Mac Pro 5.1 (2012) + W3690 Temperatures ? ? ?

[MikkelAD]

Hey everyone!

Recently bought a Mac Pro with a W3565. Thought I would get the most out of my system so upgraded to W3690. It was my first time applying thermal paste (watched youtube guides) and I think it went pretty well.

Thought I would be safe about the installation so installed "Macs Fan Control". I read online about CPU stress test and this is the result after 12 * "yes > /dev/null &" in terminal for 25 minutes.

Do you think it looks fine? ? ?

In terms of the fans I observed the following when "CPU A core from PCECI" got close to 100C:

EXHAUST 600 RPM --> 800 RPM, approximately
INTAKE 600 RPM --> 800 RPM, approximately
BOOSTA 800 RPM --> 1200 RPM, approximately

After 10 seconds the system sort of stabilized and the 3 fans slowed down a bit as seen on the picture.

Overall I think that the fans of the system are working properly since they also slow down to stock RPM as soon as the heavy CPU load is cancelled.

I have attached a picture of the idle temperatures to - If something looks weird please do tell !

Any point in messing with settings of the fans "custom" vs "auto" unless you have installed another GPU and experience problems with "PCI" & "PS" fans (read that one thread the other day) ? ? ?

Thanks in advance !

 

[h9826790]

Hey everyone!

Recently bought a Mac Pro with a W3565. Thought I would get the most out of my system so upgraded to W3690. It was my first time applying thermal paste (watched youtube guides) and I think it went pretty well.

Thought I would be safe about the installation so installed "Macs Fan Control". I read online about CPU stress test and this is the result after 12 * "yes > /dev/null &" in terminal for 25 minutes.

Do you think it looks fine? ? ?

In terms of the fans I observed the following when "CPU A core from PCECI" got close to 100C:

EXHAUST 600 RPM --> 800 RPM, approximately
INTAKE 600 RPM --> 800 RPM, approximately
BOOSTA 800 RPM --> 1200 RPM, approximately

After 10 seconds the system sort of stabilized and the 3 fans slowed down a bit as seen on the picture.

Overall I think that the fans of the system are working properly since they also slow down to stock RPM as soon as the heavy CPU load is cancelled.

I have attached a picture of the idle temperatures to - If something looks weird please do tell !

Any point in messing with settings of the fans "custom" vs "auto" unless you have installed another GPU and experience problems with "PCI" & "PS" fans (read that one thread the other day) ? ? ?

Thanks in advance !

Congratulations on your 1st CPU (with thermal paste) upgrade, all parameter looks normal to me.

83C CPU core temperature is very normal, that’s exactly what you can expect with the Apple native fan profile.

 

[MikkelAD]

Congratulations on your 1st CPU (with thermal paste) upgrade, all parameter looks normal to me.

83C CPU core temperature is very normal, that’s exactly what you can expect with the Apple native fan profile.

Thanks!

Well I thought "CPU A core from PCECI" was the important one?

 

[Darmok N Jalad]

CPU A core from PCECI is a hard one to explain, but that number can get quite high without issue. I have the exact same CPU, and the reading on mine gets that high under load. The W3960 just gets really warm being at the upper limits of this particular CPU architecture.

Personally, I set my BoostA fan to read the CPU Diode from a range of 64C to 84C. My Intake to the CPU A Heatsink, 50 to 67, and my Exhaust to CPU A Heatsink 51 to 69. This setup seems to keep the CPU diode to around 70C and the heatsink readings in the mid-50s under load without crazy amounts of fan noise. You can set yours as you wish!

As for the PCI and PS settings, your mileage may vary. Sometimes just running something that loads your GPU down will help fix the way those fans operate. Luxmark is an easy way to do that. I find that if you dual boot MacOS and Windows, these 2 fans seem to act up.

 

[MikkelAD]

CPU A core from PCECI is a hard one to explain, but that number can get quite high without issue. I have the exact same CPU, and the reading on mine gets that high under load. The W3960 just gets really warm being at the upper limits of this particular CPU architecture.
View attachment 751695
Personally, I set my BoostA fan to read the CPU Diode from a range of 64C to 84C. My Intake to the CPU A Heatsink, 50 to 67, and my Exhaust to CPU A Heatsink 51 to 69. This setup seems to keep the CPU diode to around 70C and the heatsink readings in the mid-50s under load without crazy amounts of fan noise. You can set yours as you wish!

As for the PCI and PS settings, your mileage may vary. Sometimes just running something that loads your GPU down will help fix the way those fans operate. Luxmark is an easy way to do that. I find that if you dual boot MacOS and Windows, these 2 fans seem to act up.

Yeah okay I see. Well thanks for the answers so far!

 

[Physalia]

Hello!

I'm new to macrumors, and have a 5,1 cMP mid-2012, just upgraded the CPU to a W3690, and like MikkelAD, I have my own set of related questions on the subject for you experts in the field. I have multiple questions, so please bear with me... all I know I've learned by reading and educating myself in the internet, and I'm not even close to be an expert by any means... thank you for your patience in advance.

I researched the internet including YouTube for guidance about the process of replacing the CPU, got a hold of the Technical Handbook for the 5,1, and acquired all the materials and tools needed to safely perform the job. The information is vast, the expert tutorials some were good, some bad and a few ugly. So after finishing my project and running into 'issues,' I searched and found more information that I wish I would have know prior. Things I believe are stable now with the cMP, but want to be reassured by you experts before my cheese grater MAC turns into a real plain cheese grater.


1) About the process of placing the thermal paste: Attached you will find an attempt to re-create the looks of how my thermal paste looked on the CPU (this is not thermal paste in the picture/the entire contact area on the CPU was covered,) I later found a smooth surface and perhaps less material would be optimal, but this was what I had at the time and went along with it. All the previous steps were accomplished to prepare the HeatSink and CPU for the material.

2) About the process of removing the HeatSink: I unscrewed each screw as indicated on the Technical Handbook and followed the instructions to replace them. The guidance was terrible on the technical handbook... they were using "finger tight" as a description for 'tightness' to be followed by 1/4 turn... well for me 'finger tight' was the end of the road, the screws stop moving so I left them there.

3) The computer booted up without any issues and a 2-startup chime PRAM reset performed. The 'About this Mac' system information properly displayed the new CPU. Macs Fan Control App started showing a marked increase in temperature for the CPU A core for PCECI with the fans on AUTO. I turned off the test when it reached 99C and the temperature quickly started dropping. I did see the red LED flashing on the motherboard, and reminded me of a YouTube video where this issue came up, the problem being the screws were too tight on the HeatSink. So I turned (according to the Technical Handbook screw removal order) each screw one full-turn back to release pressure. I don't know if this made a difference or not, maybe slightly, but the temperatures kept rising rapidly anyway with the fans on 'auto.'

4) Once tweaking with the Macs Fan Control all seemed very stable. I was surprised that the 'auto' mode never kicked in at high temperatures.

My concerns are:
1) Did I use to much thermal paste, incorrectly, is this a re-do for me or should I leave it alone?
2) Should I re-tighten the screws until they stop turning or leave them alone?
3) Can the 'auto' function be adjusted so that it kicks in, so I don't have to rely on the Macs Fan Control App?

Below are the images with their descriptions,

Thank you for your time and attention!

 

[h9826790]

Hello!

I'm new to macrumors, and have a 5,1 cMP mid-2012, just upgraded the CPU to a W3690, and like MikkelAD, I have my own set of related questions on the subject for you experts in the field. I have multiple questions, so please bear with me... all I know I've learned by reading and educating myself in the internet, and I'm not even close to be an expert by any means... thank you for your patience in advance.

I researched the internet including YouTube for guidance about the process of replacing the CPU, got a hold of the Technical Handbook for the 5,1, and acquired all the materials and tools needed to safely perform the job. The information is vast, the expert tutorials some were good, some bad and a few ugly. So after finishing my project and running into 'issues,' I searched and found more information that I wish I would have know prior. Things I believe are stable now with the cMP, but want to be reassured by you experts before my cheese grater MAC turns into a real plain cheese grater.


1) About the process of placing the thermal paste: Attached you will find an attempt to re-create the looks of how my thermal paste looked on the CPU (this is not thermal paste in the picture/the entire contact area on the CPU was covered,) I later found a smooth surface and perhaps less material would be optimal, but this was what I had at the time and went along with it. All the previous steps were accomplished to prepare the HeatSink and CPU for the material.

2) About the process of removing the HeatSink: I unscrewed each screw as indicated on the Technical Handbook and followed the instructions to replace them. The guidance was terrible on the technical handbook... they were using "finger tight" as a description for 'tightness' to be followed by 1/4 turn... well for me 'finger tight' was the end of the road, the screws stop moving so I left them there.

3) The computer booted up without any issues and a 2-startup chime PRAM reset performed. The 'About this Mac' system information properly displayed the new CPU. Macs Fan Control App started showing a marked increase in temperature for the CPU A core for PCECI with the fans on AUTO. I turned off the test when it reached 99C and the temperature quickly started dropping. I did see the red LED flashing on the motherboard, and reminded me of a YouTube video where this issue came up, the problem being the screws were too tight on the HeatSink. So I turned (according to the Technical Handbook screw removal order) each screw one full-turn back to release pressure. I don't know if this made a difference or not, maybe slightly, but the temperatures kept rising rapidly anyway with the fans on 'auto.'

4) Once tweaking with the Macs Fan Control all seemed very stable. I was surprised that the 'auto' mode never kicked in at high temperatures.

My concerns are:
1) Did I use to much thermal paste, incorrectly, is this a re-do for me or should I leave it alone?
2) Should I re-tighten the screws until they stop turning or leave them alone?
3) Can the 'auto' function be adjusted so that it kicks in, so I don't have to rely on the Macs Fan Control App?

Below are the images with their descriptions,

Thank you for your time and attention!


View attachment 807574

View attachment 807575

View attachment 807576

View attachment 807577

View attachment 807578

View attachment 807579

View attachment 807580

View attachment 807581

View attachment 807582

Your CPU installation / thermal paste application looks normal to me.

You problem is "you focus on the WRONG parameter". Forget about that PCECI reading, that's very meaningless on cMP. Focus on just the Diode temperature.

The native Apple fan profile (auto) will keep the CPU at around 85C when under full stress. So, your CPU warm up to 86C, which is absolutely normal.

And NEVER set the fan base on ambient, MacsFanControl will completely override the system protection. Base on wrong parameter can (and will) lead to component overheat. NEVER do that again.

Base on CPU diode temperature is the way to go, min 72, max 80 seems too aggressive to me. You CPU stabilise at 74C, nothing wrong about that, it's your own choice. Definitely safe. Just the fan noise is bit high (IMO).

And if you want the fans to share the load (more fans to spin up, but lower fan speed on a particular fan), you may make Intake and Exhaust fan also base on CPU Diode temperature. Therefore, when CPU warm up, all 3 fans will spin up, but not only rely on the Booster fan.

With all 3 fans spin up, you can safely set the max to 90 or higher. The CPU will stabilise at a little bit higher temperature, but shouldn't able to touch 80C in any case, and the fan RPM will be lowered (so less fan noise).

 

[MIKX]

It should be mentioned that 4.1 cMPs ( at least ) have blue "Locktite" on the heatsink bolts .. THAT makes judging the appropriate bolt tension almost impossible. I used sparsely applied Eucalyptus oil to 'emulsify' most of the Loktite.

mynext ( & last ) cMP will ne a 2012 model. . . . dead easy by comparison.

 

[Jayson A]

They say with Xeon processors, you should put a vertical line down the middle with the thermal paste. Is that what you did?

 

[Physalia]

Your CPU installation / thermal paste application looks normal to me.

You problem is "you focus on the WRONG parameter". Forget about that PCECI reading, that's very meaningless on cMP. Focus on just the Diode temperature.

The native Apple fan profile (auto) will keep the CPU at around 85C when under full stress. So, your CPU warm up to 86C, which is absolutely normal.

And NEVER set the fan base on ambient, MacsFanControl will completely override the system protection. Base on wrong parameter can (and will) lead to component overheat. NEVER do that again.

Base on CPU diode temperature is the way to go, min 72, max 80 seems too aggressive to me. You CPU stabilise at 74C, nothing wrong about that, it's your own choice. Definitely safe. Just the fan noise is bit high (IMO).

And if you want the fans to share the load (more fans to spin up, but lower fan speed on a particular fan), you may make Intake and Exhaust fan also base on CPU Diode temperature. Therefore, when CPU warm up, all 3 fans will spin up, but not only rely on the Booster fan.

With all 3 fans spin up, you can safely set the max to 90 or higher. The CPU will stabilise at a little bit higher temperature, but shouldn't able to touch 80C in any case, and the fan RPM will be lowered (so less fan noise).

Thanks for your reply! I knew I was going to learn a few things. I will follow your suggestions on the fan setup -and- to never ever using ‘ambient.’ Wow, do I feel lucky! You saved me from disaster, thanks again! I’m still amazed that the red led was lighting up on the motherboard while the fans were on auto and not ‘reving’ up! The culprit I suppose (correct me if I’m wrong) was the highest temperature and only temperature reaching the 99-100 mark: PCECI. Now I understand the relationship with the diode and they work hand in hand. Correct me if I wrong again please, but it seems that if a red led light is lighting up on the motherboard something must be inherently wrong; thus, PCECI worth being monitored since is hovering at much higher temperatures than the diode. How about focusing on the diode while keeping an eye on the PCECI?

Can the ‘auto’ parameters be adjusted within the OS X (Mojave)?

Thanks again for your time... looking forward to your advice!
[doublepost=1543603392][/doublepost]

They say with Xeon processors, you should put a vertical line down the middle with the thermal paste. Is that what you did?

Thanks Jayson A! Yes, the technical handbook does illustrate one line inside an oval. I struggle with this a bit because of some YouTube guides, including the ones that provided a hyperlink to the handbook, prepared the CPU by spreading the paste over the contact area. I’ve seen it used this way in other HeatSink installations, so I figure lets follow their lead. What would you recommend? Can I measure success on the temperature readouts or is this an inevitable re-do? Would like to hear your thoughts, thanks in advance!

 

[Darmok N Jalad]

Thermal compound should be used sparingly—a very thin coating is all that is needed. Too much and the heat sink won’t sit tightly to the CPU lid. For what it’s worth, when I had the cMP 4,1, all the hex bolts seemed to tighten down to a limit, so “finger tightening” wasn’t really the issue, I just stopped when the hex locked.

I don’t know if you said, but I’m assuming you have the single CPU 4,1, you used a lidded CPU, and you flashed to 5,1 before doing your upgrade?

 

[h9826790]

Thanks for your reply! I knew I was going to learn a few things. I will follow your suggestions on the fan setup -and- to never ever using ‘ambient.’ Wow, do I feel lucky! You saved me from disaster, thanks again! I’m still amazed that the red led was lighting up on the motherboard while the fans were on auto and not ‘reving’ up! The culprit I suppose (correct me if I’m wrong) was the highest temperature and only temperature reaching the 99-100 mark: PCECI. Now I understand the relationship with the diode and they work hand in hand. Correct me if I wrong again please, but it seems that if a red led light is lighting up on the motherboard something must be inherently wrong; thus, PCECI worth being monitored since is hovering at much higher temperatures than the diode. How about focusing on the diode while keeping an eye on the PCECI?

Can the ‘auto’ parameters be adjusted within the OS X (Mojave)?

Thanks again for your time... looking forward to your advice!
[doublepost=1543603392][/doublepost]
Thanks Jayson A! Yes, the technical handbook does illustrate one line inside an oval. I struggle with this a bit because of some YouTube guides, including the ones that provided a hyperlink to the handbook, prepared the CPU by spreading the paste over the contact area. I’ve seen it used this way in other HeatSink installations, so I figure lets follow their lead. What would you recommend? Can I measure success on the temperature readouts or is this an inevitable re-do? Would like to hear your thoughts, thanks in advance!

If you read that parameter carefully, that's CPU A core from PCECI. That suppose a count down temperature. And it was a count down temperature few years ago. However, I couldn't remember since which update, the parameter become erroneous, and very hard to determine its real meaning on cMP. So, my personal suggestion is simply ignore that.

And even base on that 96C PCECI reading, you should be still 4C away from ProcHot (the max temperature reported by the CPU itself. So, should be a correct number to use even Apple never sell W3690 cMP). So, that red light shouldn't be there.

No, that "auto" cannot be adjusted, you must use some fan control software to override the native setting.

 

[USmacs]

That's a lot of paste. Not that it should hurt. I've swapped well over 100 of these.

For anyone that's not certain and you have a decent tube of paste, I would recommend you put some paste on, put it back together, and then take it apart again. If it looks good, add a little pea to the center and put it back together

I only use Arctic Silver, start in the center and work my way out in a spiral, then add some to any areas that look sparse. The center of the CPU is where all the heat load is.

EDIT:
I felt I should add that the tendency is always to put too much paste on. It actually takes experience to know how little to put on. When I mention a tiny pea above i mean a little dot. Think of how much you cleaned off the mating surfaces when you took it apart.

 

[h9826790]

Thermal compound should be used sparingly—a very thin coating is all that is needed. Too much and the heat sink won’t sit tightly to the CPU lid. For what it’s worth, when I had the cMP 4,1, all the hex bolts seemed to tighten down to a limit, so “finger tightening” wasn’t really the issue, I just stopped when the hex locked.

I don’t know if you said, but I’m assuming you have the single CPU 4,1, you used a lidded CPU, and you flashed to 5,1 before doing your upgrade?

W3690 can only be used on single processor 5,1. So, must be lidded.

 

[Darmok N Jalad]

W3690 can only be used on single processor 5,1. So, must be lidded.

That’s how it should be, but I’d rather OP clairifies to eliminate the obvious. If he has a delidded CPU, he won’t be making good contact with the heatsink. From what I can see, it is already a 5,1, so that question is solved. I had the same CPU in my old cMP, and I never got above 75C under load.

And I may be mistaken, but I believe you can put a non-X-series CPU in a dual socket cMP, but you can’t run two CPUs, and it will make the fans go full bore.

 

[Physalia]

If you read that parameter carefully, that's CPU A core from PCECI. That suppose a count down temperature. And it was a count down temperature few years ago. However, I couldn't remember since which update, the parameter become erroneous, and very hard to determine its real meaning on cMP. So, my personal suggestion is simply ignore that.

And even base on that 96C PCECI reading, you should be still 4C away from ProcHot (the max temperature reported by the CPU itself. So, should be a correct number to use even Apple never sell W3690 cMP). So, that red light shouldn't be there.

No, that "auto" cannot be adjusted, you must use some fan control software to override the native setting.

Thanks again, amazing information!
[doublepost=1543616172][/doublepost]

That's a lot of paste. Not that it should hurt. I've swapped well over 100 of these.

For anyone that's not certain and you have a decent tube of paste, I would recommend you put some paste on, put it back together, and then take it apart again. If it looks good, add a little pea to the center and put it back together

I only use Arctic Silver, start in the center and work my way out in a spiral, then add some to any areas that look sparse. The center of the CPU is where all the heat load is.

Thanks USmacs! So in your opinion, with these temperatures I’m good to go? Thanks again!
[doublepost=1543616444][/doublepost]

Thermal compound should be used sparingly—a very thin coating is all that is needed. Too much and the heat sink won’t sit tightly to the CPU lid. For what it’s worth, when I had the cMP 4,1, all the hex bolts seemed to tighten down to a limit, so “finger tightening” wasn’t really the issue, I just stopped when the hex locked.

I don’t know if you said, but I’m assuming you have the single CPU 4,1, you used a lidded CPU, and you flashed to 5,1 before doing your upgrade?

Thanks Darmok, yes this is a 5,1 from the box, mid 2012, single processor. Do you think this must be taken apart again and re-done?

 

[h9826790]

That’s how it should be, but I’d rather OP clairifies to eliminate the obvious. If he has a delidded CPU, he won’t be making good contact with the heatsink. From what I can see, it is already a 5,1, so that question is solved. I had the same CPU in my old cMP, and I never got above 75C under load.

And I may be mistaken, but I believe you can put a non-X-series CPU in a dual socket cMP, but you can’t run two CPUs, and it will make the fans go full bore.

Of course we can install non "X" CPU in dual processor cMP (e.g. W5590), because that "X" has nothing to do about single or dual. The real problem is the 3xxx series CPU cannot be used in dual config, therefore, W3690 won't work in dual processor cMP.

And yes, technically we can install a single W3690 only on to the dual processor's cMP socket A. Fans will spin up to 100%. But his screen capture clearly shows not in that case.

If the ambient temperature is low enough, the CPU may never reach 75C even under full stress. But for normal room temperature (e.g. 20-25C), with native Apple fan profile. I am 100% sure a W3690's diode temperature can reach (and go beyond) 75C.

If you still have the cMP on hand, try run Prime95 (small FFT) for 10min please. I will be very surprised if your ambient temperature isn't that low, all fans at idle, but the W3690 can't even reach 75C.

 

[USmacs]

Thanks USmacs! So in your opinion, with these temperatures I’m good to go? Thanks again!
[doublepost=1543616444][/doublepost]


Yes, I think you're worrying too much. The built in fan speed control works perfectly. Don't bother taking it apart again now unless there's an issue.

 

[Darmok N Jalad]

Of course we can install non "X" CPU in dual processor cMP (e.g. W5590), because that "X" has nothing to do about single or dual. The real problem is the 3xxx series CPU cannot be used in dual config, therefore, W3690 won't work in dual processor cMP.

And yes, technically we can install a single W3690 only on to the dual processor's cMP socket A. Fans will spin up to 100%. But his screen capture clearly shows not in that case.

If the ambient temperature is low enough, the CPU may never reach 75C even under full stress. But for normal room temperature (e.g. 20-25C), with native Apple fan profile. I am 100% sure a W3690's diode temperature can reach (and go beyond) 75C.

If you still have the cMP on hand, try run Prime95 (small FFT) for 10min. I will be very surprised if your ambient temperature isn't that low, all fans at idle, but the W3690 can't even reach 75C.

Yeah, I was trying to go through everything on my iPhone at the time. Hard to look at the attachments!

 

[h9826790]

Yeah, I was trying to go through everything on my iPhone at the time. Hard to look at the attachments!

I often make the same mistake as well.

 

[Physalia]

Thanks USmacs! So in your opinion, with these temperatures I’m good to go? Thanks again!
[doublepost=1543616444][/doublepost]


Yes, I think you're worrying too much. The built in fan speed control works perfectly. Don't bother taking it apart again now unless there's an issue.

Outstanding! Thank you and everyone in this thread for your help. Your science is like a work of art and I have a lot of respect for your love and passion for these machines, and of course for being here to help. Thanks again!

Next project bootable PCIe adaptor for multiple M.2 SSD drives... just getting my feet wet on this... maybe I’ll meet you guys in one of the other threads.

 

[alphaod]

For those worried about the amount of thermal paste to put on... get one of those graphite thermal pads.

I know they're kind of expensive compared to paste, but you can't screw it up and you never have to worry about the paste drying up.

I've redone about six computers with them and they are the same if not better than before in terms of thermal performance. Of course liquid metal is better, but for the no-mess, it's hard not to like.

 

[theone29]

Hi all, sorry for hi-jacking the thread. I upgraded my cMP5,1 processors about a month ago. Lately it seems like the CPU B Core, Relative to Prochot and the X58 IOH Diode are running hotter than usual. Can you guys take a look and help on what I can customize with the Mac Fan control to low these temp? Thanks ahead for your help.

 

[h9826790]

Hi all, sorry for hi-jacking the thread. I upgraded my cMP5,1 processors about a month ago. Lately it seems like the CPU B Core, Relative to Prochot and the X58 IOH Diode are running hotter than usual. Can you guys take a look and help on what I can customize with the Mac Fan control to low these temp? Thanks ahead for your help.
View attachment 807867 View attachment 807868 View attachment 807869

All at normal range.

IOH diode in dual processor cMP is quite warm. Often reach mid 70 (considering your system ambient is 32C).

CPUs are far from overheat.

 

[Darmok N Jalad]

Hi all, sorry for hi-jacking the thread. I upgraded my cMP5,1 processors about a month ago. Lately it seems like the CPU B Core, Relative to Prochot and the X58 IOH Diode are running hotter than usual. Can you guys take a look and help on what I can customize with the Mac Fan control to low these temp? Thanks ahead for your help.
View attachment 807867 View attachment 807868 View attachment 807869

I found that sometimes when the health readings were going amiss, I’d just run a benchmark that loaded the system to 100%. After I stopped that bench, things settled back down to normal. Also, you might pull your CPU tray and chech the chipset HeatSink for dust. It might be hard to check, as it will be under one of the massive CPU HeatSinks. If you have an air duster, you may be able to get it cleaned out without removing anything.