MP 1,1-5,1 2009 Mac Pro - PSU fan loud - bad PSU?

[Tordenver]

My system:

. Mac Pro 2009 5,1 / A1289 / - SMC 1.39f5 (upgraded in 2016 from 4,1)
. Nvidia Titan X GPU (non EFI) - I use ATI Radeon HD 5770 whenever I need to update OS and/or maintenance
. Mac OS High Sierra 10.13.6

This system has been working great for the last 3-years until last week. I updated the Mac OS to the latest Apple Security. After done with installation (remember I have to switch GPU cards) I had some intermittent crashes, where the whole system goes to black without warning. It seemed to be a hardware issue to me. I thought that I may have one of my PCI cards or GPU not well seated, and decided to use that time to cleanup my machine (dusting off, which I try to do on regular basis).

I took out the CPU tray, reseated each RAM memory banks. I also took out my power supply unit, which I have to admit I had never done before. I blew dust out of it and clean up the fan as well.

I put back everything together and cleaned up my PRAM, reset the SMC (unplug the power plug for 30 sec. and wait 5 sec. before restart). The system restarted just fine except now the PSU fan is running very loud! I cannot remember last time I had this issue. As far as I can tell, this is only the PSU fan. What is strange is that iStat is showing the PSU fan running at 0 rpm!!

Now I am trying to figure out what can be the issue. Could it be a bad PSU? because of the noise and the worry I could damage the system even more, I have not run the system long enough / have not stressed it out to see if I get any more intermittent shutdown...

Following are a few screen shots from iStat menus showing the reading from my system. Do you think they look unusual? any ideas what could be the issue with my system? what do you think I should check next?

Thanks

 

[mikas]

Northbridge temp is a little bit upwards i think. Maybe you should check the rivets.
What about system monitor, was there any activity at the moment of the screenshots. You are using 230 watts at the moment, which would be not so much. For me it fluctuates between 150-230W at almost idle, with about 5% load. I think CPU T-diode A seems a bit high too if you don't have any demanding processes running at the moment.

Maybe try another software too, there have been some strange differences between these softwares.

What I would check:
- system monitor info (comparing the temp reading with it).
- another temp, amp, fan revolution software (just to be sure).
- northbridge rivets (they will breake some day, so..).
NB rivets are the hardest one, because you need to remove CPU heatsink to get under the boaard. one of the screws to remove CPU board is under the heatsink. You might be able to check them if they are alright with help of a zoom lens or so, but repairing is another thing.
- then it might be the PSU too as you suspect already (we all know they won't last forever).

 

[amedias]

some thoughts....

1. The PSU fan is not inside the PSU, it's between the PSU and the front optical bays, and it's plugged into the logic board at the front underneath the optical bays so a 'bad' PSU shouldn't have any impact on the fan as it's driven from the LoBo. *If* your PSU temps were high then the SMC would increase the PSU fan RPM to increase cooling, so a bad sensor in the PSU could cause high RPM, but your PSU temps are fine so I don't think you have a PSU issue.


2. The fact that it's reporting 0RPM means that the sense/tacho signal on pin3 of the fan connector is absent. This could be a bad cable, or it couldbe a bad fan. This is why your PSU fan is running fast because the default response from the SMC to a zero/low RPM reading is to increase the voltage to the fan to try and increase the RPM. So because there's no RPM signal it's ramped up the voltage (on pin 4).
Firstly, check the cable at the LoBo and make sure it's seated properly but since you haven't been working in that area it's likely fine. Then check the other end, which ll be harder! It's possible you damaged the other end (at the fan) when you removed the PSU or were cleaning the fan out. I've seen old dry/crappy solder joints broken just form a blast of compressed air before :-(


3. Your NB temps are high, was that at idle or under load? It's not scary high if under load, but still worryingly high. The screenshot suggests it wasn't under load as the CPU temps were fine...If that was at idle then as mikas posted above I'd be checking your NB heatsink and pins *right now* especially since you have >20C different between the NB Diode and heasink, this suggests poor thermal contact between the NB and the heatsink. This could very well be the cause of your shutdowns as if it's runnin 83C at idle then it'll be running much hotter under load and can cause the shutdowns. Those rivets are getting old now and brittle, they can fail on their own without any interference but any knocks and bumps from taking a tray or PCI card in and out could have been the last straw and caused one to pop.

So in summary I think you may have two issues!

 

[Tordenver]

First thank you for looking into my issue!

Northbridge temp is a little bit upwards i think. Maybe you should check the rivets.

Excuse me but I have no idea what the northbridge rivets looks like. Do you have a link for a picture?

What about system monitor, was there any activity at the moment of the screenshots.

No major activity at the time of the screen shot. I use this system for video editing but for the past week prior to the screenshots, I did not have any major workload on the CPU or GPU.

Firstly, check the cable at the LoBo and make sure it's seated properly but since you haven't been working in that area it's likely fine. Then check the other end, which ll be harder! It's possible you damaged the other end (at the fan) when you removed the PSU or were cleaning the fan out. I've seen old dry/crappy solder joints broken just form a blast of compressed air before :-(

Thanks for your detail explanation. I checked the cable at the LoBo and made sure it was seated correctly. Now I think you might be on something on the other end. I was very careful when removing the PSU but it might well be the blast of the compress air like you mention ! I will check this...

What I would check:
- system monitor info (comparing the temp reading with it).
- another temp, amp, fan revolution software (just to be sure).

I am not sure what you mean by checking "the system monitor info"? isn't it what my screen grabs show?
Any suggestions for another software temp, amp, fan software?

Thank you

 

[tommy chen]

Excuse me but I have no idea what the northbridge rivets looks like. Do you have a link for a picture?

 

[Tordenver]

Thanks Amedias for your help!

some thoughts....

1. The PSU fan is not inside the PSU, it's between the PSU and the front optical bays, and it's plugged into the logic board at the front underneath the optical bays so a 'bad' PSU shouldn't have any impact on the fan as it's driven from the LoBo. *If* your PSU temps were high then the SMC would increase the PSU fan RPM to increase cooling, so a bad sensor in the PSU could cause high RPM, but your PSU temps are fine so I don't think you have a PSU issue.

That makes sense. What about the fluctuations in the voltage/amp of the PSU? is that normal behavior?

Oops sorry I see I misquoted Mikas when it was you explaining the possible bad cable connection to the PSU fan!

If that was at idle then as mikas posted above I'd be checking your NB heatsink and pins *right now* especially since you have >20C different between the NB Diode and heasink, this suggests poor thermal contact between the NB and the heatsink. This could very well be the cause of your shutdowns as if it's runnin 83C at idle then it'll be running much hotter under load and can cause the shutdowns. Those rivets are getting old now and brittle, they can fail on their own without any interference but any knocks and bumps from taking a tray or PCI card in and out could have been the last straw and caused one to pop.

Sorry for asking but NB is for "Northbridge" right? Once I find out what a rivet looks like, how difficult is it to replace? I am the one who did the cpu upgrade 3 years ago. could it be I damaged something at the time?

Again the shutdowns happened just after I did the software update and I had to swap my GPU cards etc... before that everything seemed to work just fine. So I was wondering if they could be related to dust in the PSU? right now I do not want to stress out the system because of the fan issue, so I cannot test if the shutdown still happens... yesterday I ran the computer for 30 min. just to get some files from my internal RAID with no shutdown if this is any indications...

The PSU fan noise started after I removed the PSU, dust it off with air pressure... so I could see the problem you are describing with the cable being real... maybe I should just replace the fan with a new cable, what do you think?

Thanks again for your help. Very much appreciated.
[automerge]1587746565[/automerge]

View attachment 908868

Great, thanks IndioX!

Are they good looking rivets on your picture?
So I guess if the rivets are bad, the heatsink is not touching the board anymore... hence the temperature increase?

 

[Snow Tiger]

Are your thermal sensors reporting in Fahrenheit or Celsius ?

At any rate , it looks as if you damaged or impaired the PSU fan during your recent servicing . The fact that it runs at its default ( highest ) rotationals but does not report itself properly is evidence of that .

A thorough tech would run an ASD for verification , but we know what the answer would be .

So , it appears your chassis died , as the PSU fan is a part of the chassis and not physically connected to the PSU . And if the PSU fan failed , it is generally considered non replaceable so a case swap seems to be in order .

You can try replacing that fan , if you like . It requires the use of four proprietary rubber grommets to be retained in the chassis . If you don't install it properly , it might rattle . And of course a System tear down is required due to Apple's cable management . The data / power cable of that fan is snaked behind the backplane board ( logic board ) but is connected to a header on the front side of the board .

It looks like this and insist the seller of the replacement provide the four grommets , if you go that route .

Personally , I'd just re-case your Mac . I'm pretty certain that's what Apple would do as there is no operation discussed in the Services Sources on this topic .

There's an outside chance you actually damaged your backplane board instead and your fan is working properly ...

 

[tommy chen]

Are they good looking rivets on your picture?
So I guess if the rivets are bad, the heatsink is not touching the board anymore... hence the temperature increase?

thats right

if bad the rivet is missing

Mac Pro 5.1 crashing after updating to Catalina

Tsialex provided the correct link on fixing the Northbridge. Usually you need to gently remove the Northbridge heatsink, apply a new thermal paste and a new rivet screw. Check the link tsialex gave and here’s a video on repairing the Northbridge.

forums.macrumors.com

but please be more careful than the guy in the video ; -)



.

 

[amedias]

replacing the case for the sake of a £10-£20 fan seems a drastic suggestion!

I’ve changed those fans before, it’s time consuming but not technically hard.

Putting my little one to bed so will reply in more detail later

 

[Snow Tiger]

thats right

if bad the rivet is missing

Mac Pro 5.1 crashing after updating to Catalina

Tsialex provided the correct link on fixing the Northbridge. Usually you need to gently remove the Northbridge heatsink, apply a new thermal paste and a new rivet screw. Check the link tsialex gave and here’s a video on repairing the Northbridge.

forums.macrumors.com

but please be more careful than the guy in the video ; -)



.

The Northbridge Controller Heatsink is retained with two proprietary push pins . They are not rivets . The Heatsink was designed to slightly rise and fall , based on the NB chip's temperature , to prevent the die from cracking . Rivets will fix the NB heatsink in place and this runs the risk of the die cracking ( and will kill the CPU Tray's PCB ) . Please only use the proper push pins as replacements . Avoid rivets .

 

[amedias]

Since you've confirmed no real load at the time of the screenshots then > 80deg C for the NB is not good, I really do suspect the heatsink fasteners now as the cause of your instability.

The North Bridge is as identified in the photo posted above, you'll need to remove the CPU heatsink to see it fully, and in order to replace the fasteners you need to remove the CPU/Ram daughter board from the mounting tray as you need to get to the back side of the board.

Failure of the fasteners is common at this age (and sooner!) and is characterised by high NB diode temps, and especially if you see high NB temp but with a lower than expected NB heatsink temp, or a large difference between them. Normally when the NB and heatsink have good thermal contact the temps will be close or within ~10deg C of each other. Once the pins degrade or break the NB is no longer in good contact so heat transfer between the NB is reduced so the NB temp goes up and heatsink temp does not increase as much with it like it should.

Push-pin/spring rivet is a fairly interchangeable terms and they're listed as both on various electric retailers. but DO make sure you get the right ones though, as actual rivets would be very bad! The reason springs are used is to allow for thermal expansion of the NB and to maintain correct contact.

I you can't source the original Apple ones then other compatible pins are available, and the same function can be achieved with nylon screws, washers and springs. There's various threads on here and other places detailing their fitment.

Regarding the PSU fan issue...

There's nothing to indicate that your PSU or the temp sensors for it are damaged as they are still working and giving readings so I think it is *just* the sense/tacho signal on the fan that is broken, either a faulty fan or a damaged connection for the wire that carries that signal.

The Voltage variations in your screenshot are between 12.23V and 12.29V, it may look like a big spike on the graph but if you look at the values it's only actually a tiny fluctuation, just magnified as they've graphed between max an min values, not max and zero.
Fluctuations in current (amps) will be due to load, higher load will draw more current, again though the variations are not out of spec so nothing to worry about there.

So my internet diagnosis (which could be wrong as I'm not physically there to poke about) is that either the PSU fan died coincidentally at the time of your work, or that you have accidentally damaged the fan or it's cable when you were working on the machine.

The connector at the LoBo end is unlikely to have been an issue so I'm not surprised it was OK. The connection at the fan end is soldered to the PCB in the fan, and it's a delicate connection at the best of times, a tug or knock on an already taught wire can easily cause issues, especially if the joint wasn't great to begin with.

Annoyingly you'd need to remove the fan to test it to confirm, and once you've gone to the bother of removing it you might as well replace it anyway rather than attempt to fix it. I personally wouldn't advise a chassis swap, it's unnecessarily expensive when the fan can be replaced, admittedly you have to do 90% of the work of a chassis replacement to do it though!

If you decide to go down the replacement fan route then I would order one in advance. When it turns up you can then disconnect the bad fan, and connect the new one to the header on the LoBo to confirm it operates as you expect.
If it does then you've proven the fan was faulty and can continue with replacing it. If the new fan doesn't work as it should then you've proved the is something else going on and perhaps saved yourself the bother of dismantling your machine to replace that fan.

My money is on the fan needing replacement, fortunately they're relatively cheap to buy!

 

[tommy chen]

The Northbridge Controller Heatsink is retained with two proprietary push pins . They are not rivets . The Heatsink was designed to slightly rise and fall , based on the NB chip's temperature , to prevent the die from cracking . Rivets will fix the NB heatsink in place and this runs the risk of the die cracking ( and will kill the CPU Tray's PCB ) . Please only use the proper push pins as replacements . Avoid rivets .

yes rivet is the wrong expression but unfortunately it is burned in ; -)

 

[mikas]

I had to google rivet.

It is indeed a wrong term to use here.
A spring rivet seems to be more accurate.

Thanks to you all for pointing that out.

I have yet to suffer from broken spring rivets myself, but I have got those ready already for all three of my Mac Pros. There has been discussions about the material of spring rivets too, weather nylon is good enough for the temperatures found on Mac Pro north bridge. I believe my spare spring rivets are the nylon ones. It won't be melting down in there, but I really am not sure what happens before melting happens.

 

[Tordenver]

Thanks Snow Tiger for your feeddback.

Are your thermal sensors reporting in Fahrenheit or Celsius ?

This is correct - Celsius

At any rate , it looks as if you damaged or impaired the PSU fan during your recent servicing . The fact that it runs at its default ( highest ) rotationals but does not report itself properly is evidence of that.

Yes I agree. I believe Amedias explained it pretty well... what might have happened.

So , it appears your chassis died , as the PSU fan is a part of the chassis and not physically connected to the PSU .

I am not sure I understand why you say "the chassis died" when we are talking about the fan being the problem. I get it it is a tricky replacement.

It requires the use of four proprietary rubber grommets to be retained in the chassis .

I am not sure where these are located exactly. Do you have a handy picture?

The data / power cable of that fan is snaked behind the backplane board ( logic board ) but is connected to a header on the front side of the board .

If I get a brand new fan, what about if I disconnect the old cable, leave it in place, and run the new one from above. I do not have a DVD player anymore there, I replaced it a long time ago by some small SSD drives. They don't take much room there. I still need to check the cable path but it seems to me it could work?

Personally , I'd just re-case your Mac

I am not sure what that means

There's an outside chance you actually damaged your backplane board instead

I am not sure what is the "backplane board"?

Thanks again
[automerge]1587836345[/automerge]

The Northbridge Controller Heatsink is retained with two proprietary push pins . They are not rivets . The Heatsink was designed to slightly rise and fall , based on the NB chip's temperature , to prevent the die from cracking . Rivets will fix the NB heatsink in place and this runs the risk of the die cracking ( and will kill the CPU Tray's PCB ) . Please only use the proper push pins as replacements . Avoid rivets .

Thanks for the explanation
[automerge]1587837099[/automerge]

Since you've confirmed no real load at the time of the screenshots then > 80deg C for the NB is not good, I really do suspect the heatsink fasteners now as the cause of your instability.

The North Bridge is as identified in the photo posted above, you'll need to remove the CPU heatsink to see it fully, and in order to replace the fasteners you need to remove the CPU/Ram daughter board from the mounting tray as you need to get to the back side of the board.

This make a lot of sense. Now I am planing to confirm your hypothesis this weekend. I appreciate you sharing your experience what to watch for.

There's nothing to indicate that your PSU or the temp sensors for it are damaged as they are still working and giving readings so I think it is *just* the sense/tacho signal on the fan that is broken, either a faulty fan or a damaged connection for the wire that carries that signal.

Got it. I am going to try to get some other software reading to confirm.

So my internet diagnosis (which could be wrong as I'm not physically there to poke about) is that either the PSU fan died coincidentally at the time of your work, or that you have accidentally damaged the fan or it's cable when you were working on the machine.

I think you are doing a pretty good job for someone who has not seen my machine ?... and again I appreciate the troubleshooting thinking based on facts and what could have happened.

If you decide to go down the replacement fan route then I would order one in advance. When it turns up you can then disconnect the bad fan, and connect the new one to the header on the LoBo to confirm it operates as you expect.

That makes sense. One thing I am not clear... can I test to see if the new fan is working without closing the computer? can I just connect to power without anything else connected (GPU to monitor, etc...) ? or will it interfere with my system behavior?

 

[Tordenver]

View attachment 909096

Thanks Mikas for the picture

 

[Snow Tiger]

Snow Tiger said:
Are your thermal sensors reporting in Fahrenheit or Celsius ?

"This is correct - Celsius"

OK , if the Northbridge chip at System idle is reporting a temperature of 83 degrees C , then this is really bad . It's way too hot . You are burning out your Mac . it's like running an automobile engine without enough oil .

Your Northbridge Controller Heatsink needs to be rebuilt - this means you need to remove the old dried factory thermal paste on this chip and replace it with fresh and better paste . I recommend Arctic MX-4 . This is not the same company that make Arctic Silver 5 . You also need to examine the push pins of the Northbridge heatsink to see if they are broken .

Take a look at the Northbridge Controller Chip temperature in one of my more recent Builds . This is for a properly rebuilt and cooled Mac , using MFC fan utility at the appropriate settings . The chip's temperature is at 64 degree C . This is for a workstation with all the major components ( CPU , memory , GPU and NVMe drive ) concurrently at full load for 24 continuous hours - that's really stressful . It is not at idle . The memory test just ended as I took the screen grab . Other tests are still active in the background .

 

[Snow Tiger]

So , it appears your chassis died , as the PSU fan is a part of the chassis and not physically connected to the PSU .

"I am not sure I understand why you say "the chassis died" when we are talking about the fan being the problem. I get it it is a tricky replacement."

What I mean to say is , Apple would probably regard your System Chassis in need of replacement . In the USA , we would say it needs a re-casing . This is because the time to replace a failed PSU fan takes too long and that fan is regarded as being a permanent part of the chassis . And the tech would have to replace four custom grommets into the case with a replacement PSU fan . There might be another reason , too .

Maybe Apple doesn't like its techs playing around with proprietary grommets ? Could be a vibrational concern . If a locally installed grommet was improperly installed , it could cause a lot of rattling for the customer . That would definitely bring a customer back to the shop .

So maybe Apple likes it techs to only use fan assemblies with grommets preinstalled .

The PSU fan , if installed at the factory in the chassis , has its grommets preinstalled .

The Processor Cage intake and exhaust fan assemblies have factory preinstalled grommets .

The CPU Heatsink fan assemblies also have factory preinstalled grommets .

The PCIe fan assembly also has factory preinstalled grommets .

All the technician replaceable fans come with preinstalled grommets in these machines .

 

[Snow Tiger]

It requires the use of four proprietary rubber grommets to be retained in the chassis .

"I am not sure where these are located exactly. Do you have a handy picture?"

They look similar to these :

But Apple uses proprietary grommets .

Some people call these fan mounts or screws .

They are installed in holes in the fan . Then you pull them by their skinny ends through the holes located in the chassis . They are supposed to secure themselves at that time .

The problem is , Apple used proprietary sized grommets and all of my PC industry standard grommets do not work properly with the cMP PSU Fan . They are too loose . And Apple will not sell the grommets by themselves .

 

[Tordenver]

Take a look at the Northbridge Controller Chip temperature in one of my more recent Builds . This is for a properly rebuilt and cooled Mac , using MFC fan utility at the appropriate settings . The chip's temperature is at 64 degree C . This is for a workstation with all the major components ( CPU , memory , GPU and NVMe drive ) concurrently at full load for 24 continuous hours - that's really stressful . It is not at idle . The memory test just ended as I took the screen grab . Other tests are still active in the background .

Thanks for sharing this screen shot. Indeed always nice to be able to compare to another machine even if my build I am sure is quite different than this one.
[automerge]1587967158[/automerge]

The problem is , Apple used proprietary sized grommets and all of my PC industry standard grommets do not work properly with the cMP PSU Fan . They are too loose . And Apple will not sell the grommets by themselves .

Is it possible to reuse the grommets that are right now used for my PSU fan if they are still in good condition? how difficult is it to remove them without breaking?

If not do you have a place to recommend to get some new ones similar to the Apple ones? (I am located in Colorado, USA) if not what would be best to replace with? I am definitely concerned by noise...

 

[Snow Tiger]

Thanks for sharing this screen shot. Indeed always nice to be able to compare to another machine even if my build I am sure is quite different than this one.
[automerge]1587967158[/automerge]


Is it possible to reuse the grommets that are right now used for my PSU fan if they are still in good condition? how difficult is it to remove them without breaking?

If not do you have a place to recommend to get some new ones similar to the Apple ones? (I am located in Colorado, USA) if not what would be best to replace with? I am definitely concerned by noise...

Grommets are a consumable item . They are intended to be cut off during fan removal , so they are too damaged to be used again .

Your best bet is to buy a wide variety from a number of different sources . I haven't found any yet , though , that are suitable .

There was one fellow here at MR that replaced the factory PSU fan with a Noctua . Maybe he had better luck using Noctua supplied grommets to properly secure one of their fans in the cMP chassis . Trouble is , SMC won't recognize it ( it is not Apple branded ) so it won't have dynamic RPMs controlled by the System . This fan would run at some constant default setting . There is the possibility of using a physical fan control unit to control the rotationals of non-Apple System fans in a cMP . This is done on the PC side with fancy enthusiast builds .

You'd be forced to use a physical controller like this :

But you would need to closely monitor the two PSU thermal sensors in another program ( like Bresink Hardware Monitor ) and make manual adjustments .

 

[amedias]

Is it possible to reuse the grommets that are right now used for my PSU fan if they are still in good condition? how difficult is it to remove them without breaking?

I have re-used the Apple grommets when replacing fans, I've replaced about 15-20 fans now (so ~60-80 grommets re-used) and only ever broken one grommet while doing so

Just be careful, use a small screwdriver and push the barbs of the grommet back through the hole from each side and do it slowly. The grommet will pop through and can be re-used.

As for using other brands, the ones that come with Coolink fans fit fine, you just need to trim the round end down a tiny but but the diameter of the pin of the grommet fits perfectly. The grommets that come with Nanoxia fans can also be made to fit with a tiny trim.

 

[amedias]

Trouble is , SMC won't recognize it ( it is not Apple branded ) so it won't have dynamic RPMs controlled by the System

FWIW this is not entirely true.

There is no communication in a 'data' sense between the SMC and the fans (even the Apple ones) so there is no recognition of brand/hardware ID etc, the SMC cannot tell whether it is or is not an Apple fan.

The SMC manages the fans using feedback on the tacho sense pin to read the RPM of the fan (the same tacho signal that other/PC fans generate), and then adjusts the control voltage to the fan on pin 4 to either increase or decrease the RPM in response to temperature readings from various sensors read by the SMC.

The only thing stopping you using non-Apple fans in a Mac Pro is the fact that nobody else makes fans that use the same method of input control voltage on pin 4 (PC fans use square wave PWM with variable duty cycle, but not variable voltage). But there are ways of using non-Apple fans with SMC control in a Mac Pro by either converting the signal to PWM or driving a fan directly off the control pin.

I've detailed it in this thread:

Mac Pro (4.1/5.1) aftermarket fans WITH automatic SMC control

I've been reading around this subject for a while as a pet peeve of mine is noisy computers. My normal workstations are entirely passively cooled and silence is bliss, but I do still use my cMPs when I need machines with a bit more power or more cores for testing virtual environments and the...

forums.macrumors.com

 

[MIKX]

I replaced my 4,1>5,1's PSU fan with a Noctua.

Noctua NF-A12x25 PSU fan for 4,1 & 5,1 cMP

Today I did a test install of the Noctua NFaA12x25 120mm fan for my 4,1>5,1's PSU. The fan comes with a "low noise / LOWER RPM" adapter cable connected to the original PSU fan socket on the motherboard, NOTE 1 : I had to shave off one of this cable's 'raised locater guides" to allow the Nocuta...

forums.macrumors.com

Apple grommets can be removed using TWO pairs of "needle nose" pliers; Use the pliers to grab the grommet from BOTH ends .. now STRETCH the grommet so that it becomes 'thinner' in diameter - now you can work the slimmer grommet back through the hole in the fan. I've done this 5 or six times successfully.

 

[amedias]

@MIKX if you read my thread above you'll see you can swap a pin on your Noctua and run it off the Mac Pro header on the board WITH control via the SMC. No need for your adapters or to run at constant speed.

 

[Snow Tiger]

FWIW this is not entirely true.

There is no communication in a 'data' sense between the SMC and the fans (even the Apple ones) so there is no recognition of brand/hardware ID etc, the SMC cannot tell whether it is or is not an Apple fan.

The SMC manages the fans using feedback on the tacho sense pin to read the RPM of the fan (the same tacho signal that other/PC fans generate), and then adjusts the control voltage to the fan on pin 4 to either increase or decrease the RPM in response to temperature readings from various sensors read by the SMC.

The only thing stopping you using non-Apple fans in a Mac Pro is the fact that nobody else makes fans that use the same method of input control voltage on pin 4 (PC fans use square wave PWM with variable duty cycle, but not variable voltage). But there are ways of using non-Apple fans with SMC control in a Mac Pro by either converting the signal to PWM or driving a fan directly off the control pin.

I've detailed it in this thread:

Mac Pro (4.1/5.1) aftermarket fans WITH automatic SMC control

I've been reading around this subject for a while as a pet peeve of mine is noisy computers. My normal workstations are entirely passively cooled and silence is bliss, but I do still use my cMPs when I need machines with a bit more power or more cores for testing virtual environments and the...

forums.macrumors.com

Thank you for these insights .

I had naturally assumed Apple used proprietary components or methods and it was not possible to hack a non-Apple branded system fan so that cMP factory SMC could control it .

The problem for the long term durability of these cMPs is not so much the quality of the case fans - they are top notch enterprise grade , though they do fail . The problem is the proprietary rubber grommets that secure the PSU fan to the chassis . After a decade , they are denaturing and failing to secure the fan properly in place - among other concerns , this could cause rattling .

I experienced the same issue with the system fans in the Power Mac G5s once they reached their ten year birthday - the rubber grommets frequently started to disintegrate .

If we can get quality third party replacement system fans ( genuine enterprise grade or high quality consumer grade , like the Noctuas ) installed in these machines , they will last longer .