Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
I am kind-of late to the party here, however, I think this is the best place to make this post. I have a slightly tweaked approach to cooling the Northbridge that may be of interest to people going through this thread. This thread was of great use to me in my quest to bring my Northbridge temps down, so I thought I should contribute. I may make another post elsewhere still, as my modifications crossover into other territory a bit.

To start with, I upgraded my 2010 5,1 to two 3.46Ghz Xeon Processors. Things are fine there, but I noticed that the Northbridge was quite hot, (likely because of the upgrade), so I started to take steps to lower it. I de-dusted, replaced the Northbridge heatsink clips (with metal ones) and repasted. After all that the temp only came down a couple of degrees. It was running in the high 70s to low 80s.

I fiddled with Mac Fans Control, and that helped a little, but to bring down the temp significantly required the fans to really be blasting, so I gave up on that approach for the most part.

Then I started adding in mini fans. I tried 5 different fan sizes, ranging from blower types, to tiny 3cm fans. Testing with those, the 4cm fan came out on top for cooling, though the 5cm was close, and much quieter. For now I am using the 4cm fan (10mm thick). The thinness helps in getting the fan to point down on the bulk of the heatsink.

However, I didn't like the cables running off the mainboard. (It even caused a bent pin in my processor/mainboard connector once!) I got a great idea from this thread Mac Pro (4.1/5.1) aftermarket fans WITH automatic SMC control
and had power running from a splitter hooked into intake fans 4pin header.

This lead me to my next idea, which was to run a splitter off of the booster fan header. However, it is a, I think, proprietary, 6 pin header. Thanks again to the posters and info in the thread mentioned above, I was able to splice in a fan hub to power the fan 'on' the processor board, via the booster fan header. That should reduce the cable issues.​

I had to customize the hub a bit...
IMG_3875.jpeg

Joined everything with 4pin fan connectors
IMG_3880.jpeg

Here is the hub installed...
30C550A2-7D0D-4EA8-A72A-6D18480CE69E.jpeg

I went one step further and added an extension to the heatsink itself. I noticed a small space between the part of the heatsink that extends into the processor heatsink, and the fan, and so I hacksawed a slice of an old heatsink and slid it into the top of the Northbridge heatsink. You wouldn't think it would work, but it did, dropping temps between 3-6 degrees.​

Hacksawed a slice of heatsink...
IMG_3701.jpeg

Added it to the heatsink (flipped it upside down in the final version, fitting the fins into the Northbridge heatsinks fins)
IMG_3705.jpeg

I had a few extra ports on my spliced in fan-hub, so I added some RGB lights for my memory (shock. outrage.), and linked them via an WLED device, so I can control them from the Menu Bar (with a little app) or App/Web interface. Very factory. But not really relevant to this post.​

RGB Horror/Delight
5FB0E241-8267-4F2A-A23D-543EA35110C7.jpeg

The final result is I have a 4cm x 10mm fan mounted against the Northbridge heatsink, powered from the processor board itself. The Northbridge (diode) temps have been reduced from high 70s and low 80s to high 50s and low 60s. About a 20degree drop overall. And this is in high summer where I am, so winter might yield even better results (started this project in the winter, so original high temps were derived from there).​

F81E068E-F1E3-48D7-AC7C-960D12E38E8E.jpeg83992189-9FEA-4D6B-A24A-35FE025CA005.jpeg

I would like to thank all the posters in this thread, and the aforementioned one, for all their input, as it was invaluable in making my modifications.

I post here now in the off chance that this could help someone else in their efforts.

If there is any interest, I could elaborate further on my RGB mod elsewhere. Or elaborate on the Northbridge modification itself.

Cheers!​
 
Last edited:
I am kind-of late to the party here, however, I think this is the best place to make this post. I have a slightly tweaked approach to cooling the Northbridge that may be of interest to people going through this thread. This thread was of great use to me in my quest to bring my Northbridge temps down, so I thought I should contribute. I may make another post elsewhere still, as my modifications crossover into other territory a bit.​
Thankyou for posting, this is invaluable to all.

To start with, I upgraded my 2010 5,1 to two 3.46Ghz Xeon Processors. Things are fine there, but I noticed that the Northbridge was quite hot, (likely because of the upgrade), so I started to take steps to lower it. I de-dusted, replaced the Northbridge heatsink clips (with metal ones) and repasted. After all that the temp only came down a couple of degrees. It was running in the high 70s to low 80s.​
I completed the same and my investigations led me to believe that the Apple fan controller set up from factory complies with intel's max temp that seems why their thermal engineering team put that sub optimal heat sink on the NorthBridge upstream of the CPUA.

I fiddled with Mac Fans Control, and that helped a little, but to bring down the temp significantly required the fans to really be blasting, so I gave up on that approach for the most part.​
I was able to get this result on Northbridge HeatSink Temp using a ghetto hack. I would be interested to compare to yours with a fan. Note this is My Heat Sink Temp, the Diode Temp is +11ºC and I think that is because the Apple Heatsink is Al not Cu or Au, It has crossed my mind to find some Cu shims to act as a heat spreader but bonding Cu and Al on a large surface is notoriously fiddly and thus probably fruitless.
1663293220971.png


Then I started adding in mini fans. I tried 5 different fan sizes, ranging from blower types, to tiny 3cm fans. Testing with those, the 4cm fan came out on top for cooling, though the 5cm was close, and much quieter. For now I am using the 4cm fan (10mm thick). The thinness helps in getting the fan to point down on the bulk of the heatsink.

However, I didn't like the cables running off the mainboard. (It even caused a bent pin in my processor/mainboard connector once!) I got a great idea from this thread Mac Pro (4.1/5.1) aftermarket fans WITH automatic SMC control
and had power running from a splitter hooked into intake fans 4pin header.​
Thanks for the Link I have

This lead me to my next idea, which was to run a splitter off of the booster fan header. However, it is a, I think, proprietary, 6 pin header. Thanks again to the posters and info in the thread mentioned above, I was able to splice in a fan hub to power the fan 'on' the processor board, via the booster fan header. That should reduce the cable issues.​
I am wondering if you have measured the Total A draw on the SMC ?

Fans consume power in an unusually expensive way following a set of relationships with power essentially = cube of the speed of the fan. Of course this is a multi dimensional problem so If you have been measuring the power that is the easiest way to see how much additional work and there for air is moving. It may also forestall a Baseboard power issue if you're pulling to much on the SMC circuits.
I went one step further and added an extension to the heatsink itself. I noticed a small space between the part of the heatsink that extends into the processor heatsink, and the fan, and so I hacksawed a slice of an old heatsink and slid it into the top of the Northbridge heatsink. You wouldn't think it would work, but it did, dropping temps between 3-6 degrees.​

Hacksawed a slice of heatsink...
View attachment 2048028

Added it to the heatsink (flipped it upside down in the final version, fitting the fins into the Northbridge heatsinks fins)​
Really good idea, I like the elegance of finding a way to extend the heat sink fins. Did you mechanically bond the fin - fin contact ? I l looked around for a similar solution but no one is selling their norrth bridge heat sinks ( yet ) in favour of fancy AIO water loops.
The final result is I have a 4cm x 10mm fan mounted against the Northbridge heatsink, powered from the processor board itself. The Northbridge (diode) temps have been reduced from high 70s and low 80s to high 50s and low 60s. About a 20degree drop overall. And this is in high summer where I am, so winter might yield even better results (started this project in the winter, so original high temps were derived from there).​

I would and I am sure the other posters would like to see your longer term results.
 

Attachments

  • 1663290567792.png
    1663290567792.png
    57.8 KB · Views: 94
Last edited:
Thankyou for posting, this is invaluable to all.

My pleasure. It's nice to show the work to people who might care to see it. (My wife does not)
I was able to get this result on Northbridge HeatSink Temp using a ghetto hack. I would be interested to compare to yours with a fan. Note this is My Heat Sink Temp, the Diode Temp is +11ºC and I think that is because the Apple Heatsink is Al not Cu or Au, It has crossed my mind to find some Cu shims to act as a heat spreader but bonding Cu and Al on a large surface is notoriously fiddly and thus probably fruitless.
View attachment 2065327

What app are you using to get the graph? I could do the same for mine to compare. I can predict though, based on my casual tests, that mine will run slightly hotter than that, though without any tweaking of the fans with MacFansControl (which I think you are using?)

I am wondering if you have measured the Total A draw on the SMC ?

Fans consume power in an unusually expensive way following a set of relationships with power essentially = cube of the speed of the fan. Of course this is a multi dimensional problem so If you have been measuring the power that is the easiest way to see how much additional work and there for air is moving. It may also forestall a Baseboard power issue if you're pulling to much on the SMC circuits.

No, no measuring of any kind. I just went for it. Though I did test it out in stages, and did some stress testing after installation. I was worried that the extra 'draw' on the mainboard might be an issue. However, I have run the fans at full bore, with 2 attached fans and the RGB controller all attached running at full speed/power, several times now, for a goodly amount of time, and have had no issues at all. So far...

Really good idea, I like the elegance of finding a way to extend the heat sink fins. Did you mechanically bond the fin - fin contact ? I l looked around for a similar solution but no one is selling their norrth bridge heat sinks ( yet ) in favour of fancy AIO water loops.

It was some work to cut that heatsink. I did use some er..Grizzly thermal paste to attach/connect the fins. So some decent heat transfer there I think. They are also pressing against each other very tightly. It's a little unstable though, and it would have been nice to weld it together slightly.
I would and I am sure the other posters would like to see your longer term results.

I'll certainly post back here in a few more weeks/months, good or bad.

Also, I have to modify my post above, as I was off by 10mm on the fans. I am using 50mm fans, as they were the ones working best. So my previous post should have 10mm added to the size of each of the mentioned fans.

Apologies for that.
 
My pleasure. It's nice to show the work to people who might care to see it. (My wife does not)


What app are you using to get the graph? I could do the same for mine to compare. I can predict though, based on my casual tests, that mine will run slightly hotter than that, though without any tweaking of the fans with MacFansControl (which I think you are using?)



No, no measuring of any kind. I just went for it. Though I did test it out in stages, and did some stress testing after installation. I was worried that the extra 'draw' on the mainboard might be an issue. However, I have run the fans at full bore, with 2 attached fans and the RGB controller all attached running at full speed/power, several times now, for a goodly amount of time, and have had no issues at all. So far...



It was some work to cut that heatsink. I did use some er..Grizzly thermal paste to attach/connect the fins. So some decent heat transfer there I think. They are also pressing against each other very tightly. It's a little unstable though, and it would have been nice to weld it together slightly.


I'll certainly post back here in a few more weeks/months, good or bad.

Also, I have to modify my post above, as I was off by 10mm on the fans. I am using 50mm fans, as they were the ones working best. So my previous post should have 10mm added to the size of each of the mentioned fans.

Apologies for that.
iStat
 
Ah, okay. I tried that before but trashed it. Never noticed the reports as above. Maybe I'll give it another shot. My 14 days may be up though.
 
northbridge_temp.png


There are my temps on the Northbridge Heatsink last night. As I thought, a few degrees above yours. This was with my 5cm fan mounted.
 
Yeah I had all sorts of issues with my NB running 80º+. Got a replacement CPU tray just in case... when my old one died I found this was from a flashed 4.1 and I had an OEM 5,1. Bought another tray, holy smokes it was SO clean that it looked brand new. Boom, almost 2 years now and my NB is running 50º-51º. For years I had run my 5,1 in an unairconditioned environment... never realizing I was almost cooking it. Then I started using MacFansControl and saw the issue... it needed a fair amount of fan too keep it under 80º.Think both the bad tray and the amazing one cost me under $100. If your NB is out of control, dare I suggest this is one move you can make?
 
Last edited:
View attachment 2068974

There are my temps on the Northbridge Heatsink last night. As I thought, a few degrees above yours. This was with my 5cm fan mounted.
Excellent- I would love to see a bit more detail on how you connected the fan in from the booster fan power.

I also had another 40mm fan but had ugly cables running up to the top of the machine for power. Ugly yes but better to have NB under 60°C.
 
Excellent- I would love to see a bit more detail on how you connected the fan in from the booster fan power.

Okay...however, I don't think I have the best pictures of this procedure.

In any case, I started by getting the pinout for the heatsink fan (from the above mentioned thread). Then I carefully marked off the wires for the fan, and set the temp-sensor wires to the side (not to be cut)

IMG_3638.jpeg


Then, I cut the wires, and use a crimper to attach a male and female 4pin fan connection on either side:

IMG_3645.jpeg


This kept things neat, no soldering, and I could just restore the default setup by plugging these two connections into each other.

IMG_3878.jpeg


Then I made a custom length 4pin fan wire, to be connected on one end to the connector I made for the fan on the heatsink, and the 1st port of the hub. No pic for this, sorry.

The female connector that was attached on the end of the proprietary apple connector (which attaches to the motherboard) was then plugged into the main 'input' for the fan hub. The pic above does not show this well. However, you can at least see the cable from the hub wrapping around the bottom of the ram ports.

The images and my explanation do not really make this clear. I wish I had made a more obvious picture of the setup before I put everything together. However, I have repasted the heatsink at least 5 times, and don't want to do it again.

Basically though, what I have done, is insert the hub between the cable/power coming from the board to the fan, then rerouted the fan into the first port for the hub.

Note: The Deepcool 4 port hub is the best for this job. I tested others, and it is the only one that fits:

IMG_3735.jpeg


It is a tight squeeze, but no issues. If you wrap the wires around the back edge of the processor board like I did, it makes closing the processor tray 'slightly' tight, but fine.

Hopefully that sheds a little more light on the process. I fear not though...
 
Wow, this obsession with NB temperature. NBs are designed for/known for running hot (90°C is nothing special). All you guys are doing is redirecting/obstructing the airflow by placing anything in between, causing the fans to run faster and have more noise for the same airflow.

Placing a cheap fan generating lots of EMC noise above the RAM doesn’t seem to be a good idea either.

If the NB really runs hot, clean the dust out. If this doesn’t help, reapply good quality thermal paste. If this still doesn’t help, try another PS as the caps in it may have gone bad (NBs need clean power!).
 
Wow, this obsession with NB temperature. NBs are designed for/known for running hot (90°C is nothing special).
Well Intel's design may have been great for system like the IBM X346 that lived in a cushy 19ºC Data centre bac kin 2004 But our CMP's generally live under a desk. So 90ºC Design spec is just not how some of us want it. Yes you are right we are obsessed, otherwise we would all have moved on to the MacPro new platform

All you guys are doing is redirecting/obstructing the airflow by placing anything in between, causing the fans to run faster and have more noise for the same airflow.
Yes and No, redirecting over the NB Heatsink is generally the goal IF you choose to run cooler NB.
Placing a cheap fan generating lots of EMC noise above the RAM doesn’t seem to be a good idea either.
Yes and isn't the whole inside of the case EMC noisy ?
If the NB really runs hot, clean the dust out.
Yes Absolutely still need to hygiene and of course more fan speed / fans = more dust intake, in general terms.
If this doesn’t help, reapply good quality thermal paste.
This is how we all started the journey, Kernal panics and mystery shut downs lead to troubleshooting and one of the common CMP problems is overheating NB chipsets. Which is why we are here obsessing about it.
If this still doesn’t help, try another PS as the caps in it may have gone bad (NBs need clean power!).
First time I've heard that but as it turns out I have have changed my PS and it removed some ongoing issues with my configuration.

Be welcome and should you feel the need ... the need for cooling then jump back here.
 
Replacing thermal paste, cleaning out dust and replacing the old push pins is the first thing many people have done, but someone still has to suggest it anyway.

My dual CPU 5,1 is running well with the mods. The only thing that failed was a CPU.
 
It just seems to be a weak point in the design that is fortunately fixable. A popped rivet is the only problem I've had with my CMP in the 14 years it's been running. A couple nylon screws and some new thermal paste has kept it running.
 
  • Like
Reactions: RolfNoot
Well Intel's design may have been great for system like the IBM X346 that lived in a cushy 19ºC Data centre bac kin 2004 But our CMP's generally live under a desk. So 90ºC Design spec is just not how some of us want it. Yes you are right we are obsessed, otherwise we would all have moved on to the MacPro new platform
Time's changed. Northbridges just run hot, I've seen so many servers have their NB's running at 100ºC with zero problems.

Yes and No, redirecting over the NB Heatsink is generally the goal IF you choose to run cooler NB.
There's more than just the NB. Redirecting flow to the NB means less flow elsewhere. In general, more obstructions means less flow in, less flow out, easy as is.

Yes and isn't the whole inside of the case EMC noisy ?
Far from noisy as EMI near the wires of the fan. When you look at the PC designs you'll never see fan-power traces nearby high-speed digital lines.

This is how we all started the journey, Kernal panics and mystery shut downs lead to troubleshooting and one of the common CMP problems is overheating NB chipsets. Which is why we are here obsessing about it.
Are these kernel panics and mystery shutdown linked to the NB? As long as the NB was within spec (5-104ºC, see attachment), you shouldn't have had issues. What temp did you NB get?

First time I've heard that but as it turns out I have have changed my PS and it removed some ongoing issues with my configuration.
Yes, NB may run hot when the PS is bad. A bad PS actually leads to lots of weird issues. So the combination of the hot NB and issues you described leads me to think your PS is bad. I actually replaced one from my working machine recently. Especially with the power hungry X5690s and the modern graphic cards a good PS (or even recapped) is key for a stable system.

Anyway, if keeping your system running stable with these mods helps for you, that's good.
 

Attachments

  • Schermafbeelding 2023-02-26 om 17.43.23.png
    Schermafbeelding 2023-02-26 om 17.43.23.png
    34.7 KB · Views: 95
Are these kernel panics and mystery shutdown linked to the NB? As long as the NB was within spec (5-104ºC, see attachment), you shouldn't have had issues. What temp did you NB get?
Unfortunately, quite a few users experienced stability issue when the NB diode temperature reported above ~90°C.

This is well within the spec, but the reality is that the cMP was de-stablised. And once they spin up the fans / re-paste etc to cool down the NB. The cMP works normal again.

This is why some people tends to cool down the NB more than they have to. Their actions are driven by their own experience.

I am not sure why the cMP was de-stablised even the NB temperature still well within spec. May be some surounding electronics are too old, and more sensitive to temperature. So, it's not really the NB failed, but something else was affected by the NB's high temperature.

I am with the single processor cMP. So, no NB temperature issue. Just keep the heatsink clean, and spin up the fans a bit, than the NB will stay below 50°C.

But for those dual processor cMP with a few NVMe installed. It seems the NB can go really hot and at some point, de-stablised the cMP.

Anyway, I think it's very personal if the user want to mod their cMP. It's their own computer. We can provide all the info we know (including the arguments / knowledge on both sides). But in the end, they make the decision and take the consequence.
 
Last edited:
  • Like
Reactions: tommy chen
Unfortunately, quite a few users experienced stability issue when the NB diode temperature reported above ~90°C.
Where is the NB diode on the MacPro situated? Is it near the die? Maybe the thermal paste wears off causing the diode to measure the wrong temperature and not transferring the heat sufficiently to the heatsink (so actually the NB die gets much hotter than the diode temperature reading).

It still doesn't explain why (lots of) users have zero problems with the standard config while (some) others need a mod. That's why I suspect the causation may probably be elsewhere. It's well known caps wear out over time causing instabilities, that's why I suggest replacing the PSU. A bad PSU can cause the NB to heat up and then additional cooling may (falsely) seem to solve te issue.

So maybe to dig further:
- does the overheating still occur with re-applied paste?
- what's the NB voltage reading (can we actually read it)?
- does swapping the PSU help?
 
Oh and by the way, LOTS of people are applying the thermal paste in a wrong way. People spread out the paste causing air to be trapped between the surfaces.

Look at this video
 
I have a 4.1 cpu board in my test pool that I need to have the cpu temperature below 40 Celsius. Can do that with MacsFanControl. Most of the time I use it to diagnose 5.1 (as the Fans go wild with a 4.1 cpu board in a 5.1 that damage does not hurt).

CPU is not damaged, swapped it of course.
 
I am with the single processor cMP. So, no NB temperature. Just keep the heatsink clean, and spin up the fans a bit, than the NB will stay below 50°C.

But for those dual processor cMP with a few NVMe installed. It seems the NB can go really hot and at some point, de-stablised the cMP.

The single CPU version is much better, it runs far cooler and quieter. Single X5690, 64gb ram and NVMEs is a nice combination. Better still with modern GPU.

The dual CPU models run warmer and sound like a wounded banshee when they are running. The highest I saw the NB get to was about 73°C from memory. With the extra measures I kept it down below 60° and then a quiet running machine.

I never checked with slower CPUs but with all of us using X5690s that can’t help the temperatures either.
 
  • Like
Reactions: h9826790
Where is the NB diode on the MacPro situated? Is it near the die? Maybe the thermal paste wears off causing the diode to measure the wrong temperature and not transferring the heat sufficiently to the heatsink (so actually the NB die gets much hotter than the diode temperature reading).
Not sure. I think the diode is a part inside the chip.

From those who have the rivet broken, then NB diode temperature goes really high (e.g. over 120°C), but the NB heatsink temperature is low. So, it's a sign that even the cooling system is borken down, the NB diode can still report high temperature (correctly?).
 
thanks to all for sharing all the great ideas. I am thinking about doing it myself (dual cpu 5,1).
so iam curious about how you guys experience your mods for cooling Northbridge over the years - be it the added aluminium foil tape on cpuA heatsink mod, add a plastic card that redirects the air from cpuB heatsink toward the nb or an extra fan for the nb.

@drolson I wonder if cpuA (wich already runs a bit hotter than cpuB) will suffer by the Steel/aluminium foil tape mod and getting less cooling?
 
Last edited:
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.