatsink compound. I don't think the liquid ever gets very warm, so low levels should not be the main factor. Just my opinion, mind.

 

I've had mine apart now something like three or four times. I think I'll soon have it down to a fine art.

[Byrd]

@beachycove Thanks for the additional info, much appreciated.  Agree that the sloshing may not just indicate low coolant; I'll be interested in seeing what crud comes out.

 

Sounds like it's a case of "do it properly" then make some minor improvements in tubing, coolants and double check all so that when servicing is required in a couple of years it won't be as arduous.  Obviously wasn't hoping for that and was feeling lazy

 

Thanks

 

JB

[Byrd]

An update, I tore down the G5 revision 1 pump, drained liquid (~ 150ml all up), filled with vinegar to flush and primed it to cycle through.  The existing coolant was without any sedimentation, and the CPU blocks completely spotless.  I think the pump is the issue - it seems to be stuck, with fault and unable to spin up properly which is suggestive of the pristine condition of the tubing and blocks!

 

Looks like I'll have to chop off the tubing and start fresh.  I think I've a spare Laing DDC-1 pump in an old PC which should hopefully be a direct swap.

 

JB

[PowerMac_G4]

@Byrd How long did you spend flushing with vinegar? Did you shake the loop vigorously while it had vinegar in it? Did you use a mix of hot water and vinegar or just vinegar?

 

It is highly unlikely that there is no sediment at all in there. The fact that your pump is faulty suggests to me that there has in fact been some decomposition. After fifteen years it's pretty much a chemical certainty that the Dexcool has decomposed at least a little bit. The issue is that it coats the inside parts of the system that you can't see and then flakes off - it doesn't just turn to sediment. It takes a really long time to get all the stuff off the insides of every part of the system. I'm talking an hour or two of shaking the LCS with vinegar in it. I ended up draining and refilling with vinegar about ten times.

 

I was in exactly your situation when I cleaned out my dual-pump Quad. I thought I had a very lucky system. Draining and flushing the first time around yielded no detritus at all yet a few weeks after rebuilding the machine my temperatures skyrocketed. A second, far more thorough flushing brought a whole lot of schmutz out with it.

 

The DDC pumps are easily cleaned if they get clogged up - it's a fairly brilliantly designed bit of kit and very easy to take apart and inspect.

 

I don't mean to rain on your parade at all – I'm just warning you to be cautious and maybe take a second run at it before you start swapping pumps or reassembling it. The LCS is a tricky beast that takes a while to get under control. 

[Byrd]

Thanks @PowerMac_G4; I shook that thing for a good hour, releasing bubbles and it’s all come out clean. Used some hot water and “double strength” white vinegar.

 

You are right - there must be something big blocking it up as there is definitely some resistance of water flow through the loop. I think the pump is struggling to get enough water to flow through.

 

Since my post I’ve chopped out the tubing and will replace with clear vinyl 3/8”/10mm and new hose clamps all around. The filters (just outside of the inlets to both CPUs) were spotless. The PC watercooling pump I have is a Laing D5, not compatible but will help me in flushing the loop. Should the pump also be bad I can possibly incorporate this (it

has an integrated rad head so too big in the enclosure), or a compact AIO system is another thought.

 

thanks

 

JB

Edited May 18 by Byrd

[PowerMac_G4]

@Byrd Well colour me impressed! You might just have the cleanest Quad ever found!

 

Good luck with the rest of the build, be sure to post some pics when you're finished. One of the fun things about servicing Quads is that everyone ends up with a slightly different system at the end.

[mg.man]

36 minutes ago, Byrd said:

Should the pump also be bad

 

I probably have at least one, maybe two "spare" LCS units... neither have been run for a very long time... but could be useful for parts... Unfortunately, they're in the UK = other side of the planet to you! 

[mePy2]

@mg.man Are they version two? I would like one/both !

[mg.man]

Hmm... they seem to have multiplied while in storage...  I have four!  I don't know about V1 vs V2... three are single pump, one is dual pump...

 

 

 

I'm pretty sure the one with the single processor card is a dual-2.5Ghz... i.e. from a 'Quad... IIRC, I used this for testing the logic 'boards... using the Apple HW diags to bypass the "missing" CPU error, running the machine as a "dual 2.5Ghz" machine... 

 

As I said, none of these have been run for donkeys... so YMMV absolutely applies!!  @mePy2, PM me if you'd like to discuss... I'd like to hear from @Byrd, since he might just need a pump (which should be easier to test)...

[mePy2]

ok

[Byrd]

@PowerMac_G4 That's true about every watercooled G5 being different once rebuilt; also in current times there isn't much availability locally (Australia) for replacement parts, and PC watercooling is a dead hobby with decent AIO units now.  I've purchased some Thermaltake 3/8" soft kink-resistant tubing, Thermaltake blue coolant (had to happen - with the internals being blue as well). hose clamps and fill bottle.  I used to enjoy watercooling various PCs so this will be a fun enough project.  This Quad G5 came from a post processing house in South Melbourne, it appeared to have had a hard life but reckon the LCS was replaced at a time by Apple not long before the computer was shelved.  Be good to see it up and running without concern, DDR2 RAM is so cheap now I'll upgrade it to the max and put in an SSD.

 

@mg.man Thanks for the offer, they are multiplying!  I'll see how the rebuild goes and if the pump is shot, I'd love to take you up on your offer.  I hope the pump is OK, just low on coolant/flow to operate properly.

 

Thanks

 

JB

[Byrd]

More tests - with all parts lined up ready to fit, but not there yet.

 

The Laing DDC 3.25 18W water pump has failed, or was always no good   Wired up a test rig on a bench PSU - the pump's power cycles on and off constantly, you can hear the rotor jolt for a moment then stop.  I tried several methods to siphon water through the pump with no action at all.  It was fully cleaned with a small amount of black particle around the rotor, I'll tear it down some more but think it's toast.

 

I'd love to know if the G5 Quad won't ramp up fans to crazy speeds without PWM monitoring on the pump - if so, I could source another small 12V pump and be done with it.  I've a Laing D5 on hand, but that won't fit.

 

Thanks

 

JB

[siliconinsider]

I have saved a quad, it was quite a bit of work and a learning experience, but I ended up rebuilding the entire pipework of the LCS.

I have removed all of the ancient pipework either metal or rubber. The smaller rubber pipes that go to the copper cooling blocs were also problematic and the copper blocks were clogged.

- Ran warm water and vinegar through the heatsink, at high pressure at the end to make sure that no solid residue was left

- Immersed the copper block in a beaker of concentrated sulfuric acid as they were almost entirely clogged.

- Rebuilt the cooling circuit with transparent garden hoses, two diameters were needed, and sealed all the joints with epoxy glue. I have also added a T connection for filling.

- Purging air bubble is the challenging part, consisted of shacking the unit and moving it back and forth to force the bubbles up the T connector and filling coolant with a syringe. I was also able to run the pump with a power supply.

Now it runs super quiet and COY are at a very reasonable temperature at full load.

 

The 970MP model bring power to the CPU card with two bolts that have a serrated washer. This can damage the PCB if forced too much., It could be possible to change the washer, or use a flat one, as long as the contact is good enough to carry the many amps needed to power the CPU card.

 

Photos below are of my first attempt, the second attempt I replaced the metal pipes and the small plastic pipes that connect to the copper blocks with nothing but transparent tubing.

 

 

[mePy2]

Cool!

 

I would have loved to see the new one... my radiator is coggled too. I am undecided whether to remove the bottom hoses too or not.