MP 1,1-5,1 CPU A 15+ degrees hotter than CPU B

[iChan]

I know CPU A should be hotter than CPU B, but it used to only be around 7-8 degrees delta. Now it's routinely 15+ degrees.

Is this something I should worry about? and if it is, what should I do to fix it?

This only really started happening after I put new thermal compound on both the CPUs and northbridge chip. I'm comfortable enough with the northbridge Diode being at 69 cos I know that is designed to run hot anyway and I have a backup Dual CPU tray. Someday I might mod it to take a fan directly on the NB diode, but right now I'm just curious about the large temperature difference between CPU A & B, thanks!

Edit: I forgot to mention that I also use Macs Fan Control and I got tired of managing the fans and now just have auto everything.

Edit 2: This is a 4,1 flashed to 5,1.

 

[KeesMacPro]

CPU A runs hotter than B , this is normal.
The Delta CPU A heatsink -diode is too big (if at idle) so there's poor contact heatsink-die.
I'd very carefully try if you can turn the heatsink screws a tiny bit more.
If there's no effect I'd repaste them both CPU A and B.
NB delta is 15celsius too, which is not very good, It should be between ~7-12 after a fresh repaste.
You could leave the NB , but practically would be to repaste the NB too in case you repaste the CPUs.

 

[iChan]

Thanks, I didn't even notice the delta on the NB. It does suggest bad contact between the chip and sink. I'll do a full repaste this evening and post an update here. Perhaps I was a little too stingy with the paste. Will also tighten everything a little more than last time.

 

[iChan]

Just a note to say that last time I did it, I bent some of the cpu B pins on the tray because the CPU is delidded and doesn't really have anything holding it down so it came up with the paste on the underside of the heatsink and then dropped onto the pins.

The result was a read error on one or two of the RAM slots. ( I can't remember exactly)

I also removed a tiny strand of fibre which may have helped resolve the issue. In the end, everything is fine, but it's definitely a procedure that isn't difficult, but is kinda fraught with dangers if you're not careful and have a clean work area.

I bent a side fin of the NB heatsink too because that thing is ridiculously flimsy.

Somehow I feel the compound I used isn't the best. I got this one:

To be perfectly honest, I really think I tightened the the CPU A heatsink enough and it's the second repaste of the CPUs I've done on this Mac.

The first time I used the older version:

...And the temperatures were really fine. I just did them all again with the new paste because it was definitely time to do the NB.

I might just go back to the 2018 cos I still have enough left for one more repaste.

 

[KeesMacPro]

Repasting the delidded CPUs in a dual MP4,1 is a sensitive job with risks involved, as you describe .
If you think you've tightened the screws properly, go for a repaste.

The MX-4 is one of a lot of very good quality pastes on the market, I think applying it correctly makes more difference than the brand and type of thermal paste, based on my own experiences.

make sure there's no black sealer residu left on the upper side of the CPU , because this will in combination with the plastic spacers avoid good contact between heatsink and die.

 

[iChan]

So I removed the tray to check the screws, and yep, they were as tight as they would go. So I said "f this", I'm going to repaste.

I took a photo of what was there.

Seems OK, don't you think? The viscosity of the new formula seems be a lot runnier than the old one, but decent spread and decent even contact, I don't see anything wrong to be honest. Please let me know what you think.

So I cleaned it off and did reapplied, this time with the older MX-4:

(sorry for the dust). What do you think of that application? too much? Too little. Interested to hear what you have to say.

Incredibly, I see an instant major improvement upon reboot.

I'm quite shocked actually.

When I was buying the new formula of MX-4, I read many reviews of people saying the same thing that I'm experiencing. That it's runnier and doesn't work as effectively. Chalk this experience down to me becoming a believer. Maybe there are some really big differences in performance between these two pastes, or it could be down to user error, I'm not quite sure.

 

[KeesMacPro]

TBH didnt know about this "new" MX-4 , looks almost liquid.
Looks like the right amount and Delta looks good too (2Celsius).

I do see a lot of the black sealer on the CPU though, which makes the CPU with the spacer on that surface too high = bad contact.
Or did you install the CPUs without the plastic spacer on the die?

 

[iChan]

I also now regret not doing the NB, because I don't have enough left for repasting all three I think.


What's curious is why the 2019 paste seems to be working fine for CPU B. My guess would be because maybe I'm just not taxing the system at all - just generally doing browsing and watching YouTube. I'm planning on get a single CPU tray this week because I don't need all this power actually for my daily driver.

Here's the temps with the fans running full blast for a couple of minutes:

There's a 1deg delta on CPU A but a 5deg delta on CPU B. I think it's definitely the paste. Maybe it's fake. I got it from AliExpress.

 

[iChan]

TBH didnt know about this "new" MX-4 , looks almost liquid.
Looks like the right amount and Delta looks good too (2Celsius).

I do see a lot of the black sealer on the CPU though, which makes the CPU with the spacer on that surface too high = bad contact.
Or did you install the CPUs without the plastic spacer on the die?

There's a plastic spacer? I have none! Do I need them?

And I did check the rubber when I did the delidding and it sits much lower than the contact surface so shouldn't impede.

 

[iChan]

That delta on the NB after running fans for a good 5 minutes on Full Blast is insane. (I had literally just turned it back to auto as I took this screen)

Can't be having that. I'm going to do the NB and report back.

 

[KeesMacPro]

Just to see if i understood all correctly and my view:
-Temps should be read with all fans on auto
- dont buy paste at aliexpress , just dont
- the OEM CPUs are installed with a spacer like a plastic flat frame placed on top of the CPU , to protect the socket pins for too much pressure .
Some people dont use them, personally i always use them but the CPU surface should be perfectly clean ( no black sealer left) otherwise there's bad contact heatsink-die
-Delta NB is 15 degrees , which is not dramatic , but i would repaste it

I would buy a new paste ( not from China) and redo both CPUs and the NB...
MX-4 (original!) is fine
Arctic Silver 5 is fine
Thermal Grizzly Kryonaut is very good
Just to name a few...

 

[iChan]

@KeesMacPro

I reapplied everything (switching the A & B CPUs for the hell of it) and ran my full blast test again.

Seeing better temperatures all-round.

Still not great on the NB, but at least 12deg seems be within spec.

And it is idle:

I'm happy enough now.

I did take some pictures of the "2019" MX4 on the NB too...

It looks real tidy if I do say so myself! Strange that it performs so badly.

Interestingly, the 2019 compound cleans off like a liquid, whereas the "old" compound I had just applied on CPU A cleaned off in clumps.

It's like the difference between gel toothpaste and normal toothpaste.

 

[iChan]

Just to see if i understood all correctly and my view:
-Temps should be read with all fans on auto
- dont buy paste at aliexpress , just dont
- the OEM CPUs are installed with a spacer like a plastic flat frame placed on top of the CPU , to protect the socket pins for too much pressure .
Some people dont use them, personally i always use them but the CPU surface should be perfectly clean ( no black sealer left) otherwise there's bad contact heatsink-die
-Delta NB is 15 degrees , which is not dramatic , but i would repaste it

I would buy a new paste ( not from China) and redo both CPUs and the NB...
MX-4 (original!) is fine
Arctic Silver 5 is fine
Thermal Grizzly Kryonaut is very good
Just to name a few...

Thanks for all your input.

I tried looking for my older chips I removed when I upgraded, the spacers are in with them I think, but I don't know where they are right now

I will try Thermal Grizzly next time. I might switch to my other dual tray to see if I get an improvement with the NB once I get it, and remove the rubber.

Thanks again! You were very helpful

Right now, I'm happy with what I've got. Thanks again!

 

[KeesMacPro]

The second MFC screenshot looks excellent!

I'm convinced this Aliexpress tube actually is toothpaste.

EDIT: with the Temps like this, i'd leave it as it is , and go for a beer somewhere ... o no, go to the fridge

 

[Stex]

There's a plastic spacer? I have none! Do I need them?

And I did check the rubber when I did the delidding and it sits much lower than the contact surface so shouldn't impede.

take a look at my lovely icon... plastic spacer in action from cMP 4,1 manual! I tried to attach it here but it is too large...

 

[Sean__Wong]

@KeesMacPro

I reapplied everything (switching the A & B CPUs for the hell of it) and ran my full blast test again.

Seeing better temperatures all-round.

View attachment 1748696
Still not great on the NB, but at least 12deg seems be within spec.

And it is idle:
View attachment 1748703

@iChan Go Ghetto on your NB It is easy to get it down another 5ºC but the delta T will remain, IMHO the NB Heatsink should be copper based as the transfer rate is not high enough.

https://forums.macrumors.com/attachments/macpro5-1-nb-cpu-a-baffle-jpg.922245/

5,1 Mac Pro water cooled project - Working, with photos

@p11hlf Thanks for the post the watercool and your neat metal work looks good. Will you post your NB Heats sink shroud on thingiverse or another site ? I'd like to get some basic measurements on the Heatsink, I'm a bit stuck as I've only got one cMP 5,1 and I use it ! I'm keen to get a shroud...

forums.macrumors.com