Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
The thermal pad replaced paste... I saw the before and after pictures the tech took. While I have only one reading prior, it does show a big drop on the heatsink with a slight rise on the diode. Certainly with all the questions I've been asking and the replies I have been getting, it sure seems the NB qualifies for the "needing thermal paste" category! It was the tech's decision to use a pad, based on my telling him I had a 166F(74ºC) reading in the NB.

Once I get the backup tray I ordered, I will open up my current one. I should be able to tell if there is any issues securing the NB heatsink, if not, I will replace the pad with paste.
 
My backup tray was damaged to unusability in shipping, the vendor immediately sent a replacement (I sent him pictures to bolster any claims he has with the shipper). Meanwhile I keep a very close eye on ALL my temps. Yes I am reluctant/nervous to take my tray apart. I DO see what I think may be the cause of many of the broken clips, you have to remove the whole PCB from the metal chassis and squeeze the pins on the underside to release them to remove the NB heatsink. I can SEE that in my unit, the plastic rivets are through the PCB and seemingly holding.

What DOES seem kinda telling to me is that from a cold start, as soon as I get my desktop, that NB chip is running 71º (heatsink at 34º). Somehow I find it hard to believe it gets that hot that fast. Under heavy load, yes with fans spinning, the heatsink goes up to 36.8º with the NB diode at 76.5º. With no fan control, the diode can run 86.8º with the heatsink at 46.4º. CPU runs at 46º.

I still can't help thinking the diode is simply not reading the correct temp of the NB chip. Obviously it's varying, but it's like it starts at very high "reading" to begin with. I mean, how can one possibly have a 30º differential (assuming 10º is normal)? AND, while a lot subtler, if it REALLY was running tha hot, how come everything else in there isn't running a lot hotter than normal?
 
What DOES seem kinda telling to me is that from a cold start, as soon as I get my desktop, that NB chip is running 71º (heatsink at 34º). Somehow I find it hard to believe it gets that hot that fast. Under heavy load, yes with fans spinning, the heatsink goes up to 36.8º with the NB diode at 76.5º. With no fan control, the diode can run 86.8º with the heatsink at 46.4º. CPU runs at 46º.

That is a pretty much textbook example of bad thermal contact between the NB and the heatsink. It can and will get hot very quick if it's not making decent contact. I'm not saying it can't be a duff sensor, but if it is that would be the first one I've ever heard of on a Mac Pro, where as if it's just bad thermal contact that'd probably be example number twenty something that' I've personally heard of...

if it REALLY was running tha hot, how come everything else in there isn't running a lot hotter than normal?

There's no much else in very close proximity to the NB, and certainly nothing else with a temp sensor. The nearest thing with a temp sensor to the NB is the CPU, and that has a big enough heatsink (and forced airflow) that even when the NB is burning up it wouldn't appreciably affect the CPU temp readings. The Ram isn't far away really but an overheating NB is a fairly localised affair, and the NB diode is a very localised reading form within the die itself, and the die is physically quite small and even if it were heating up the air and board around it there's enough opportunity for that heat to dissipate without really causing noticeably elevated readings in other components.

As several of us have already said though, an IOH tdiode temp in the mid to high 70s is not abnormal for a dual CPU Mac Pro, and even into the 80s under load is not unusual. What is unusual is the big difference between the diode and the heatsink, normally a ~80deg C Tdiode temp would have the heatsink sitting in the mid-high 60s, which is why I suspect poor thermal contact in this case.
 
Last edited:
Interesting... btw, it's a single processor machine, not dual. I'm kind of discounting the fact it has a new thermal pad instead of paste (from everything I have learned and been told, that's a few degrees at best)... so the only thing left is the heatsink is not really pressing down at all, it's just maybe laying there? Wouldn't that also indicate the spring loaded clips are not spring loaded (although, from what I have read, they are there to accommodate metal expansion/retraction and heat and at that, maybe a millimeter I'd guess)? If I DO see that there is no real play in the heatsink and that trying to rock it from front to back shows no wiggle, wouldn't that indicate good thermal contact? I DID play a bit with the damaged tray, it's heatsink did not wiggle at all going front to back (perpendicular to its length) BUT I could twist it a tiny bit when looking down from overhead... not so sure THAT can be called bad thermal contact. Look, I am NOT trying to be argumentative, I AM trying to learn stuff. Most all the rather extensive threads about overheating seem to boil down to broken clips and what to do about them... and I think some ham-handedness is in play as well. Something I am very specifically guarding against!
 
You’ve had countless people tell you what’s wrong but you continue to believe your notions instead. I’m not sure why you’re here xD
 
  • Like
Reactions: h9826790
Hi ,
I'm new here and would like to share my solution for the iohd issue.
After reading a lot on this forum and searching the internet, I found a guy on YouTube who solved it with nylon nuts and bolts.
I suppose the original plastic pins are made this way because some poor guy had to assemble 100 of MP daily.In other words: its fast to assemble.
A few weeks ago I upgraded my CPU's and did the IOH as well, with nylon nuts, bolts (m3x16)and washers.
To avoid too much pressure ,before I removed the heatsink, I measured the distance between the head of the pin and the heatsink, and adjusted the new screws at the same hight.
I used Arctic Silver 5 and noticed that the difference in Temp between the heatsink and the diode stays between 9-10 degrees Celsius.
I attached some Fotos and a screenshot of the Temps based on a regular load (audio recording, editing).
Hoping to be of some help to someone.
cheers
 

Attachments

  • IMG_2299.JPG
    IMG_2299.JPG
    370.5 KB · Views: 148
  • IMG_2300.JPG
    IMG_2300.JPG
    472.1 KB · Views: 140
  • IMG_2302.JPG
    IMG_2302.JPG
    544.7 KB · Views: 155
  • Screenshot 10.png
    Screenshot 10.png
    314.4 KB · Views: 143
Hi ,
I'm new here and would like to share my solution for the iohd issue.
After reading a lot on this forum and searching the internet, I found a guy on YouTube who solved it with nylon nuts and bolts.
I suppose the original plastic pins are made this way because some poor guy had to assemble 100 of MP daily.In other words: its fast to assemble.
A few weeks ago I upgraded my CPU's and did the IOH as well, with nylon nuts, bolts (m3x16)and washers.
To avoid too much pressure ,before I removed the heatsink, I measured the distance between the head of the pin and the heatsink, and adjusted the new screws at the same hight.
I used Arctic Silver 5 and noticed that the difference in Temp between the heatsink and the diode stays between 9-10 degrees Celsius.
I attached some Fotos and a screenshot of the Temps based on a regular load (audio recording, editing).
Hoping to be of some help to someone.
cheers
Don't ever use constant speed (not unless it is at maximum) in MFC if you don't want to fry something. Always base it on the sensor temp or leave the apple default. Your exhaust and intake are set for constant speed.
 
  • Like
Reactions: h9826790
Don't ever use constant speed (not unless it is at maximum) in MFC if you don't want to fry something. Always base it on the sensor temp or leave the apple default. Your exhaust and intake are set for constant speed.
Thx Startergo!
What would you recommend ? default or based on which Sensor/Part?
 
I bought them in an electronics store (Netherlands), but you might get them easily in a supplier of parts for radiographic toys as well.

By the way, I just setup the fans and it looks all very cool:)
One more question I have:
I use a GTX780TI and did the Pixlas mod a while ago, and had the fan for the PS depending on PSMI supply 2 because it runs warmer then supply 1.
It stays normally between 35 and 40 degrees Celsius, should I leave it like you did it?
 
if apple engineers used thermal paste it is for a reason.
replacing the paste with some good goop such as artic or any overclocker brand after carefully cleaning the die and laping the radiator thermal contact surface on a piece of 2000 grid water sanding paper , and applying the right amount of paste without spreading it, will get you a max delta of 10*C between diode and radiator sensor.
if you have more than 10*C of delta, your thermal conductivity is bad and the Northbridge is not cooled properly.
you want to use the rice grain technique.
a lot of guy buy cheap thermalpaste and do their best to spread it evenly with a credit card or a plastic spunger and get a beautiful air bubble trapped in the paste.
the pressure of the spring is not enough to evenly spread the thermal paste and surely not strong enough to expulse a air bubble trapped in thick cheap paste.
once plastic clip reattached, you want to very gently “massage the chip” with the radiator until you can see paste getting out on the side with a flashlight.
the thiner the better.
My machine now usually have 5/7 *C of delta and never raise over 10*C
my fans are set up to ramp up full speed to never alow the pch go over 70*C, but on very intensive works it will go up to 75* for 1 or 2 second before getting back to 70*C.
as soon as workoad is less the fan wil ramp down within 10seconds...
please remember that those machine were developed in an eara were the fastest storage on earth was 700mb/s fiber sas 3 array... and the fastest gpu had 240 cores.
now we are hitting those puppys with 12x3,46 128gb of 1333ddr3 , 5Gb\s pcie flash array and 3 gpu with 2800 core each....

so clearly the Ioh is the busy part of the machine and I’m not sure that back in the day engineers even planed that so many of those machines would be hit this hard with workload...
 
  • Like
Reactions: Eschers
if apple engineers used thermal paste it is for a reason.
replacing the paste with some good goop such as artic or any overclocker brand after carefully cleaning the die and laping the radiator thermal contact surface on a piece of 2000 grid water sanding paper , and applying the right amount of paste without spreading it, will get you a max delta of 10*C between diode and radiator sensor.
if you have more than 10*C of delta, your thermal conductivity is bad and the Northbridge is not cooled properly.
you want to use the rice grain technique.
a lot of guy buy cheap thermalpaste and do their best to spread it evenly with a credit card or a plastic spunger and get a beautiful air bubble trapped in the paste.
the pressure of the spring is not enough to evenly spread the thermal paste and surely not strong enough to expulse a air bubble trapped in thick cheap paste.
once plastic clip reattached, you want to very gently “massage the chip” with the radiator until you can see paste getting out on the side with a flashlight.
the thiner the better.
My machine now usually have 5/7 *C of delta and never raise over 10*C
my fans are set up to ramp up full speed to never alow the pch go over 70*C, but on very intensive works it will go up to 75* for 1 or 2 second before getting back to 70*C.
as soon as workoad is less the fan wil ramp down within 10seconds...
please remember that those machine were developed in an eara were the fastest storage on earth was 700mb/s fiber sas 3 array... and the fastest gpu had 240 cores.
now we are hitting those puppys with 12x3,46 128gb of 1333ddr3 , 5Gb\s pcie flash array and 3 gpu with 2800 core each....

so clearly the Ioh is the busy part of the machine and I’m not sure that back in the day engineers even planed that so many of those machines would be hit this hard with workload...

A delta of 5/7 is incredible !
I did view lots of video's watching guys doing it the wrong way...
So I read the tutorial of Arctic Silver, but didn't "massage the chip" as you explained...
I'll keep an eye on the Temps and if it rises , I'll do the trick again , but with the extra's you explained!
Thx gentlemen, I've found so much knowledge here!
cheers
 
I use a GTX780TI and did the Pixlas mod a while ago, and had the fan for the PS depending on PSMI supply 2 because it runs warmer then supply 1.
It stays normally between 35 and 40 degrees Celsius, should I leave it like you did it?
1573306101894.png
 
just one last thing : dont use liquid metal on IOH radiator: the radiator will weld itself to the chip... dont ask me how i know it...

i use artic silver 5 on everything it is easy to find.
 
thx again!
This may be informative for you regarding a Northbridge temps.

My current system temps according to iStat Menus.
Now that Summer is over I have reverted all my fans to System control & shut down macs Fan Control until 2020 Summer.

Nov 10-2019 iStats menus.jpg
 
Last edited:
This may be informative for you regarding a Northbridge temps.

My curent system temps according to iStat Menus.
Now that Summer is over I have reverted all my fans to System control & shut down macs Fan Control until 2020 Summer.

View attachment 876127
Thx for the detailed info!
Actually I cleaned my MP as well a few days ago , and noticed the amount of dust accumulated especially in the Cpu part. I like the idea of keeping the optical drive bay slots open, might do that too during the summer.
 
As an Amazon Associate, MacRumors earns a commission from qualifying purchases made through links in this post.
  • Like
Reactions: KeesMacPro
Found this (if the 15mm doesn't engage all the threads on the nut, one can cut down the 20mm one):

Nice!
The ones i used are 16mm and are more then long enough , so the 15mm could be about the exact length necessary.
You might still need the (nylon) washers, especially under the head of the screw to support the pressure of the springs , maybe the ones i put under the pcb are not necessary...
 
As an Amazon Associate, MacRumors earns a commission from qualifying purchases made through links in this post.
My Northbridge Diode was always 77c - steady as a rock, up to that within a minute of boot, pretty much never wavering.

Pulled my CPUs for a clean, and check of the Arctic MX4, and decided to check the Northbridge heatsink rivets - they were still unbroken. I was able to unfasten, and remove the heatsink relatively easily. The thermal compound was brittle and crumbly, so wiped it off with isopropyl, and regreased with mx2, using a thin film with credit card.

Northbridge diode is now 78c pretty dead steady. Heatsink is 65c, system ambient 35c. I have 4 internal spinners, 2 sata ssds a couple of usb disks, multiple usb peripherals, wifi ethernet & bluetooth active - pretty much every form of I/O is doing stuff. I put MX4 on the cpus because it was non-conductive. I'm not sure if it would be better to put Arctic Silver 5 on the northbridge.
 
Last edited:
My Northbridge Diode was always 77c - steady as a rock, up to that within a minute of boot, pretty much never wavering.

Pulled my CPUs for a clean, and check of the Arctic MX2, and decided to check the Northbridge heatsink rivets - they were still unbroken. I was able to unfasten, and remove the heatsink relatively easily. The thermal compound was brittle and crumbly, so wiped it off with isopropyl, and regreased with mx2, using a thin film with credit card.

Northbridge diode is now 78c pretty dead steady. Heatsink is 65c, system ambient 35c. I have 4 internal spinners, 2 sata ssds a couple of usb disks, multiple usb peripherals, wifi ethernet & bluetooth active - pretty much every form of I/O is doing stuff. I put MX2 on the cpus because it was non-conductive. I'm not sure if it would be better to put Arctic Silver 5 on the northbridge.
Arctic Silver helps but ultimately it is the ambient temperature and the airflow across the diode area from my tests. 35 is pretty high so MFC with appropriate settings do the job perfectly. I have to say my cMP sits in the den and if the door is closed and the AC setting is at 82 F and the ambient detects 32C overnight it gets pretty hot. With AC at 78F and ambient of 30 it looks good. There is a lot of heat accumulated in the case which has to get out of the case.
 
  • Like
Reactions: KeesMacPro
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.