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

flyproductions

macrumors 65816
Original poster
Jan 17, 2014
1,104
468
So, as even with the slightest load getting very loud, i found it time to service my Quad's cooling again. Last time i did this was exactly ten years ago with six of them it even beeing my daily driver and productive machine for graphic design work. So it was about time.

What i did was draining and refilling the coolant, opening the CPU-blocks to check for clutter and of course renewing the thermal grease. All looked nice so far. No leaking whatsoever. In particular i was positively surprised to find close to zero residue inside the CPU-blocks...after such a long time.

So temps and fanspeeds are ok again.

Idle not very low but fair enough.

Idle.jpg

At least all of the big fans and the pump reach their lowest speeds.

Rendering on all cores in Cinebench ok too.

load.jpg

Temps a bit more even across the cores not hitting 70ºC with fans around 2.000, so still 1.600 below their max.

So in all: Success!

...with one exception: Straight from boot up the Main Logic Board Backside fan starts to crawl up in speed from it's low of 1.100 rpm to over 2.800 without any particulary high temperatures in this area and does not drop back below 2.600 even without any workload. In effect the whole thing is annoyingly loud again. From this single fan. And without any reason.

Did anybody experience this too or knows what can be done? Can dust behind the board be a reason or is recallibration needed?

Thanks in advance for any suggestion.
 
Last edited:
So, as even with the slightest load getting very loud, i found it time to service my Quad's cooling again. Last time i did this was exactly ten years ago with six of them it even beeing my daily driver and productive machine for graphic design work. So it was about time.

What i did was draining and refilling the coolant, opening the CPU-blocks to check for clutter and of course renewing the thermal grease. All looked nice so far. No leaking whatsoever. In particular i was positively surprised to find close to zero residue inside the CPU-blocks...after such a long time.

So temps and fanspeeds are ok again.

Idle not very low but fair enough.

View attachment 2104175
At least all of the big fans and the pump reach their lowest speeds.

Rendering on all cores in Cinebench ok too.

View attachment 2104176
Temps a bit more even across the cores not hitting 70ºC with fans around 2.000, so more than 1.000 below their max.

So in all: Success!

...with one exception: Straight from boot up the Main Logic Board Backside fan starts to crawl up in speed from it's low of 1.100 rpm to over 2.600 without even some high temperatures in this area or any workload. In effect the whole thing is annoyingli loud again from this single fan without any reason.

Did anybody experiance this too or knows what can be done? Can dust behind the board be a reason or is recallibration needed?

Thanks in advance for any suggestion.
There's a thermal calibration that needs to be done after working on the G5 CPUs. You didn't mention that you did this, but I'm not assuming anything. Did you?
 
There's a thermal calibration that needs to be done after working on the G5 CPUs. You didn't mention that you did this, but I'm not assuming anything. Did you?
No, i didn't. But i also didn't the last time, i did this service. To my knowledge this only needs to be done in case one or both of the CPUs are replaced, but not if they are just unmounted and remounted again.
 
No, i didn't. But i also didn't the last time, i did this service. To my knowledge this only needs to be done in case one or both of the CPUs are replaced, but not if they are just unmounted and remounted again.
OK, that's the only question I had. I've never done this myself, I just knew this was something that had been mentioned before so thought I'd ask.
 
So i tried an SMU-reset, as suggested here. Did not help anything!

After boot the speed of Main Logic Board Backside fan is 1.100 rpm. It's like that for two minutes or so, while the temperature reaches 43 - 43,5ºC and stays there. But then this stupid thing starts crawling up. Even with the temperature not rising at all. This continues up to 2.700. Also without any noticeable effect in significantly lowering the temperature. But - of course - producing a solid noise level.

I'm close to just unplugging it! 😡
 
I see that your sensor readings lack one important temperature: memory controller.
Here's iStatMenus example of sensor readings with memory controller:

leo1.png


Please, try to install iStatMenus 3.19 and let's see what is says. Normally it should be in range of 55-65 Celsius.

"Memory controller" is most likely an U4 north bridge, which tends to dissipate lots of head.
It has thermal paste on it (which should be replaced during machine maintenance), and has plastic clips holding it's cooling assembly. Those clips sometimes tend to break on their own with ageing.

So, my bet is that U4 in your quad has no proper cooling, which causes it's temperature to rise, resulting in high backside cooler RPMs.
 
I see that your sensor readings lack one important temperature: memory controller.
Here's iStatMenus example of sensor readings with memory controller:
Please, try to install iStatMenus 3.19 and let's see what is says.
Bresnik get's this too:

Memory Controller.jpg

It gets a bit warmer than yours. But even the 66 it reaches with the shown full load should be perfectly fine for a part like this. And once the load is over, with the mainboard fan ramped up like this, it falls back well below 64 again.

Wouldn't tend to see this as a reason for the behavior. Meanwhile i think about running the calibration, as eyoungren suggested. But i'm a bit afraid about how much it stresses this old machine.

Edit: Sorry, just noticed that the pic was taken after the processor load had allready finished, but fans and pump not yet throtteled down.
 
Last edited:
But even the 66 it reaches with the shown full load should be perfectly fine for a part like this. And once the load is over, with the mainboard fan ramped up like this, it falls back well below 64 again.
Yep, this is fine and definitely not a reason for such an issue.
 
So i finally ran the Thermal Calibration from the Apple Service Diagnostic DVD yesterday. Both of the CPUs passed without any issues. But in the end it didn't change anything.

Basically the temps have been ok before and so are now.

To begin with the better, temperatures under load are very nice! Stabilizing well below 70ºC within a range of < 5ºC across all four cores. Even with multiple consecutive runs of Cinebench. I was not able to get the pump passed 2.700 rpm and all the big fans keep their speeds around 2.000, not exceeding 2.200.

Idle-temps, on the other hand, could be a bit lower. With the lower CPU in the high 30ies with one core hitting 40 from time to time and the upper one in the mid to high 40ies, the Quad's best values after the last cooling-service have been +/- 5ºC lower over all. One reason could be the thermal compound used: Then i used CoolLaboratory's Liquid Metal while now i only had some Arctic Silver Ceramique 2 available which even was very old. So in particular with the coolant circulating somehow slowly with the pump running at it's lowest speed, the metal should be more efficient in dissipating the heat of the CPU-dies. But at least all the fans and the pump stay at or very close to their low limits when idling. The difference between the two CPUs however is ok and was expected. In the one-pump-system the cold coolant coming from the radiator has a much shorter and more direct way to the lower CPU while the upper on is additionally heated from below. These effects decrease as soon as the pump-speed rises the coolant starts to circulate at a higher speed and pressure.

Nice so far! But, what did not get changed by the calibration, is the much to high speed of this stupid mainboard fan, permanently reaching 2.600 to 2.700 rpm (which is just 300 below it's max!) even at idle and with the sensor's temp not exceeding 43ºC. So next, i will try, is blowing some compressed air behind the board to eventually remove some dust-accumulation. Next could be removing the board to have a look behind. But this is exactly what i try to avoid, as it requires a complete disassembling of the machine, while i do not even exactly know, what to look for. 😟
 
Maybe this is of help?
Yes, it was! 😃

In fact, making use of your link finally solved the problem!

There might be things, which simply have to be done...after more than 15 years of use.

quad_mainboard.JPG


So i did what i was hoping to get around and stripped the Quad litterally to the last screw last night. But the reward was satisfying!

I renewed the (surprisingly not even all dryed out) thermal paste on the northbridge. But the much more important thing was cleaning the fan itself, as described in the linked article. This part was just a chunk of dust with the individual blades barely beeing visible. No surprise, this could rotate as fast as it liked without beeing able to move any relevant amounts of air or cool anything.

So, after reassembling, the "turbo boost" of the mainboard fan was just gone! I was not able to make it speed up of it's base speed of 1.100 rpm as hard as i tried. Not even with constant full load over extended periods of time. It maxed in temperature at about 47ºC but didn't trigger the fan which before started to ramp up from anything above 41 and easily reached it's max of 3.000 after minutes of load. The "Main Logic Board Backside" temperature now idles around 43.5 degrees. This is about 1 degree higher than before. But it's 1.100 vs. 2.600 rpm of fanspeed!

So i'm very happy now with a G5 Quad beeing as silent as it gets and appearing a bit like brandnew. But, if anybody else also has a G5 with a roaring mainboard fan, don't hope to get away for cheap! 😝
 
Normally it should be in range of 55-65 Celsius.

"Memory controller" is most likely an U4 north bridge, which tends to dissipate lots of head.
Just noticed, that with my "mainboard service" the temperature of the memory controller decreased for nearly 10º!

New Temps.jpg
 
Last edited:
I wish my Quad was operational...
Yes, i understand! Even now using a 5,1 which is in every way nearly ten times as fast, i still love this monstrosity.

But, if it is the liquid cooling, that fails, it isn't that hard. The hardest part is opening these 8 phillippshead-screws...

quad_lcs.jpg


...for the first time. After all this leaking of the first version of the LCS, used with the Dual 2.5/2.7, Apple used lots of red Loctite to make sure they don't fall off. But, if this is done successfully, the rest isn't that hard. I even reused the o-rings for the second time now.
 
I'm shaking at the very idea to dig in Quad's LCS... :D
I'm shaking at the very idea to dig into Linux... 😂

But it makes me a bit sad to see the most powerful Powermac ever existed to vanish from the planet for just that: a relatively easy maintenance task. People even attempt to switch to air cooling just to avoid that. Meaning swapping some piece of technical art for some less effective profane chunk of metal. No one would ever expect to ride a car for over 15 years without once refilling the coolant. And in fact it is the very same thing.
 
Last edited:
Nice to see new thermal compound working even better after a while of use. So now i get some really nice temperatures for the Quad, shortly after booting right before the liquid starts to warm up a little. 😁

Temps.jpg

But even after hours of idling, CPU A maxes out in the low 40s now with pump and all the fans including the one of the Main Logic Board Backside staying at their lowest rpm. So this should be done for a while and i wouldn't trade it for Air Cooling! 😉
 
  • Love
Reactions: philgxxd
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.