Working on the G5 quad liquid cooling system

Heikki Lindholm wrote:

As far as I can tell, and according to the service manual, there are two versions of the quad G5 LCS: version 1, which is made by Delphi and has one pump, and version 2, which is made by Cooligy and has two pumps. I've never seen a Panasonic cooling system mentioned in relation to the quad G5, but I've no inside track on these things…

I have no inside track either and had also not read any mention of Panasonic in this regard. It was forum user romko23 who brought up the Panasonic name first, and the ethically challenged tech I mentioned seemed to run with it quite naturally. He seemed very knowledgeable and friendly, other than his penchant for making fraudulent charges to a customer's credit card within a few minutes of having gotten a hold of the information. 😠

Heikki Lindholm wrote:

…Maybe he was talking about the dual G5 not the quad. The late 2004 service manual does mention a Panasonic cooling system.

We were definitely discussing the Quad, specifically mine which was built in late June of 2006, but he may have been mistaken about the Panasonic in the Quad.

Thanks for that very helpful clarification.

if I switch the CPUs and don't do a thermal calibration, the system will work, but with fans full blast

That is the case.

I'd just like to switch them for the sake of testing whether the "hot cpu" is more caused by the cpu than the cooling system or not.

Switching CPU positions is a very good troubleshooting technique.

if I later change the CPUs back to how they originally were, and never do a thermal calibration, does the system still think it needs a calibration and sets fans full blast?

The system should "think" that all is normal if all is as originally installed.

The majority of cases don't need calibration, though many do it to be thorough.

Also, the use of the ASD disc sometimes "cures" cases of "gremlins in the system"; strange oddities which are hard to pin point yet occur.....

Should you desire a copy of 2.6.3, send me an email via my website listed in my profile.

I can help get you a good copy.

Hi,

Ok, so the single pumps that you and I have are by Delphi. Great! But unless someone can get a real close up of the version 1 LCS and identify it by name of Panasonic or even Delphi, I will put it as a question mark. I have taken my G5 off line due to some messages I read regarding about a DigiNotar ROOT CA certificate and the fact that we are no longer getting software security updates anymore, which put us PowerPC users at high risk for our information to be stolen. Evidentally, Apple does not want us using the PPC macs.. Unless someone can really REALLY translate what all this means from www.lowendmac.com, I have decided to keep my G5 off line until further notice as I don't want someone to steal my information and use it against me.

I am clueless when it comes to the single pump though, as it seems like it was only added in the mid to end cycle of the G5 Quad.. figure around April to August of 2006.

I agree.. I am just glad I have a single pump and not the dual pump. Strangely enough though, 2.5 years ago when I had my 1st G5 Quad it had a dual pump and it never ever gave me any problems at all.. though I used this G5 to sell and get the 2006 Mac Pro.. and the rest was history then as the selling of my first G5 paved the way for me to complete my transition into the Intel Mac Pro. And today I have my trusty G5 Quad with the revision 1 or more reliable LCS and the Mac Pro 2010 model. Its not really a 2010 model, I bought the boards and changed it from a 2009 to a 2010 since all the 2010 is, is a firmware update to allow Westmere processors to run.

Well, it wasn't I who was claiming anything having to do with Panasonic; romko23 brought it up initially.

I also have no reason to believe that Panasonic units were used in the Quads, though a Wikipedia article mentions them in connection with the G5s in general without citing a source.

Also, nowhere do I find confirmation that a single pump is better (or worse) than a dual system.

Heikki Lindholm wrote:

Ramón G Castañeda wrote:

Also, nowhere do I find confirmation that a single pump is better (or worse) than a dual system.

Such confirmation would be hard to get from anywhere…

Those are my sentiments exactly. That's why I was wondering why Romko had been and remains so insitent that single pump is better. Finally, my above comment reflected that.

The sum all the anecdotal evidence I've come across seems to point to the LCS failing often in the G5 Duals when they were first released, not the LCS in the Quads, regardless of what they are.

It's all a moot point for everyone involved in this discussion at this point, and I think it would be of interest only to someone contemplating the purchase of a G5 Quad today—and good luck finding a seller who can reliably tell you what exact LCS the machine he's offering for sale has. You practically have to take the whole darn machine apart in order to tell.

Following this thread, I have endeavored on rebuilding my LCS dual Delphi unit. Flushed out with compressed air, removed all hoses and clamps, and epoxy on the ends of everything Apple apparently was scared sh!tless of leaking. I filled the little radiator, pipes and the cpu heat transfer thing with Muriatic Acid, let it do its thing (wear protective eyewear!!). Then let it sit, flushed everything with Baking Soda dissolved with distilled water. Installed new clear hoses, new clamps (available at Koolidge). Flushed with distilled water. Removed the bleed/vacuum caps on little pipes and on radiator (4 all together, 2 each side). Had to Dremel the ends off as the factory squeezed those shut and then epoxied the whatnot out of it and put the caps on. Tapped the openings for a #8 machine screw about 3/8"length, made sure I blew out any crud from tapping the threads. Then made sure the v-cap (plain vanilla) machine screw would sit flat on the inlets. Tightened them somewhat and they held vacuum. Then got a bucket with 50% car antifreeze (the good orange stuff: Dexcool or something). Then while submerged in the bucket, used an automotive brake bleeder to suck coolant into the system, while turning it upside down, making sure it remained submerged, or the open pipes at least. Made sure pumps were filled with water also. Then hooked up all the hoses. Then opened the screws I installed on the bleeder/fill ports, kept everything submerged and sucked about 1 liter through the system each side, looking for bubbles. put RTV sealant on the #8 screws' threads and let that cure for 10 mins, then installed the screws, keeping everything submerged. Same process for each processor side. Quite a chore. Then let it dry a bit by blowing compressed air allover everything. Made sure there were no bubbles as I turned it up and down and sideways. And dddough there were still little tiny bubbles but I figured it must be almost impossible to evacuate all and any air out of this contraption. Then put under-water two component epoxy (amazingly enough available at Osh) on all the ends of pipes, hoses, and filled vacuum plugs (from Autoparts store) with same and put those over the screws I installed on the bleed/fill ports (4 all together). While trying not to get the stuff everywhere, I made sure it was where it counts.

Let that dry overnight. And installed it today.

My cpu b used to run almost 2x as hot as cpu a before. So now the difference is only 10F between cpus, and it's cpu a that is hotter.

Basically I could tell that the cooling system for cpu b was missing coolant since it made a chugging noise when turned around and upside down.

Now this seems to have cured the cooling part, but since both cpus are a lot cooler now around 111f and120F . BUUUT my fans still start running high at 3k (intake and exhaust, drive bay and back side are low and quiet) and the pumps running at 3.6k. Gradually increasing from much lower to that level in about 1 min from having booted. This is oSx Tiger 10.4.11 btw, my fave by far for ppc.

So now I try to run 2.6.3 Apple Service Diagnostic, and it hangs on the test for cpu A pump. I get overtemp led and checkstop with frozen computer. If I uncheck that test everything passes. So then I figure I try the thermal calibration and that hangs after about 3secs, same leds on as above and computer frozen.

Also tried ASD 2.5.8 and that won't load the firmware interface whatnot.

So basically I think this is just a bad design all together, just looking at it seeing all that epoxy from the factory all over the place. And then on top I do not think ASD 2.6.3 works for the 2 Delphi pump setup, my hunch.

The rebuilt LCS seems tight and I may just use it this way, seeing there is no way to thermal calibrate this thing. B@st@rd boots very fast with this ssd I got also. It seems stable so far with me tinkering around in photoshop and lightroom, which is all I really do. Just the noise that's annoying.

This is a fine example on how not to design a pc, reminiscent of Italian performance cars design philosophy.

I just posted this so no one else wastes their time. Not worth it. Duuuh. New hardware coming my way.

Ok. So it basically would run with fans blowing high, could not calibrate cpus (ASD 2.6.3 would hang 5 sec into calibration, and diagnostic would hang when testing cpu A pump (fan). Then I noticed that taking the clear big plastic air circulation side cover off would actually lower the fans, especially the intake ones, yet it would still show 3k rpm in iStat Pro, though they were blowing very slowly. So I left the plastic off for several hours, did some processor intensive stuff and no difference in fan speed, iStat showing still 3k, but it was nowhere near that from looking at the intake fans. CPU temps would increase, and decrease according to load. After about 2 hours, about half of it idle, it was frozen again and fans blowing full speed (not like when I first took covers off) when I got back to it (left unsupervised for 1 hour or so). Would not start back up again at all, overheat led and checkstop both would come on 10 sec from pushing power button. Tried resetting defaults, nvram, reset-all commands and still no go. Got my hands on a known good logicboard of the exact same type and same symptoms. Removed cpu b and same, swapped B to A and same problem. Yet with this logicboard I would only get overtemp led, again about 10 sec from pushing power button. Would not boot (folder with ?) and then freeze as led for overtemp came on. Pumps seem to be running judging from touching them and vibration. Since my tubes are clear I imagined I would see coolant flowing but it is undiscernible, even the small bubbles I still have in the system would not seem to move or wobble. Never did a flow test on the pumps checking volume, just made sure they were pumping when I rebuilt the whole enchilada. It's basically a doorstop now. Probably fried the cpus by letting the big plastic vent cover off and running it for 2 hours.

Btw one can't really peer through the cpu cooling blocks, there are small passages on each side surrounding it I would assume but definitely nothing to see there. There was no evidence of corrosion anywhere: not in the lines, not in the radiator, not in the cooling blocks or surrounding. The muriatic acid mostly did its trick in the radiator core I think, as it did improve flow for the lower (cpu B) cooling set up. My fans were ramping up within 1 minute from booting to 3k, cpu temps would be around 160 F when I was stressig cpus, otherwise at 120F-130F, one about 10F hotter than the other. It did that before rebuilding the cooling system when cpu B was up to 50F hotter than A, and it did that after with only 10F difference between the two and A running hotter now. I am going to drop an air cooled 2.3 dual core cpu from another g5 that bit the dust and see how that goes over. Cheers.