MP 1,1-5,1 Mac Pro (4.1/5.1) aftermarket fans WITH automatic SMC control

[Nilzy1]

HI Nilzy1

Looking at the specs of the NF-S12A ULN , max rpm is 800....
Unfortunately this is not enough to keep all cool at medium or higher loads.
I'm sorry but I recommend to replace them (all 3 of them) for the NF-A 12x25ULN (runs at 1200 max).
If you did the pin swap on an extension cable, the fan replacement will be far less time consuming than the first time (plug and play so to say).

Yes. the PCI fan is a patience test to get it in
I used the noctua rubbers the other way around to get it in (first in fan then through holes casing).

The NF-A8 FLX (as CPU Boost fan) has not been tested AFAIK , and runs at 2000rpm max , so there is a chance that it will not run fluently at default min rpm = 800...
To stay on the safe side, you could install the NF-R8 redux 1800 , which has been confirmed by others , including myself, to run perfectly as a CPU Boost fan.
In case you prefer to test the NF-A8FLX keep us posted about the results.


In all cases (if you havent ) I'd install e.g. MacsFanControl to check if all is working properly after the fans replacement.

Sorry for the bad news on the NF-S12A ULN !

Thank you for the answer.

I was expecting that I have to replace the NF-S12A ULN fans.
This is a sample of temperatures I have got. The one that worries me the most is the power supply temperature. What is a normal temperature range for that?

Should I be able to see the GPU temperature as well?

Haven't decided if I'm gonna test the NF-A8 FLX fan. I was sure someone already had tested it, but going thru the previous posts I realise I was wrong.

I haven't seen the exhaust/intake fans spin up more than 600-ish yet, when set to auto.

 

[KeesMacPro]

Hi Nilzy1

The one that worries me the most is the power supply temperature. What is a normal temperature range for that?

The PSU Temps can vary a lot depending on different things, so IMO there is no "normal" PSU T under all conditions.
Besides, there are 2 Brands of PSUs used in the MP4,1/5,1 : ACBel and Delta.
I suspect that your PSU is a ACBel , as they are known for higher Temps.
To keep the T lower, I'd set the PSU fan like this:
- sensor PSU power supply 2
- T min 35 degrees
- T max 55 degrees
This will keep the T somewhere in the 40's; the expected life time of capacitors shortens with higher Temps.

Should I be able to see the GPU temperature as well?

Only with (newer) GPUs that support that e.g. RX580, RX Vega etc.

Haven't decided if I'm gonna test the NF-A8 FLX fan.

In case the fan doesnt run at default min rpm , you'll have to do the procedure twice, so I understand your doubts ; )

I haven't seen the exhaust/intake fans spin up more than 600-ish yet, when set to auto.

Your MP is single CPU , they run quite a bit cooler than the dual CPU.
Generally, all fans running at "auto" in this setup works ok; your Temps look all good (except for the PSU IMO).
When under heavy load, I would check the NorthBridge diode Temp.
To keep it very safe, you could set the Intake and Boost fans both like this:
- sensor NB diode
- T min 55 degrees
- T max 75 degrees
This keeps the CPU and NB T low under high load (other fans on "auto" except PSU).
Apple typically sets the SMC by default with priority at low noise and not low Temps, so to get the fans in an Apple device spinning up, Temps are quite high already....

 

[Minimum91]

I want to thank everyone for the helpful information they have provided in this topic. It guided through the topic well enough I could follow as a not too technical person.
I have installed all the Noctua fans correctly and have no trouble with pulsing fans or otherwise.

Intake, exhaust, PS: Noctua NF-A12x25 ULN 120 mm
Boosta: Noctua 80 x 80 x 25mm NF-R8 redux-1800
PCI: Noctua NF-A9 FLX 92 mm Case Fan

However I noticed a significant increase in base temperatures for my 3,46Ghz X5690 with base fan speeds. I didn't take a screenshot beforehand, but I am pretty sure that before changing the fans the CPU A diode temps were around mid 40's while idling. Now they are around 65. I used Arctic MX-5 paste. I now have to increase the base fan speed to about 900rpm with TG Pro to keep it under 60 degrees. I am looking at this wrong?

 

[KeesMacPro]

I want to thank everyone for the helpful information they have provided in this topic. It guided through the topic well enough I could follow as a not too technical person.
I have installed all the Noctua fans correctly and have no trouble with pulsing fans or otherwise.

Intake, exhaust, PS: Noctua NF-A12x25 ULN 120 mm
Boosta: Noctua 80 x 80 x 25mm NF-R8 redux-1800
PCI: Noctua NF-A9 FLX 92 mm Case Fan

However I noticed a significant increase in base temperatures for my 3,46Ghz X5690 with base fan speeds. I didn't take a screenshot beforehand, but I am pretty sure that before changing the fans the CPU A diode temps were around mid 40's while idling. Now they are around 65. I used Arctic MX-5 paste. I now have to increase the base fan speed to about 900rpm with TG Pro to keep it under 60 degrees. I am looking at this wrong?

Good to hear that the info provided here was helpful to you!

Actually after a replacement of all fans for the models you installed , all Temps should be equal or a bit lower (and much lower noise level) with the fans running the same rpm as before the replacement.
If the CPU A Temp is higher , how are the NB diode and heatsink Temps?

If all fans work correctly the only thing I can think of is that you accidently installed a fan reversed so blowing in the opposite direction...

If you could post a screenshot of TG Pro with all fans and Temps visible , that might be helpful.

 

[Minimum91]

Good to hear that the info provided here was helpful to you!

Actually after a replacement of all fans for the models you installed , all Temps should be equal or a bit lower (and much lower noise level) with the fans running the same rpm as before the replacement.
If the CPU A Temp is higher , how are the NB diode and heatsink Temps?

If all fans work correctly the only thing I can think of is that you accidently installed a fan reversed so blowing in the opposite direction...

If you could post a screenshot of TG Pro with all fans and Temps visible , that might be helpful.

All of the fans including the Boosta are installed with the front facing forward, which I think is correct. What kind of processor is your machine sporting and how are your CPU temps on idle?

 

[KeesMacPro]

What kind of processor is your machine sporting and how are your CPU temps on idle?

I have different MPs so I attached a screenshot of a single CPU X5690 , exactly like yours.
Note that the ambient T here at the moment is lower than yours so all T are a bit lower.
Nevertheless I think delta CPU diode- heatsink is too high in your case (if at idle) : 8 degrees (should be just ~3 Celsius).
I would do a repaste of heatsink CPU A to start with.
I've never personally tested mx-5 , but mx-4 and arctic silver are standard thermal pastes that work fine.
Thermal Grizzly kryonaut is outstanding but expensive (and not really cost effective in this case IMHO).

All of the fans including the Boosta are installed with the front facing forward, which I think is correct.

To avoid confusion about what's the "front" of the fan I'll try to describe it a bit different:
When I open up the MP and look at the fans, all of them are installed with the labels facing backwards ( the model and brand label).
BTW there's also an arrow mark on the frame of the noctuas (indicates the airflow direction)....
Airflow should be from frontside MP= intake to backside MP=exhaust.

Hope to be of some help!

 

[avro707]

All of the fans including the Boosta are installed with the front facing forward, which I think is correct. What kind of processor is your machine sporting and how are your CPU temps on idle?

This looks like a fan or fans might be pushing the wrong way or maybe the thermal compound is not applied properly? The fans should all be operating in a push /pull arrangement. The intake fans are moving air into the case, and the ones at the back move it out of the case. All should be in the same direction.



My six core X5690 runs at 33°C (tdiode) at idle with stock Apple fans and default Apple fan control. The heatsink reads about 5 degrees C higher. I used Arctic silver 5. I’m in a warm climate too.

My dual 6 core machine also runs cooler than your single processor one too, so something seems odd with yours. Hmm.

 

[KeesMacPro]

My dual 6 core machine also runs cooler than your single processor one too, so something seems odd with yours. Hmm.

Exactly, a single CPU should run way cooler than a dual.
The DIMMs T is quite high ~ 50 Celsius perhaps the Exhaust fan is reversed...

 

[avro707]

Exactly, a single CPU should run way cooler than a dual.
The DIMMs T is quite high ~ 50 Celsius perhaps the Exhaust fan is reversed...

Yes, good catch. Here are mine for comparison on a 128GB machine with RX580 and 2x X5690:

Only non-standard modification is a little Noctua 40mm 12v fan put in front of the Northbridge heatsink to add a bit more airflow there, something that is necessary for these dual processor machines like mine which is at a spec that Apple couldn't build at the time and probably didn't consider back then.

Another thing worthwhile doing is comprehensively de-dusting the whole computer, but this sounds like a fan not pointed the right way.


For the topic, I did consider Noctua fans to replace the default ones but ended up buying a a number of full sets of case fans for the Mac Pro 5,1, enough to keep my two 5,1 machines going.

Noctua's R&D area are aware of this particular topic and that we are using their fans for our old beloved workstations.

 

[Minimum91]

I have different MPs so I attached a screenshot of a single CPU X5690 , exactly like yours.
Note that the ambient T here at the moment is lower than yours so all T are a bit lower.
Nevertheless I think delta CPU diode- heatsink is too high in your case (if at idle) : 8 degrees (should be just ~3 Celsius).
I would do a repaste of heatsink CPU A to start with.
I've never personally tested mx-5 , but mx-4 and arctic silver are standard thermal pastes that work fine.
Thermal Grizzly kryonaut is outstanding but expensive (and not really cost effective in this case IMHO).


To avoid confusion about what's the "front" of the fan I'll try to describe it a bit different:
When I open up the MP and look at the fans, all of them are installed with the labels facing backwards ( the model and brand label).
BTW there's also an arrow mark on the frame of the noctuas (indicates the airflow direction)....
Airflow should be from frontside MP= intake to backside MP=exhaust.

Hope to be of some help!

This looks like a fan or fans might be pushing are the wrong way or maybe the thermal compound is not applied properly? The fans should all be operating in a push /pull arrangement. The intake fans are moving air into the case, and the ones at the back move it out of the case. All should be in the same direction.



My six core X5690 runs at 33°C (tdiode) at idle with stock Apple fans and default Apple fan control. The heatsink reads about 5 degrees C higher. I used Arctic silver 5. I’m in a warm climate too.

My dual 6 core machine also runs cooler than your single processor one too, so something seems odd with yours. Hmm.

I have tested the fan direction by waving a smoking stick around the fans and it seems like the airflow is correct – from the front, through the heatsink, out the back. The fans are placed with the sticker facing the back. Replaced the thermal paste to MX-4.

This seems to have produced no changes and rules out thermal paste and fan positioning as the culprit. I am also wondering - how can it be that the temperatures for the RAM also got higher by 15°C. Starting to get worried.

FYI to be able to change the PSU fan I had to take out the motherboard, since the fan cable is tucked behind it. Whilst doing so I completely cleaned the computer inside out. Not a speck of dust left inside. I was using an extremely soft brush to remove off the stuck dust with the vacuum cleaner about 5 centimeters away from any capacitors or electronics so that I don't accidentally damage anything.

Do you guys have any idea what I could test next besides obtaining a spare CPU tray?

 

[KeesMacPro]

I have tested the fan direction by waving a smoking stick around the fans and it seems like the airflow is correct – from the front, through the heatsink, out the back. The fans are placed with the sticker facing the back. Replaced the thermal paste to MX-4.

This seems to have produced no changes and rules out thermal paste and fan positioning as the culprit. I am also wondering - how can it be that the temperatures for the RAM also got higher by 15°C. Starting to get worried.

FYI to be able to change the PSU fan I had to take out the motherboard, since the fan cable is tucked behind it. Whilst doing so I completely cleaned the computer inside out. Not a speck of dust left inside. I was using an extremely soft brush to remove off the stuck dust with the vacuum cleaner about 5 centimeters away from any capacitors or electronics so that I don't accidentally damage anything.

Do you guys have any idea what I could test next besides obtaining a spare CPU tray?

TBH this is a puzzle to me.
However, before thinking about another CPU tray, I'd start searching for the culprit methodically.
After all, all T were good before the fan swap, so obviously something's changed related to that .
I suppose the screenshot is made running at idle and the side door closed....

Looking at all Temps now:
- the Delta CPU diode-heatsink has not improved after repaste (9 Celsius)
- CPU and the DIMMs are way too hot

The DIMM T is mainly influenced by:
- the exhaust fan
- for being at the end of the airflow in the CPU compartment , it also gets warm air from the hot parts that are positioned in front or close in this "tunnel" i.e. CPU , NB .

As far as I can tell , based on the information here, I'd say that the CPU Boost and/or Exhaust fan are not blowing in the correct direction.
In both cases warm air will accumulate and/or stay too long somewhere in the middle/ end of the "tunnel".
This is exactly what the Temps show.....

 

[Minimum91]

TBH this is a puzzle to me.
However, before thinking about another CPU tray, I'd start searching for the culprit methodically.
After all, all T were good before the fan swap, so obviously something's changed related to that .
I suppose the screenshot is made running at idle and the side door closed....

Looking at all Temps now:
- the Delta CPU diode-heatsink has not improved after repaste (9 Celsius)
- CPU and the DIMMs are way too hot

The DIMM T is mainly influenced by:
- the exhaust fan
- for being at the end of the airflow in the CPU compartment , it also gets warm air from the hot parts that are positioned in front or close in this "tunnel" i.e. CPU , NB .

As far as I can tell , based on the information here, I'd say that the CPU Boost and/or Exhaust fan are not blowing in the correct direction.
In both cases warm air will accumulate and/or stay too long somewhere in the middle/ end of the "tunnel".
This is exactly what the Temps show.....

Yes, when the screenshot was made, the door was closed. There is a lot of space behind the machine.

To get the incorrectly positioned fans theory out of the way, here's a crude airflow test:

 

[MacNB2]

I have tested the fan direction by waving a smoking stick around the fans and it seems like the airflow is correct – from the front, through the heatsink, out the back. The fans are placed with the sticker facing the back. Replaced the thermal paste to MX-4.

As @KeesMacPro suggests, there seems to be an airflow issue.
Take out the CPU tray and shows pictures of the Intake & Exhaust fans.
Similarly, take pictures of the CPU A and CPU B heatsinks so that the fans are visible.

Here are pics of my Stock Fans:
Intake & Exhaust:


CPU A & B fans:


The other issue could be that you may not be tightening the CPU heatsink bolts enough but please be careful when taking this advice as you could damage your board & CPU's. An extra 1/16th or 1/8th turn on each bolt might help.

 

[MacNB2]

Yes, when the screenshot was made, the door was closed. There is a lot of space behind the machine.

To get the incorrectly positioned fans theory out of the way, here's a crude airflow test:
View attachment 1901428

I guess were we both posting at the same time

OK so it looks like both the Intake & Exhaust are facing the same way.
There seems be some turbulence in the smoke between the Intake and CPU heat sink.
That is, I would have expected the Intake fan to blow the smoke straight into the CPU and the CPU fan to suck the air straight in but it's "hesitating".

Try increasing manually the CPU fan speed to full and see what happens to the smoke test between the Intake the CPU heatsink.

 

[arw]

Concerning the whole mod, could there be future problems as the main current draw is from the „controller pin“ of the original fan instead of the original VCC?

 

[KeesMacPro]

Concerning the whole mod, could there be future problems as the main current draw is from the „controller pin“ of the original fan instead of the original VCC?

I suppose that with:
- "the whole mod" you refer to a fan replacement with a 3-pin 3rd party fan
- "controller pin" you refer to the variable voltage pin 4 of the LogicBoard header
- "original Vcc" you refer to pin 2 of the LogicBoard header

If so, I understood correctly ; )

Pin 2 in an untouched setup is providing a constant 12V to feed the controller inside the OEM Delta fans and draws very little current.
Pin 4 in an untouched setup is providing a variable Voltage and the fan motor itself draws all current from this pin.
So after a fan mod as described here there's essentially nothing changed concerning current draw.
At least not for pin 4 (variable voltage).

Another more practical proof that this mod is absolutely safe is the fact that , as many many others, I've done this mod almost 2 years ago on different MP 4,1/5,1 and never experienced any issues whatsoever.

 

[arw]

I suppose that with:
- "the whole mod" you refer to a fan replacement with a 3-pin 3rd party fan
- "controller pin" you refer to the variable voltage pin 4 of the LogicBoard header
- "original Vcc" you refer to pin 2 of the LogicBoard header

If so, I understood correctly ; )

Pin 2 in an untouched setup is providing a constant 12V to feed the controller inside the OEM Delta fans and draws very little current.
Pin 4 in an untouched setup is providing a variable Voltage and the fan motor itself draws all current from this pin.
So after a fan mod as described here there's essentially nothing changed concerning current draw.
At least not for pin 4 (variable voltage).

Another more practical proof that this mod is absolutely safe is the fact that , as many many others, I've done this mod almost 2 years ago on different MP 4,1/5,1 and never experienced any issues whatsoever.

Thank you, yes, you understood correctly.
I on the other hand was obviously wrong in thinking, the logic of the Delta fans was like the PowerMac G5 ones‘ (variable voltage only for controlling, although the controlling voltage there was 0 to 5V).
Thanks for enlightening me, must have overread it.

 

[KeesMacPro]

Thank you, yes, you understood correctly.
I on the other hand was obviously wrong in thinking, the logic of the Delta fans was like the PowerMac G5 ones‘ (variable voltage only for controlling, although the controlling voltage there was 0 to 5V).
Thanks for enlightening me, must have overread it.

You're very welcome!

 

[KeesMacPro]

Yes, when the screenshot was made, the door was closed. There is a lot of space behind the machine.

To get the incorrectly positioned fans theory out of the way, here's a crude airflow test.

Hi,

I'm curious if you found the culprit.
If the fans are not the issue, the only thing I can think of is insufficient CPU heatsink- die thermal contact.
Perhaps you're aware of this, but for the single CPU heatsink, make sure to first tighten the 4 screws close to/ around the CPU and the 5th (close to the NB) at last , without overtightening it to avoid a tilted heatsink position and/or undesired forces on the PCB.

In case you solved it, if you want , post your findings to help others running into the same issue.
Thx in advance,
Best.

 

[Minimum91]

Hi,

I'm curious if you found the culprit.
If the fans are not the issue, the only thing I can think of is insufficient CPU heatsink- die thermal contact.
Perhaps you're aware of this, but for the single CPU heatsink, make sure to first tighten the 4 screws close to/ around the CPU and the 5th (close to the NB) at last , without overtightening it to avoid a tilted heatsink position and/or undesired forces on the PCB.

In case you solved it, if you want , post your findings to help others running into the same issue.
Thx in advance,
Best.

Hi,

I did a loads of testing and what I found is quite curious, if not Kafka-esque. Still needs needs more polish though.

In short, my high CPU temperature issue was to do with an outdated IOPlatformPluginFamily.kext I installed just a couple weeks prior to the fan replacement. I did it as a recommendation via another MacRumors thread dealing with high pitched noise coming from the CPU compartment when connecting a FireWire audio device. The old kext fixed the high pitch noise, but made the CPU 15-20°C hotter (more on that later). Obviously I just did not notice this higher pitch sound prior to the fan replacement as the old fans were significantly louder. For now, I have unloaded the old kext and brought back the stock one supplied with Mojave, with which the noise has returned. While doing so I have made a couple of screenshots that understandably might not seem convincing, but here they are.


I have noticed a strong correlation between the sensors - CPU A Diode and CPU A Core, Relative to ProcHot. If one goes up, the other one goes down and vice versa. Could it be that the Kext intentionally or not swaps the stats coming from these sensors?

After returning the kexts back to stock I proceeded to monitor temps and as you can see, they did return to a more sane range at least for the CPU A Diode. But what I am noticing now is that pushing the fans to go faster does not necessarily decrease the temperature!
Take a look at these two screenshots. First one was taken with my custom settings (the way I run it all day) and the second is 8 minutes later after turning all the fans down!



My TG Pro Settings


Sincerely,
Max

 

[avro707]

Those CPU temps looks normal now.

 

[KeesMacPro]

Hi,

I did a loads of testing and what I found is quite curious, if not Kafka-esque. Still needs needs more polish though.

Hi,

Thx for the update, much appreciated!

In short, my high CPU temperature issue was to do with an outdated IOPlatformPluginFamily.kext I installed just a couple weeks prior to the fan replacement. I did it as a recommendation via another MacRumors thread dealing with high pitched noise coming from the CPU compartment when connecting a FireWire audio device. The old kext fixed the high pitch noise, but made the CPU 15-20°C hotter (more on that later).

Back in the day, I've had the Duet FW too , for simple editing, mastering, arrangements and compositions in my Home Studio.
I remember the high pitch noise: very annoying!
The driver for this device is just "more or less" compatible with an OS like Mojave...
Besides installing a KEXT from 10.6 is adventurous IMHO...
There is a DUET(2) USB model and recently a 3rd version came out; might be worth upgrading to one of them , sound quality is far superior and they are fully supported running Mojave and later.

Obviously I just did not notice this higher pitch sound prior to the fan replacement as the old fans were significantly louder.

Sounds familiar: )
This goes for all sounds the MP internals produce . One of the reasons, besides I/O speed, why I replaced all mechanical HDDs in my MPs for blades and/or SATA SSDs.
Still use some spinners though, but only as external for e.g. backups.

I have noticed a strong correlation between the sensors - CPU A Diode and CPU A Core, Relative to ProcHot. If one goes up, the other one goes down and vice versa.

Relative to ProcHot = how many degrees away from thermal throttling, that's why it's counting "reversed".

After returning the kexts back to stock I proceeded to monitor temps and as you can see, they did return to a more sane range at least for the CPU A Diode. But what I am noticing now is that pushing the fans to go faster does not necessarily decrease the temperature!

I fully agree with @avro707 : all Temps look good .
A single CPU stays relatively cool without a lot of fan settings, so with the fans at min rpm it will run quite cool and definitely safe at idle.
More interesting is what happens at heavier loads , personally I'd set the fans according to the Temps when running heavy loads of the kind you do on this device.

Good to hear your MP is running fine!

 

[avro707]

The single processor MP5,1s are really good machines, even upgraded with a single X5690 and 64GB ram (4x16GB) those are excellent solid machines, they run pretty cool with standard original Apple cooling and fairly quiet.

I run my less powerful MP5,1 every day when I do my 2 hours per day of riding on Zwift and it runs faultlessly with that. I prefer it to Apple TV4K (which I also have) for Zwift, it can run higher detail. Although admittedly a remote control device similar to the new Apple TV one (the silver coloured remote) might be handy for Zwift.

That's a narrow use case - but given that machine was originally only purchased for spare parts and not used, it's great. I figured it was cheap enough to upgrade it into a fairly up to date working machine, why not. It drives a 55" Hisense 4K television that is wall-mounted. Happy days. It even has thunderbolt, as the other one does too. And with that they happily run the LG 5K Ultrafine display at 5K resolution no dramas at all.

Looks like Minimum91 has a great result there.

 

[davorrr]

USA CASE
I only make music on a pretty low volume level on my maxed out cMP 5.1 dual cpu. I'd like to replace all the fans by this elegant method to get the machine quiet.

Sometimes I think I am the most sensible person in the world regarding the fan noise. I officially suffer under hyperaccusis. Before you recommend me a fanless Macbook Air, I have to point out that I need to use Mojave on cMP 5.1 for my TC PowerCores: Got many of these DSP Firewire boxes.

QUESTION
Which fans and which profile would be most radical in order to get the noise level as low as possible?

 

[mnlth0]

I finally started to replace fans in my Dual CPU Mac Pro 5.1. At least partially

A few months ago I started to collect parts but due work stuff I simply didn't manage to get to it. So for a test I tried to do the simple replacement of CPU cage fans using the pin swap method with NF-A 12x25 ULN fans. It could not have been easier! I removed fan pins from Mac Pro original fan, and re-used original header for Noctua fans. That's it! Just make sure that pins are seated firmly so they don't pushed out when you connect the header to the board.

I didn't have time to replace the PSU fan today but I will do it in the following days. Same goes for PCI fan for which I have a question: is it possible to do pin swap instead of soldering on PCI fan?

Attached a few photos of results. So far so good. I actually didn't even reset pram / nvram, no issues yet but I will reset once I swap all fans.

Thank you all for your amazing research, input and help!