PowerMac G5 Quad "The New Blood Mod" - A guide to flushing, modifying and refilling the dual-pump cooling system.

[Byrd]

It's pretty impressive! I'd be happy to see how you managed to attach the heatsinks.

Will post up a new thread soon once I'm happy, as the experiment needs more work. I used the existing 4-pronged clamp (that used to hold the waterblock), modified slightly to clamp down the new heatsink securely and equally over the core. I'm unable to calibrate using AHT - the software throws an error about max core temp being reached, even though it's sitting at 30 - 35 deg C idle. Under load, the fans kick up too high while the temps don't go above 60 degrees. Makes me think that perhaps a basic AIO cooler x 2 (now that there is space to mount a block) might be suitable to cope with the heat soakage. But at present it never crashes, no LED errors, and I can actually use the Quad happily and relatively quietly.

 

[igor_av]

Will post up a new thread soon once I'm happy, as the experiment needs more work. I used the existing 4-pronged clamp (that used to hold the waterblock), modified slightly to clamp down the new heatsink securely and equally over the core. I'm unable to calibrate using AHT - the software throws an error about max core temp being reached, even though it's sitting at 30 - 35 deg C idle. Under load, the fans kick up too high while the temps don't go above 60 degrees. Makes me think that perhaps a basic AIO cooler x 2 (now that there is space to mount a block) might be suitable to cope with the heat soakage. But at present it never crashes, no LED errors, and I can actually use the Quad happily and relatively quietly.

Could it be because the LCS is not connected anymore and AHT is expecting it to be there?

 

[Amishallin]

After leaving my G5 quad project sit for a few weeks I finally cracked open the LCU. Got maybe a combined 5oz of fluid out of both loops and hope the rest will come out while flushing with external pump and time in an ultrasonic cleaner. I noticed the "Z" tubes have some kind of rubbery/plastic thing around the end. Do these need to be removed? I'm thinking setting the ultrasonic cleaner to 45-50C at 30min with distilled water. Is this correct or do I need to add some kind of cleaner?

This side seems to be the "cleaner" of the 2 sides.


This side is very corroded. Plan on putting a clamp before and after the ridge on all 4 ports.


This was some wonderful smelling stuff.


Read from a few places that the best way to remove these clamps is to either pry them or cut the crimp off. Pretty sure other people have found this out but if you break off this small tab the clamp comes off easily.


Are these 4 hoses ok to leave? They look smaller than the hoses on top.


Also have to spend time fixing the minor damage I did to the fins on the radiator. Other than this the rest of the machine looks brand new... aside from the dried up thermal paste on all 4 chips that need it.

 

[Amishallin]

Deciding to go on I removed the pumps to begin flushing the radiator. The first thing I noticed is what appears to be corrosion from some leakage on 1 of the blocks (both CPUs came off with no signs of corrosion). When everything is done I'll have to run fluid through that side for a few days and see if it leaks. Also, there is a fingerprint "burned" into the bottom of the block. It was there when I removed the CPU and cleaned off the crusty thermal paste. Guess whoever assembled it wanted to leave their mark. The second image is the flow of 50/50 water/vinegar through the rad. Don't think buying the cheapest pump on Amazon claiming to have a flow of 90 gallons per hour was a good idea. It was only $6 and thought I could get away with it. However... it is able to push out some small pinkish crystals.

The rest of the images are of 1 of the pumps. They both are wrapped in duck tape and seem to be rather clean on the inside. Do these disassemble further? There is also more of that glue around the bottom on the outside. I'm guessing this is the glue mentioned in the 1st post and is ok to remove in a controlled fashion? Does it have to be re-glued or do the 4 screws hold it all together just fine?

 

[Amishallin]

A little update. Got a real pump for fountains at Menards. It was strong enough to cause the radiator to vomit out crystals and other debris after another ultrasonic session. All parts flow freely and clean now.

The tubing on the bottom for the copper blocks has been removed revealing some corrosion inside. After an ultrasonic treatment (with the rad) they are shinny inside. FYI i'm using 1/4" inner diameter 3/8" outer diameter silicone tubing and the same 8.3mm-10mm clamps (new).

There is 1 thing I need to know at this stage... the size of the o-rings that hold the copper block in position. 360alaska's guide states these are size 27 orings. Looking at some sizing charts size 27 seems to vary around 1.3 inches. After looking at orings at Menards and other hardware stores it seems the inner diameter is 1-7/8 (or 1.88). Are these rings actually around 1.3 inches but stretched?

The attached images are a partial reassemble of the LCU.

 

[CircuitBored]

A little update. Got a real pump for fountains at Menards. It was strong enough to cause the radiator to vomit out crystals and other debris after another ultrasonic session. All parts flow freely and clean now.

The tubing on the bottom for the copper blocks has been removed revealing some corrosion inside. After an ultrasonic treatment (with the rad) they are shinny inside. FYI i'm using 1/4" inner diameter 3/8" outer diameter silicone tubing and the same 8.3mm-10mm clamps (new).

There is 1 thing I need to know at this stage... the size of the o-rings that hold the copper block in position. 360alaska's guide states these are size 27 orings. Looking at some sizing charts size 27 seems to vary around 1.3 inches. After looking at orings at Menards and other hardware stores it seems the inner diameter is 1-7/8 (or 1.88). Are these rings actually around 1.3 inches but stretched?

The attached images are a partial reassemble of the LCU.

You can install the heat blocks without the o-rings. They don't really do very much - I think they're just a last-ditch effort to try and block a hypothetical leak in any way possible. My Quad is running fine without them.

 

[Amishallin]

Did a test fit of the clamps I got (same style that were on there) and it looks like they could tear the softer silicone tubing in the future. After bending back a couple nails and slicing my finger open I managed to get short strips of the 3/8 tubing around all 8 connections. These fit super tight and may not even need clamps. However, I still plan on applying either silicone gel or JBweld WaterWeld around the ends of the hoses.

Attached are some images of just the tubing. What do you guys think?

 

[Missedisland]

Currently working my way through this guide as well for my Quad G5. Thanks so much for writing this @CircuitBored . I'm on exactly the same steps as you are @Amishallin. However, when I started to fill up the LCS system with fresh coolant, both pump tops started to leak back here (see photo). The top doesn't fully seem to fit on the pump there. Am I doing something wrong or do I need to make a more tighter seal somehow?

 

[Amishallin]

Hmm yes been wondering about that... mine looks the same way. Was thinking about filling the void with silicone.

 

[bkmoore773]

The O-ring is what makes the seal between the pump and the heads. If it's leaking, then something's not right with the o-rings. Mine has the gaps, but has never leaked.

 

[Missedisland]

Thanks got it! Somehow both o-rings were still stuck in the old Delphi tops and I didn’t notice.. I even almost threw those in the trash already. Seems to be sealed now.

 

[Amishallin]

Not sure if this is normal or not but when the components were separate (block, radiator, z tube) the flow of fluid was quite strong... almost too strong. However, when everything it put together there is "ok" flow and slowly goes down to a constant trickle (both directions). I know for a fact there are no blockages since each part was tested individually both directions.

The pump i'm using to leak test is an 800 gal/minute pump that will eventually replace a failing pond pump. I do have a 1/4hp pump but think going from a garden hose to 3/8" will literally blow things apart... causing more harm than good. The current pump has an adapter for 1/2" and was able to fit my 3/8 tubing on it.

 

[bkmoore773]

I'm thinking you probably have an air bubble in the radiator or somewhere that's blocking the flow of water.

 

[Amishallin]

Not really sure. Tried rotating, shaking, and tapping it on the table and nothing. It's so weird since all the individual parts flow freely. The last time I flushed the rad was when each side ran clear for a day.

How much flow are the 2 pumps supposed to have? I think the DDC pumps are more suited for this than the type i'm using.

Think i'm gonna do an experiment this weekend. Completely reassemble the loops and use a flow indicator on the in/out hoses. The G5 should be fine powered on just to supply power to the pumps right?

 

[CircuitBored]

You can install the heat blocks without the o-rings. They don't really do very much - I think they're just a last-ditch effort to try and block a hypothetical leak in any way possible. My Quad is running fine without them.

I hope this comment hasn't caused confusion. I'd like to clarify: this is in reference to the two O-rings at the bottom of the LCS assembly, between the CPU card and the water cooling assembly. If you are disassembling a unit that has never been serviced before then you will likely find the disintegrated remains of the originals.

The O-rings at the top of the LCS assembly, between the pump and pump top, are absolutely essential and have to either be transferred over or replaced in the new assembly.

 

[Missedisland]

My rebuild LCS comes along nicely. However, still getting tiny leaks around the hose barb fittings. Anyone else experiencing this? I used Alphacool ones but the o-ring doesn’t seem to fully screw into the top for a snug fit.

 

[Missedisland]

Ok, I redid the tubing with new (tighter) clamps and fittings. No more leaks. Think I also got all the air out of the system. I’ve reassembled the G5 however it fails with a overheating error (red light next to the ram), all fans spin up, and ASD says “failed to communicate with CPU0 and CPU1”. Thing is the liquid cooling seems actually pretty solid now. Could there be another heat issue? I’ve repasted the CPUs and also the Northbridge on the back in process.

 

[Quadraman]

Does anyone know if there is a non-conductive, non-corrosive, fluid, with the same viscosity as water, that can be used in place of water in these?

 

[Amishallin]

Don't know about non-conductive but i'm using clear EK CryoFuel. It's non-corrosive, anti-microbial and is great for thermally demanding PCs. Why clear? Wanted to see how much sediment was left in the loops... and it was on sale.

Image on the far left: each side passed a 45min @ 8psi leak test.

Image on the far right: the coolant was a little cloudy/foamy at first but as it sat the bubbles condensed and were worked out (clear now). Relieved the pumps still work and are running strong.

 

[Amishallin]

Thermal calibration passed today after 15 min. When running the other tests all 16gb RAM failed at the pattern test. Take it that means it's all bad? After several attempts it was able to boot to Tiger desktop once where I connected to my wifi and set the date/time before it froze (no checkstop or overtemp or fan spin up). While it was fully booted I removed the air deflector and lights 3, 4 and 5 turned on (which is normal). The other times the little twirly thing goes indefinitely with no checkstop or overtemp lights. The light above the power button stays solid and doesn't blink 1-5 times.

Pretty sure it's bad RAM and OWC still sells it. Been a customer since 1999.