MP 1,1-5,1 El Capitan hot CPUs issue

[Sample]

Hi to all, I run Mojave on an SSD in one of the SATA II bays and recently also installed El Capitan (I frequently need it to use some legacy software) on a partitioned 6G SSD installed on a OWC Accelsior S PCIe card on PCI slot 4 (the other SSD partition runs Sierra).

___________________________________________________________________________________________

1) El Capitan shows “kernel_task” in the activity monitor using about 1,5 Gb of RAM and a little CPU.
I’ve already tried the popular trick of removing the kernel extensions from the following path:

System/Library/Extensions/IOPlatformPluginFamily.kext/Contents/PlugIns/ACPI_SMC_PlatformPlugin.kext/Contents/Resources

...but kernel_task keeps popping up all the same so I put the kernel extension back to where it belonged.
___________________________________________________________________________________________

2) Concerning temperatures, Mojave, Sierra and Mountain Lion all run ok while only El Capitan causes higher temperatures on the CPUs and Northbridge:

Mojave:

CPU A Diode: 43
CPU A Heatsink: 38
CPU A core from PCECI: 47

CPU B Diode: 37
CPU B Heatsink: 32
CPU B core from PCECI: 38

Northbridge Diode: 69
Northbridge Heatsink: 58

BOOSTA (fan): 1113
BOOSTB (fan): 1113


__________________________________
El Capitan:

CPU A Diode: 81
CPU A Heatsink: 58
CPU A core from PCECI: 91

CPU B Diode: 62
CPU B Heatsink: 48
CPU B core from PCECI: 69

Northbridge Diode: 77
Northbridge Heatsink: 65

BOOSTA (fan): 1247
BOOSTB (fan): 1247




Has anyone fixed this El Capitan hot CPUs issue?

Thanks for your help guys
S

 

[h9826790]

It looks like poor CPU heatsink contact / thermal paste application, but nothing about El Capitan.

It's just you coincidentally see this in El Capitan.

If you run CPU stress test in other macOS, most likely you can see even worse temperatures.

 

[Sample]

It looks like poor CPU heatsink contact / thermal paste application, but nothing about El Capitan.

It's just you coincidentally see this in El Capitan.

If you run CPU stress test in other macOS, most likely you can see even worse temperatures.

Thanks! Is there an app I can use to make this stress test safely?
I gather that if I try this on Mojave and the temps don't stay down it means it's a hardware issue, while if the temps stay down it means it's an El Capitan issue right?
Sorry never tried anything like this and I need a little advice....

I also need to point out that I just tried loading another installation of El Capitan running from an HDD in one of the SATA II bays (not from an SSD like the one with this heating up issue) and there was no issue there, low temps and no kernel_task...
Could the problem also be El Capitan running on an SSD?

thanks for your help..

 

[h9826790]

Thanks! Is there an app I can use to make this stress test safely?
I gather that if I try this on Mojave and the temps don't stay down it means it's a hardware issue, while if the temps stay down it means it's an El Capitan issue right?
Sorry never tried anything like this and I need a little advice....

I also need to point out that I just tried loading another installation of El Capitan running from an HDD in one of the SATA II bays (not from an SSD like the one with this heating up issue) and there was no issue there, low temps and no kernel_task...
Could the problem also be El Capitan running on an SSD?

thanks for your help..

Prime95

You can start with only 1 thread.

If the cooling is bad, you can stop the stress test straight away.

If the cooling looks OK, then you can increase the thread count to increase the stress.

 

[amstel78]

Thanks! Is there an app I can use to make this stress test safely?
I gather that if I try this on Mojave and the temps don't stay down it means it's a hardware issue, while if the temps stay down it means it's an El Capitan issue right?
Sorry never tried anything like this and I need a little advice....

I also need to point out that I just tried loading another installation of El Capitan running from an HDD in one of the SATA II bays (not from an SSD like the one with this heating up issue) and there was no issue there, low temps and no kernel_task...
Could the problem also be El Capitan running on an SSD?

thanks for your help..

Interesting that the temps stay low when booting El Cap from an HDD. Do you have trimforce turned on in El Cap? If so, turn it off and see if it makes a difference.

 

[Sample]

Interesting that the temps stay low when booting El Cap from an HDD. Do you have trimforce turned on in El Cap? If so, turn it off and see if it makes a difference.

thanks amstel78, yes isn't that strange??
I checked and I had never even turned on trimforce, learned of its existence from you, it is off.

@ h9826790 When I tried loading other OSs I put a hand on the CPU A heatsink and it was lukewarm, confirming Macs fan Control cool readings, then I restarted El Capitan, waited for a while and the CPU A heatsink was much hotter at the touch, confirming again Macs fan Control hot readings...

I wonder what could be the culprit??

 

[KeesMacPro]

Maybe it's a combination of things like e.g:

- bad heattransfer as stated by @h9826790 between the CPU and heatsinks ( difference Temp CPU diode and heatsink is very big indeed)
- a fresh install on the SSD (El Capitan) keeping the OS occupied indexing all disks ( you could test with only the El Capitan SSD installed ) this should be gone after a while depending on the amount of data
- the SSD is almost 100% full (or if i understood correctly the partition is too small) leading to memory paging between RAM and SSD


just some thoughts....

BTW: IMHO Trimforce should be enabled , this improves the SSD performance and keeps it clean .

 

[h9826790]

@ h9826790 When I tried loading other OSs I put a hand on the CPU A heatsink and it was lukewarm, confirming Macs fan Control cool readings, then I restarted El Capitan, waited for a while and the CPU A heatsink was much hotter at the touch, confirming again Macs fan Control hot readings...

I wonder what could be the culprit??

You need to check activity monitor to confirm that temperature different exist under the same CPU loading.

Only boot to different OS doesn't mean much.

As KeesMacPro pointed out, may be there are much more background task in El Capitan, which stress the CPU more, then of course it runs hotter.

Anyway, your biggest problem now isn't El Capitan or TRIM, but the large temperature between CPU diode and heatsink. You have to confirm if that's the same in all OS (e.g. by using Prime95 to stress the CPU).

If yes, you have to fix that physically ASAP.

 

[Sample]

Thanks KeesMacPro, I have an update, the SSD (via PCIe) has two 500 Gb partitions, one has El Capitan, the other had Sierra. I erased Sierra and installed a fresh installation of El Capitan there too using the exact same installer and this new instance of El Cap has no issue...

At this stage I gather it's probably some malware I got by installing some software on the first installation of El Cap, that I still haven’t installed on this fresh installation of El Cap, could that be?

Could someone recommend some good (free or paid) malware removal/protector for Mac? Never used and AV on Mac in my life..

(I tried booting just with the PCIe SSD unplugging the 4 SATA II internal drives but the issues in El Cap (1) was still there..)

 

[KeesMacPro]

@Sample
Good to hear !
I cant tell you if its malware related, but e.g. Malwarebytes is cool and for free


Anyway, I agree with h9826790 i think it's wise to check the CPU heatsinks and repaste them .
The difference in Temp between CPU diode and heatsink should be just a few degrees at idle and ~ 15 degrees max at (very) high loads.
Would be a shame if you cook them

 

[Sample]

Thanks again KeesMacPro and h9826790, I actually just put in new CPUs and repasted them one month ago:

- Mojave (43-38 C°)
- Mountain Lion (45-41 C°)
- Sierra (49-43 C°)
- and this new El Capitan (43-38 C°)

at idle these all show a difference between Diode and Heatsink (of CPU A) between 6 and 4 degrees, do you think that’s good enough? CPU B is lower still.
Only the “older” El Capitan, the one we suspect of malware, shows those crazy numbers (81-58 C°) and kernel_task eating up 1,4 Gb of RAM... And at the touch only this one feels pretty hot, while the others are just lukewarm...

 

[h9826790]

- Mojave (43-38 C°)
- Mountain Lion (45-41 C°)
- Sierra (49-43 C°)
- and this new El Capitan (43-38 C°)

at idle these all show a difference between Diode and Heatsink (of CPU A) between 6 and 4 degrees, do you think that’s good enough? CPU B is lower still.

That doesn't mean much, you have to stress the CPU to test the cooling system.

I actually just put in new CPUs and repasted them one month ago:

This is exactly where the problem is. Very high chance the heatsink don't have properly contact to the CPU now.

If you want to face the problem, Simply run Prime95, then you will know the answer.

As you can see from my test below. Even the CPU is under stress test, the temperature difference between my CPU diode and heatsink is only 13°. And all the fans still stay at quite low RPM.

This is a sign that the cooling is very effective. In your case, the cooling doesn't looks effective to me at all. You better check it.

 

[KeesMacPro]

Here are some screenshots of my dual X5690 with Mojave installed , just to compare if you like.
At idle de Temps differ a few degrees too which is normal, and the other screenshot is after running Cinebench CPU test.
During the test there are peaks of about ~75 Celsius (CPU A), but immediately after the test the CPUs go back to somewhere close to the heatsink Temp.
CPU B running cooler than CPU A is normal ; CPU B kicks in when CPU A is working a lot.

 

[amstel78]

That doesn't mean much, you have to stress the CPU to test the cooling system.


This is exactly where the problem is. Very high chance the heatsink don't have properly contact to the CPU now.

If you want to face the problem, Simply run Prime95, then you will know the answer.

As you can see from my test below. Even the CPU is under stress test, the temperature difference between my CPU diode and heatsink is only 13°. And all the fans still stay at quite low RPM.
View attachment 1730190
This is a sign that the cooling is very effective. In your case, the cooling doesn't looks effective to me at all. You better check it.

What application are you using to monitor fan speeds and voltages? Reminds me a little of HW Monitor made by Marcel Bresink although that application has long been deprecated.

 

[h9826790]

What application are you using to monitor fan speeds and voltages? Reminds me a little of HW Monitor made by Marcel Bresink although that application has long been deprecated.

iStat menu

 

[amstel78]

iStat menu

Thanks! Been looking for something to replace HW Monitor. Will check that out.

 

[Sample]

That doesn't mean much, you have to stress the CPU to test the cooling system.


This is exactly where the problem is. Very high chance the heatsink don't have properly contact to the CPU now.

If you want to face the problem, Simply run Prime95, then you will know the answer.

As you can see from my test below. Even the CPU is under stress test, the temperature difference between my CPU diode and heatsink is only 13°. And all the fans still stay at quite low RPM.
View attachment 1730190
This is a sign that the cooling is very effective. In your case, the cooling doesn't looks effective to me at all. You better check it.

Thanks h9826790, I finally grasp the seriousness of the situation, I'll do the test asap and post the results, I remember that when I installed the new CPUs I turned the bolts just finger tight (using thumb and index) and stopping as soon as I felt a sudden increase in resistance, around 7-8 turns for the two delidded X5675s.

Probably not tight enough I guess... do you think I risk damaging the CPU by doing the test as it is? Or should I give another turn to both heatsinks before trying?

Do you guys have a preferred number of turns, or some reference for the amount of torque needed for delidded CPUs on a dual 4,1 (flashed to 5,1), black square plastic spacer included?

Thanks for your help, really appreciate it

 

[h9826790]

Thanks h9826790, I finally grasp the seriousness of the situation, I'll do the test asap and post the results, I remember that when I installed the new CPUs I turned the bolts just finger tight (using thumb and index) and stopping as soon as I felt a sudden increase in resistance, around 7-8 turns for the two delidded X5675s.

Probably not tight enough I guess... do you think I risk damaging the CPU by doing the test as it is? Or should I give another turn to both heatsinks before trying?

Do you guys have a preferred number of turns, or some reference for the amount of torque needed for delidded CPUs on a dual 4,1 (flashed to 5,1), black square plastic spacer included?

Thanks for your help, really appreciate it

That’s why I suggest you run Prime95 with 1 thread first, then slowly increase the thread count, but not full 24 threads on the first go.

For dlidded CPU, usually the plastic retainer cause the problem (if you installed that).

For self delidded CPU, there may be some silicon seal residual, which lift the plastic retainer a bit, makes it can contact the CPU die properly.

For screws, it should be finger tight + 1/4 turn, not just finger tight. If you missed that 1/4 turn, you may do it now. Of course, if feel very high resistance, don’t use hard force.

 

[KeesMacPro]

Do you guys have a preferred number of turns, or some reference for the amount of torque needed for delidded CPUs on a dual 4,1 (flashed to 5,1), black square plastic spacer included?

Here's a link for the Service manual, always good for troubleshooting etc.etc.
There's a chapter about repasting and so on.

https://tim.id.au/laptops/apple/macpro/macpro_early2009.pdf

In case you decide to repaste the CPUs :
Personally I never count the turns because with a different CPU, fresh thermal paste (probably another brand too) and never the exact same amount of paste as before, IMHO it doesnt make much sense to count turns (we're talking about fractions of millimeters here).
The most important thing is to keep the heatsink perfectly levelled while mounting the heatsink and turn the screws crosswise , to avoid mounting the heatsink not 100% parallel to the pcb.
So e.g. turn screw 1 1turn, screw 3 1 turn, screw2 1 turn, screw4 1 turn and so on while keeping an eye on the angle of the heatsink .
After lets say 5 or 6 turns, turn every screw not 1 turn but 1/2 turn to avoid unevenness getting closer to the correct position of the heatsink.
As soon as you get at the point that you feel resistance screwing, give all screws about 1/4 turn max .
I think the "feel" of the torque is a much better indication of the right pressure , then a turning count...

Hope to be of some help...

 

[amstel78]

@Sample The below is what you should see when everything is tightened correctly and properly pasted. I'm running Prime95 using the small FFT test. All 12 physical cores are being used. Neither CPU diode should report temps above 90C. Mine loafs around 81/73C respectively (and this with the higher TDP X5690 CPUs).

Also, there shouldn't be more than a 15-20C difference between diode and heatsink temps. I could probably get the differential a *tad* lower with some new AS5. Alas, that's a project for another day.

 

[Sample]

Prime95

You can start with only 1 thread.
View attachment 1729686

If the cooling is bad, you can stop the stress test straight away.

If the cooling looks OK, then you can increase the thread count to increase the stress.

Can I ask you where you found this version of Prime95? (I'll use Mojave to test)
The only version I found is the plain text version (which I'm ok with of course) but I checked all the menus and I couldn't find the option to choose a "small FFT" test, as soon as I run the app it automatically starts a test but I stop it immediately since I can't understand which kind of test it's performing...

 

[h9826790]

Can I ask you where you found this version of Prime95? (I'll use Mojave to test)
The only version I found is the plain text version (which I'm ok with of course) but I checked all the menus and I couldn't find the option to choose a "small FFT" test, as soon as I run the app it automatically starts a test but I stop it immediately since I can't understand which kind of test it's performing...

http://www.mersenne.org/ftp_root/gimps/p95v298b7.MacOSX.tar.gz

 

[Sample]

http://www.mersenne.org/ftp_root/gimps/p95v298b7.MacOSX.tar.gz

Sorry to bother you again h9826790 but that is the same text only version I already downloaded, the only options I have under the “test” menu are “stop” and “continue”, I let it run (with its automatic default test? not sure if it's the small FTTs one) for half a minute before stopping it and yes there is a big difference in temp between the CPU A Diode and its Heatsink (77-49), here’s a screenshot:



But I don’t seem to have any option to select “Small FTTs” like in the version you have posted previously which is this:



How can I access this page?
Thanks for your patience

 

[amstel78]

Click on Options after stopping all helper threads. Torture test should be one of the choices. Click on it and you'll see the menu to select type of test, etc.

 

[Sample]

Click on Options after stopping all helper threads. Torture test should be one of the choices. Click on it and you'll see the menu to select type of test, etc.

yesss, thanks amstel78, here are the first results with small FFT, does it stop automatically after a while? After half a minute I decided to stop since I didn't want to risk: