Re-using/ Re-purposing the G5 Water cooling system

[rehpotsirhc]

From what I've read, it typically doesn't matter due to how quickly the water is moving through the loop. It still is not the best way but when you're water cooling a 2P system, you aren't going to slap a radiator in between the processors because it doesn't look good and it doesn't really provide much benefit.

Jeffinslaw

Yeah I think you'll see a 1 or 2° C drop in temps after the radiator only vs the hottest point in a water cooling system once the system has been running for a while and temps have stabilized.

This is a pretty cool guide to water cooling where it gets explained (in one of the 3 videos, I'm not re-watching all three)
http://www.youtube.com/watch?v=aHNFhTAvvBA

 

[Einsteiniac]

Hi,

This is going to be a slow burning thread, so patience required (!), but I thought it might be of interest and worth a post or two.

Recently I got hold of two defective G5 2.7GHZ dual machines. From these two I actually got one working really well and have overhauled for sale. The other though is dreadful.....a nice case (well, that's what I bought them for really) but a cooling system malfunction took out the power supply big style.

The thing is though having only seen air cooled Mac.s I was really interested by the lovely looking water cooler in both machines. The working one I re-furbished, but on the spares machine I decided that the water cooler is too good to throw away. It has a brilliant little 140 mm aluminium radiator that looks bullet proof (pic attached). It also has an interesting pump. The pump is labelled Delphi, but in reality it is a Laing pump which is the standard for most proper water cooling projects and is fairly expensive ( the retail version is about £60). I am replacing the seals on mine and going to see if it does the business.

From testing, the usual black Apple wiring to the pump plug is a six pin connector with 2 blanks. When looking at the connector to the pump end on, the pins are like this:

1, blank, 2, blank, 3 , 4.

The connections are as follows: 1 = ground, 2 = V mot (variable), 3 = Tach output, 4 = 12v motor.

To test, connect 1 to ground and 2 and 4 to 12 v. If testing dry then only run for a second as the pump uses the coolant to lubricate itself.

The failure mode of these pumps is the O-ring seal which costs £1.20 to replace. Failure can also be due to burn out of the PCB (fortunately this is not the case for mine), but there is a guy on eBAy who occasionally sells new PCBs that he makes from new surface mount components.

I'll let you know how I get on.

Oh yes and one other thing I have NEVER seen before. This G5 still had the little grey rivet in place in the processor cover. I managed to extract it without breaking it and have saved it. It's a silly little thing but finding one that has not been broken at some point is about as rare as finding a hen with teeth.

I can say that I have seen WAY too many of those stupid gray plugs then I ever care to admit in my lifetime. They were so freaking annoying and only got worse as the G5 design was revised. The tolerances just got tighter and it became more and more difficult to remove the plug without completely damaging it. Hence the reason you almost never see them. If an Apple technician ever had to remove the CPU cover chances are the plug was completely damaged in the process. Even better was Apple didn't provide replacement plugs so the machines went without and developed a slight rattle.

 

[powerpcg5]

Yes, I think all it needs is a careful rebuild hopefully and then keep an eye on it. I must say I don't care much for their CPU block design. Which O-rings on your original G5 system failed, was it the pump O-ring or the processor rings.?

The O-rings in the cpu block. The original O-rings where white/grayish. I replaced them with a different brand (black and a bit thicker).

My PowerMac has been running for almost a month now without problems, but yeah I'm definitely keeping an eye on it.

 

[minihack]

Just a little P.S. to this thread.

The cooling system is now running my Zotac Z77 and i5 2500k without issues. It is really quiet and I am glad I opted to have a reservoir on the top of the pump as without it (and being able to watch the swirling coolant) I wouldn't know the pump was working.

Stress testing and O/C'd to 4.2 GHz on "turbo" takes the system up to mid to high 40s (centigrade), whereas idle is mid to high 20s (dependent on how warm the room is of course).

I had to design a special case enclosure for the system as naturally it won't fit in my Cube build. So I am running a 7 series ITX board, a discrete graphics card, an OC'd 6-series processor, using one of those brilliant Silverstone 100% modular miniature ATX 450W supplies and a water cooling system from the G5 with a Zalman CPU water block. The system cracks me up just looking at it as it is perhaps not the natural choice of components for an ITX build. The only fan in the system is a Noctua 140mm on the front of the radiator that blows through the whole case.

Edit:
Later the same day.......well. I took the Noctua fan out to make a mount for it and forgot to put it back in. I didn't realise this until an hour after turning the computer on. When I did realise, I hurried to hardware monitor to check what the damage might be. Surprised to see that the temp.s were still idling in the high 20s! So, I thought let's stress test it (briefly) just to see what happens without a fan. The result - to be honest, nothing spectacular, it rose to around 50c in the five minutes I left it and then when I stopped the stress test it dropped pretty quick to low 30s. I guess this radiator and loop does not need that much air cooling.

 

[powerpcg5]

I guess this radiator and loop does not need that much air cooling.

Yup. If you look at a PowerMac G5, the dual front fans in front of this radiator barely spins at all... it's spinning very very slowly (you can see the blades as it spins, it's that slow) at normal operation and just a slight increase in speed during load.

It's made/designed by Delphi... the same Delphi that makes radiators and other car parts for GM/Chevy. It's really overkill.

 

[minihack]

Excellent. My corsair would not keep the loop cool for long without the fans spinning, and with the G5 Laing pump being really quiet this little (re-cycled) loop works really well for a near silent installation. Come to think of it, the aluminium radiator does look a lot like the oil cooler I used to have on one of my souped up MG Midgets......

 

[Jeffinslaw]

Excellent. My corsair would not keep the loop cool for long without the fans spinning, and with the G5 Laing pump being really quiet this little (re-cycled) loop works really well for a near silent installation. Come to think of it, the aluminium radiator does look a lot like the oil cooler I used to have on one of my souped up MG Midgets......

Is the inner chamber of the radiator aluminum or is it just the housing? If the whole thing is aluminum then you need to get that thing out of there ASAP. Mixing different metals like aluminum and copper (which is what most blocks are made of) is a REALLY bad idea.

Jeffinslaw

 

[minihack]

Yes, it is all aluminium.

If there is a problem Apple had it too because the CPU blocks they used are certainly aluminium with copper internal fixings. Having skimmed one down and also taken it apart I know that for sure.

There are articles on galvanic corrosion I have looked at: http://martinsliquidlab.org/2012/01/24/corrosion-explored/ and in my loop there is no direct aluminium copper contact (unlike the G5). I'd agree though that it is probably not a great idea, and probably was not Apples greatest move either. None of my systems is ever likely to be doing enough years without attention for it to be a real concern I think - or at least it is a chance I am willing to risk for a while.

 

[powerpcg5]

Yes... there are also copper fins inside the CPU water block of the Apple system but even after several years of use, when I opened up and checked the copper fins, they look just fine without any sign of corrosion.

I haven't seen photos of Apple LCS looking like these photos. http://martinsliquidlab.org/2012/01/24/corrosion-explored/

Most of the failures seen on the LCS of the Apple is not due to galvanic corrosion, but due to failing rubber O-rings (either in the CPU block, or the water pump).

 

[Solidww]

Glad to read this worked out great. Time to invest in making one of these work.