Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

Tazintosh

macrumors newbie
Original poster
Apr 26, 2024
13
1
Hi folks,

I recently got a Power Mac G5 2.5 GHz already max out to 16 GB of ram and with a FX 4500 for… 30€

It took a loooo…oong time to clean it all but I finally own a shiny almost brand new PM G5. I've done:

  • Full disassembly and OceanGate Titan Deep Clean Level mainly with isopropyl alcohol and air compressor.
  • Changed all thermal paste and pads for: logicboard backside, video card, CPUs (thin layer of MX-6).
  • Full clean and refill of the Liquid Cooling System (version 1): multiple pass of 80°C pure vinegar into the radiator, rinse, refill with auto green coolant, removing as much bubbles as possible (nightmare). Pump is perfectly working and finger test while using a heat gun is astonishing.
All is fully working and running 10.5.8. The computer is stable under load, running 4 instances of yes > /dev/null & for more 1 hour already (at the time of this writing).

Running ASD 2.6.3, all tests have successfully passed (takes more than 5h for 16 GB of ram), but Thermal Calibration is for now always failing with error for both CPUs

ERROR CPU0 - Processor Max. temperature exceeded
Reseat processor and retest unit. Replace processor if it continues to fail.
ERROR CPU1 - Processor Max. temperature exceeded
Reseat processor and retest unit. Replace processor if it continues to fail.
You'll find attached two captures of the screen with no load and under heavy load (> 1h).

I feel my temperatures are fine:

  • At rest:
    • CUPA ≈ 35° C
    • CPUB ≈ 45° C
  • Under >1h heavy load:
    • CUPA ≈ 55° C
    • CPUB ≈ 75° C
Two questions only:

  1. Why there is always a temperature difference between A and B? (I've seen this online many times). In my case: ≈ 10°C between A & B at rest and ≈ 20°C under heavy load.
  2. Why the thermal calibration is failing for "Processor Max. temperature exceeded" considering the CPUs under load seems absolutely stables? (All tries have been run with —all— covers)
Thanks a lot in advance for your inputs.

rich text editor image
rich text editor image
rich text editor image
rich text editor image
 
1. Why there is always a temperature difference between A and B? (I've seen this online many times). In my case: ≈ 10°C between A & B at rest and ≈ 20°C under heavy load.
Could be any number of factors, but the most common are either "user error" or "that's just how it is". CPUs are never created truly equal. While both could perform identically, one of the CPUs may generate more heat than the other and that's just how it is.

I don't have a Quad (wish I could find one without having to defer my rent and not buying food), but I do have a dual-proc PC and one of the otherwise identical CPUs runs noticeably hotter than the other.

I don't know about thermal calibration. I would just omit doing it but I'm not sure if something bad happens if you don't.
 
Thanks for the answer @Doq.
Following your thought, switching CPU should then switch temp difference between A & B placement.
But what surprised me initially was precisely because over my web searches, I was actually always finding people with A cooler than B.
 
Something new and weird happened today:
After two weeks at least not using it (basically since my above photos), I've turned on the G5 for a curious and interested friend.
Initial temperatures for CPUs were good, around 35°C has screened on my initial post.
Then, the CPUB temp started to slowly rise up (without any noticeable load).
It ended up to 98° and the G5 put itself in sleep mode but also hang.
Both led "Overtemp" and "Checkstop" were lit.

I let it rest many hours and did the test multiple times, but the behaviour remains the same --> Both led "Overtemp" and "Checkstop" ends up being lit.

I would be extremely surprised if something were wrong with my thermal compound application, specially since I stress tested initially the G5 during 1h 20 min. But before unmounting all, what's your opinion on this guys ?
 
Quick Update.
I might have found a culprit. At least I think I've to deeply investigate on this direction:
I noticed that since yesterday tests, the G5 was going "Overtemp" quicker and quicker.
Even this morning, while the machine was cold, it took less than a minute to trigger the both led "Overtemp" and "Checkstop" leds.
I've unmounted the LCS with processors, looked closely to the tubes while shaking and found many particules floating around.
My guess is that the small inlet filters are clogged. The new liquid might have helped detaching some deposit and while the system was perfectly clean hours after refilling and testing, those passing weeks may have been sufficient to built enough amount to clog the filters. And the more I was trying yesterday, the more the pump was pushing debris to the filters.
I'll probably buy an ultrasonic cleaner and make a new cleaning. Looks like mine was not good enough.
 
  • Like
Reactions: ToniCH
I've unmounted the LCS with processors, looked closely to the tubes while shaking and found many particules floating around.
My guess is that the small inlet filters are clogged. The new liquid might have helped detaching some deposit and while the system was perfectly clean hours after refilling and testing, those passing weeks may have been sufficient to built enough amount to clog the filters. And the more I was trying yesterday, the more the pump was pushing debris to the filters.
I'll probably buy an ultrasonic cleaner and make a new cleaning. Looks like mine was not good enough.
Yes, dirt coming unlodged inside the radiator could very well cause all kinds of trouble with coolant flow. After reading many of these LCS threads on various forums one common thing is that the radiator is super difficult to get properly cleaned. People needed to do it many many times before they can be sure all the crud is really out.

And then come the air bleeding troubles. I've seen people use all kinds of contraptions to help bleed air out, even vacuum pumps.

I have a 6ltr ultrasonic, I hope the thing fits into it when I start to rebuild mine. Can you please roughly measure the outer dimensions on your radiator?
 
I have a 6ltr ultrasonic, I hope the thing fits into it when I start to rebuild mine. Can you please roughly measure the outer dimensions on your radiator?

The "Bouding Box" of the piece (radiator + CPUs boards support) would be ≈ 20.5 x 18 x 14 cm
20.5 being the CPUs boards support, which belongs to the assembly.
The radiator itself is ≈ 18 x 14 x 5 cm
 
The "Bouding Box" of the piece (radiator + CPUs boards support) would be ≈ 20.5 x 18 x 14 cm
20.5 being the CPUs boards support, which belongs to the assembly.
The radiator itself is ≈ 18 x 14 x 5 cm
Thanks! If one can separate the radiator from the supports then it would fit completely submerged as the tank is 30 x 15 x 15 cm. How is the thing put together? From pics it looks like there are screws?

But, I would say this size for an ultrasonic is pretty much smallest I can recommend for general use. When I bought it few years ago I thought 6ltrs is plenty but In retrospect I would probably buy a 10ltr or even little bigger if I was looking for one now.
 
I'll go for a 15L personally (not for this, but for other projects too).
I'm not sure you care about cleaning the support plate. If the radiator itself is fully merged, you're good no?
 
Sure, 15L would probably be good if you do some car/bike/small engine etc. stuff too.

Yeah, no need to get the sheet metal ultra clean! ;)
 
Update since a month:

Process have been long (and not done yet), for a good reason (besides waiting for parts to be delivered and finding time).
I've received my ultrasonic cleaner and did a long clean of the radiator with 80°C distilled water and vinegar.

WARNING, please note that some screws coating went off during the cleaning and before I noticed, they started rusting. The super thin hexagonal metal sheet that acts like a suspension and surrounds the CPU/Cooling interface on the one pump design will also rust like super fast. Finally, something else on this same interface seems not to like much the ultrasonic waves as some kind of white "smoke" was degasing from it.

I've meticulously cleaned the filters (clogged) too, rinsed a lot the radiator and started to fill it back.
But my gust was screaming me not to mount back the assembly on the computer yet.
And the truth it I was right: I ran the system multiple hours to see how it goes. All was fine. Then I decided to let it rest multiple days, because my feeling was that the coolant could be the reason of the clog.
During all this month too, I read that the copper (CPU/Cooling interface) was reacting heavily with aluminium causing the later to corrode. —Even— when peaces don't touch physically. The simple presence of liquid will allow ions to navigate and do their job.
And after a few days at rest, I started to see some whitish depot developing around the radiator tubes connections, right on the edges where my flexible transparent tubes are connected.
I didn't wait much more and emptied the system. I'm 90% convinced my coolant (Motorkit MOT3543 green 10%) either doesn't contains any anti galvanic / corrosive stuffs, either not enough, but I couldn't find any information about the composition.
I've ordered a bottle of Corsair Corsair Hydro X Series XL5, which is said to contains those.
Now I've to restart all over again. I'll keep you posted.
 
What I did to test the functionality of my pumps and "just because" was to turn the computer on with my version 2 cooling unit outside of the computer with none of the CPUs attached. It also helped work the bubbles out of the radiator since I could move/rotate/jiggle the unit around to work the air into the clear pump tops I got. Just be careful NOT to touch the radiator fins... I spent many hours fixing mine.

However, the upper loop tends to get air bubbles that are easy to work out and top off with coolant. The lower loop seems to be fine. Not sure what's the deal is since I used a leak tester at 8psi for several hours and it didn't leak.

When it comes to coolant I used the EK cryofuel clear premix. Clear so I could see any particles. It also claims low electrical conductivity... whatever that means.

Much of my adventure is documented over at 68kmla.org here starting on page 4.
 
Last edited:
Hi @amishallin
I know your "new blood" post and thanks for it as it gave some interesting directions, though I personally used my own route somehow.

To run the system and do the filling, I wanted to keep everything as far as possible from the G5. So I used a laboratory power supply to run the pump, thus allowing me to also vary its speed.
I've also removed the one way valve from the radiator to replace it with a tube connector (it's a M12 x 1.75 thread if anyone is interested).
With this modification and the positioning as shown on the under photo, the bubbles find their way out almost naturally when running the system (takes time).

The more the bubbles leave the system, the more the level on this tube will rise, it's because of the Bernoulli effect that becomes more and more present. Since the CPU/Coolant interface is super narrow compared to the 10 mm tubing, the coolant speed on the output of the interface is higher than on the input, thus lowering the pressure. Since those tubes (Corsair) are flexible and not as rigid as the black original ones, they will squeeze under the lower pressure and obviously increase the level on my "exit" tube. This is fixed enough when the exit tube is closed, because the pressure will build up more on the system.

IMG_2751.jpeg
 
“Good news everyone!”

I received my Corsair Hydro X Series XL5 bottle but before filling the whole system, I decided to run a test:
  • I placed two brass (similar to copper) caps on a big piece of raw aluminium I had.
  • Around each cap, I dropped a spoon of my Motorkit MOT3543 green 10% for one, and Corsair Hydro X Series XL5 for the other.
  • I scratched the aluminium surface using the caps and let it rest for 5 days.
==> A big amount of depot built up around and under the sample using Motorkit MOT3543. Nothing around the Hydro X sample.

Based on what you'll see on the following photos (caps are 1/4"), imagine the amount of depot that built up into the radiator during 2 weeks !!!

I'll now fill up my LCS and let you know for the calibration test whenever I've time.

On the left: Motorkit MOT3543 green 10%. Depot can be seen all around.
On the right: Corsair Hydro X Series XL5. All clean.

IMG_2784.jpeg
IMG_2785.jpeg
IMG_2786.jpeg
IMG_2787.jpeg
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.