Resolved Mac Pro 5,1 - W6390 upgrade and KPs

[mateuszd]

Hey all, this is my second thread. I wanted to split the two issues I have into separate threads to make it easier.

So, I've upgraded the original W3565 because I've not been able to use VMware Fusion with it to a W6390 that I got from eBay, this was advertised as a tested and fully working unit. I immediately installed VMware Fusion 11 at the latest update and started playing around with a VM and let my Mac go to sleep a few times and I woke it up to find that it crashed with a KP twice, both times, this seemed to happen at time of idleness of the system. In both occasions, the crashes mentioned the AppleIntelCPUPowerManagement.kext extension but I read online that this sometimes shows up even if another extension causes it.

This raised my suspicion that perhaps Fusion is responsible for it. The question is, should I try the NullCPUPowerManagement.kext? I believe that is for Hackintoshes, and considering that my CPU is supported, I think power management should work fine, right?

Whilst troubleshooting, I decided to stress test the CPU with the yes > /dev/null command, I found that the temperature reading reached way over 90 degrees in the PCECI reading, at some point it hit 95 degrees and then the Mac rebooted, this time, no KP message. I was kind of shocked because the Mac internal fans didn't engage so then I downloaded Mac Fans Control and manually set the BOOSTA fan at high speed, keeping the CPU cool and in fact, been able to run the stress test for a few hours without a crash.

This makes me think that hopefully those two issues are unrelated, one is hopefully caused by VMware and the other, by overheating. But the question is, why wouldn't the fans engage? Is Mac relying on a temperature reading from a different sensor (the diode or heatsink)? Those two were much lower at the time of the crash at around 80 as far as I remember so perhaps I've installed the thermal paste wrong?

Could you advise me what would be an acceptable difference between the PCECI and Diode and Heatsink readings so I can determine if I should apply the paste again?

Thanks guys!!!

 

[startergo]

Hey all, this is my second thread. I wanted to split the two issues I have into separate threads to make it easier.

So, I've upgraded the original W3565 because I've not been able to use VMware Fusion with it to a W6390 that I got from eBay, this was advertised as a tested and fully working unit. I immediately installed VMware Fusion 11 at the latest update and started playing around with a VM and let my Mac go to sleep a few times and I woke it up to find that it crashed with a KP twice, both times, this seemed to happen at time of idleness of the system. In both occasions, the crashes mentioned the AppleIntelCPUPowerManagement.kext extension but I read online that this sometimes shows up even if another extension causes it.

This raised my suspicion that perhaps Fusion is responsible for it. The question is, should I try the NullCPUPowerManagement.kext? I believe that is for Hackintoshes, and considering that my CPU is supported, I think power management should work fine, right?

Whilst troubleshooting, I decided to stress test the CPU with the yes > /dev/null command, I found that the temperature reading reached way over 90 degrees in the PCECI reading, at some point it hit 95 degrees and then the Mac rebooted, this time, no KP message. I was kind of shocked because the Mac internal fans didn't engage so then I downloaded Mac Fans Control and manually set the BOOSTA fan at high speed, keeping the CPU cool and in fact, been able to run the stress test for a few hours without a crash.

This makes me think that hopefully those two issues are unrelated, one is hopefully caused by VMware and the other, by overheating. But the question is, why wouldn't the fans engage? Is Mac relying on a temperature reading from a different sensor (the diode or heatsink)? Those two were much lower at the time of the crash at around 80 as far as I remember so perhaps I've installed the thermal paste wrong?

Could you advise me what would be an acceptable difference between the PCECI and Diode and Heatsink readings so I can determine if I should apply the paste again?

Thanks guys!!!

I am confused here. Are you trying to control the fans from within VMware? You can't reach the SMC from within VMware. You have to do it in the host.

 

[mateuszd]

I am confused here. Are you trying to control the fans from within VMware? You can't reach the SMC from within VMware. You have to do it in the host.

Hey, thanks for the reply. No, I am controlling the fans from the host...

I had two different types of crashes, a kernel panic when running VMware Fusion and a no error crash (reboot) when doing stress testing which I think was caused by overheating.

My issue is that, without any fan speed controlling software, the Mac just overheated and crashed, the fans didn't engage until I installed the software and manually created a rule to action based on the PCECI value.

 

[orph]

you may not have correctly put on the CPU heatsink or thermal paist when you replaced the CPU, can you get a temp monitor app and get idle temps and load temps (like luxmark CPU stress test) so we can see what your system is hitting.


temperature monitor has always been easy to use with the 'history window' to track temps over time
https://www.bresink.com/osx/TemperatureMonitor.html
i stat menus is used by a lot of people and has a demo to
https://bjango.com/mac/istatmenus/

any cpu test will do

is it a macpro4,1 or 5,1?
if 4,1 how did you replace the cpu? de-lid or spacers?

may also be a bad aplication of thermal paist, can happen.

 

[mateuszd]

you may not have correctly put on the CPU heatsink or thermal paist when you replaced the CPU, can you get a temp monitor app and get idle temps and load temps (like luxmark CPU stress test) so we can see what your system is hitting.


temperature monitor has always been easy to use with the 'history window' to track temps over time
https://www.bresink.com/osx/TemperatureMonitor.html
i stat menus is used by a lot of people and has a demo to
https://bjango.com/mac/istatmenus/

any cpu test will do

is it a macpro4,1 or 5,1?
if 4,1 how did you replace the cpu? de-lid or spacers?

may also be a bad aplication of thermal paist, can happen.

Thanks buddy, tried istatmenus, I moved to Macs Fan Control for now.

It's a Mac 5,1 so just swapped the CPU but I have a feeling that when I installed the cooling block, I screwed it in unevenly and that perhaps the thermal paste didn't spread evenly. Do you think I should reseat it?

Can you tell me is the "CPU A core from PCECI" the right sensor to monitor?

Idle temps:

CPU A Diode: 37
CPU A HeatSink: 35
CPU A core from PCECI: 38

 

[startergo]

Hey, thanks for the reply. No, I am controlling the fans from the host...

I had two different types of crashes, a kernel panic when running VMware Fusion and a no error crash (reboot) when doing stress testing which I think was caused by overheating.

My issue is that, without any fan speed controlling software, the Mac just overheated and crashed, the fans didn't engage until I installed the software and manually created a rule to action based on the PCECI value.

The high PCECI normally indicates bad connection between the heatsink and the CPU. Are you using thermal paste or something else? Make sure the heatsink is evenly tight on all 4 corners.

 

[mateuszd]

The high PCECI normally indicates bad connection between the heatsink and the CPU. Are you using thermal paste or something else? Make sure the heatsink is evenly tight on all 4 corners.

I'm using a thermal paste, but I think I installed it wrong, I will redo it today, I screwed the cooling block in unevenly and perhaps the thermal paste didn't spread evenly or I didn't put enough of it. Will feed back later about that...

Any ideas about the power management kernel panic? That seems unrelated to the temperature situation...

 

[orph]

fix one thing at a time, take it slow and easy.

it may just be as simple as un even heatsink is the only problem, that can give bad thermal contact, bad pin contact, ram not detected and so on

 

[mateuszd]

fix one thing at a time, take it slow and easy.

it may just be as simple as un even heatsink is the only problem, that can give bad thermal contact, bad pin contact, ram not detected and so on

I shall, I'm just testing things on everyday load and so far haven't experienced any KPs even when using VMware. I just swapped out the GPU to a GTX 680 so hopefully my graphics issues cease. I am running Macs Fan Control still but the fans haven't speed up a single time. I still bear in mind that I have to redo the thermal paste because the PCECI readings were too high. What is the maximum reading I should have under normal conditions (a correctly applied thermal paste)?

 

[h9826790]

I shall, I'm just testing things on everyday load and so far haven't experienced any KPs even when using VMware. I just swapped out the GPU to a GTX 680 so hopefully my graphics issues cease. I am running Macs Fan Control still but the fans haven't speed up a single time. I still bear in mind that I have to redo the thermal paste because the PCECI readings were too high. What is the maximum reading I should have under normal conditions (a correctly applied thermal paste)?

You better ignore that PCECI temperature. High reading usually means nothing.

Focus on CPU diode, and CPU heatsink.

Your idle temperature looks very good to me. I don’t think it temperature issue.

Unless your W3690 CPU diode temperature can touch 90C. Otherwise, can’t quite because of CPU overheat.

If VM seems causing issue. Just stop using it for a day or two (but keep the cMP running 24/7). If no more KP, then 99% is VM software issue.

 

[mateuszd]

You better ignore that PCECI temperature. High reading usually means nothing.

Focus on CPU diode, and CPU heatsink.

Your idle temperature looks very good to me. I don’t think it temperature issue.

Unless your W3690 CPU diode temperature can touch 90C. Otherwise, can’t quite because of CPU overheat.

If VM seems causing issue. Just stop using it for a day or two (but keep the cMP running 24/7). If no more KP, then 99% is VM software issue.

Thanks, I don't think the other readings reached that temperature but I will test again to confirm. The two issues seem unrelated, I'm focusing on the error-less crash which happened at high temperature, I think the diode was around 85, but I need to check again to confirm.

In terms of the other issue, it's weird because I've been using VMware and haven't had a KP since I raised the thread.

Computers...

 

[h9826790]

Thanks, I don't think the other readings reached that temperature but I will test again to confirm. The two issues seem unrelated, I'm focusing on the error-less crash which happened at high temperature, I think the diode was around 85, but I need to check again to confirm.

In terms of the other issue, it's weird because I've been using VMware and haven't had a KP since I raised the thread.

Computers...

85C is normal. That's how the cMP programmed to do so, until the W3690 diode temperature goes above 80C, the fans will stay at idle. And just spin up tiny bit to keep the CPU at around 85C. In my experience, it can never touch 90C with native fan setting (assume the heatsink is reasonably clean).

Anyway, it so much easier for us to determine if the temperature is normal if you can post a screen capture. Something like this will help us to have a better picture about if the cooling is adequate.

 

[mateuszd]

85C is normal. That's how the cMP programmed to do so, until the W3690 diode temperature goes above 80C, the fans will stay at idle. And just spin up tiny bit to keep the CPU at around 85C. In my experience, it can never touch 90C with native fan setting (assume the heatsink is reasonably clean).

Anyway, it so much easier for us to determine if the temperature is normal if you can post a screen capture. Something like this will help us to have a better picture about if the cooling is adequate.
View attachment 842530

Cool, I've been yes'sing the CPU for 10 minutes now and it looks like this...

Is it me or the temperature difference between the heatsink is too much to the cores?

20 minutes later...

 

[h9826790]

Cool, I've been yes'sing the CPU for 10 minutes now and it looks like this...

View attachment 842537

Is it me or the temperature difference between the heatsink is too much to the cores?

All normal, 17C is pretty good delta temperature under that situation.

 

[mateuszd]

All normal, 17C is pretty good delta temperature under that situation.

Cool, if you look at the screenshot I just added you will see 83 degrees and 66 under heatsink, fans still at the same speed of 857.

 

[crjackson2134]

Is there any particular reason to go with a w3690 as opposed to an x5690?

 

[mateuszd]

x5690 can work in a dual cpu config, which considering how expensive dual cpu trays are, I will probably never need.

 

[crjackson2134]

x5690 can work in a dual cpu config, which considering how expensive dual cpu trays are, I will probably never need.

Right, I know that. And it can address 64GB memory easily. Was just wondering if there were some other reason in particular to not get an x5690.

 

[Demigod Mac]

Been rockin a W3690 for four years, solid as a rock. Gelid's GC Extreme thermal compound keeps it nice and cool.

 

[mateuszd]

Been ok so far, I think it started working after SMC reset.