cMac Pro restarts while rendering in Resolve

[h9826790]

96C，that's very wrong，and no wonder why the Mac shut itself down.

May be you should try something like this in MacsFanControl, and see it if help.

Anyway, you may need to open up your Mac again and re-apply the thermal paste, I know it's quite new, but may be you applied too much last time.

 

[wiski15b]

96C，that's very wrong，and no wonder why the Mac shut itself down.

May be you should try something like this in MacsFanControl, and see it if help.
View attachment 641249

Anyway, you may need to open up your Mac again and re-apply the thermal paste, I know it's quite new, but may be you applied too much last time.

I did use setting similar to that, max had to be at 70c though, not 85c, for it not to shut down. I just reapplied the paste to the CPU, I think I may have used to little last time rather than too much.

 

[wiski15b]

I just don't understand why the computer is not ramping up the fans to MAX even when temps get into the 90's. Maybe the system (being a flashed 09) just doesn't know what to do with de-lidded x5690's lol.
Also, anyone know what the intake/ exauhst fans base their temp on? Is it system ambient?

 

[Andy2k]

I did use setting similar to that, max had to be at 70c though, not 85c, for it not to shut down. I just reapplied the paste to the CPU, I think I may have used to little last time rather than too much.

Did the reapplying thermal paste help lower temps? Did CPU A temp always get that hot since you originally applied thermal compound? Who delidded the CPUs? Did you ever check the lidless CPU to make sure it's surface is relatively even? Perhaps your CPU A heat sink isn't seating correctly? How long did it take to get to 96 C. My 5690's don't heat up that fast but they're lidded on an legit 2010 oMP.

 

[wiski15b]

Did the reapplying thermal paste help lower temps? Did CPU A temp always get that hot since you originally applied thermal compound? Who delidded the CPUs? Did you ever check the lidless CPU to make sure it's surface is relatively even? Perhaps your CPU A heat sink isn't seating correctly? How long did it take to get to 96 C. My 5690's don't heat up that fast but they're lidded on an legit 2010 oMP.

A always runs hotter than B, however I assumed this was normal due to its position inside the case. The surface of the die is smooth to the eye, but who knows as any small pits can cause heat issues I think. During stress tests with geek bench,CPU A gets into the mid 80's. However, furmark, gets it way up to 97c and even caused a shutdown in the mid 80's. Reapplying the thermal paste didn't fix the issue. Not sure if the thermal pad at the base of the heatsink could be a culprit.
[doublepost=1469150632][/doublepost]Something tells me the fans are not responding the way they should. Even when it gets in to the mid 80's running geek bench, they are only spinning up to less than half their max rpm.

 

[Andy2k]

A always runs hotter than B, however I assumed this was normal due to its position inside the case. The surface of the die is smooth to the eye, but who knows as any small pits can cause heat issues I think. During stress tests with geek bench,CPU A gets into the mid 80's. However, furmark, gets it way up to 97c and even caused a shutdown in the mid 80's. Reapplying the thermal paste didn't fix the issue. Not sure if the thermal pad at the base of the heatsink could be a culprit.
[doublepost=1469150632][/doublepost]Something tells me the fans are not responding the way they should. Even when it gets in to the mid 80's running geek bench, they are only spinning up to less than half their max rpm.

A always runs hotter than B, however I assumed this was normal due to its position inside the case. The surface of the die is smooth to the eye, but who knows as any small pits can cause heat issues I think. During stress tests with geek bench,CPU A gets into the mid 80's. However, furmark, gets it way up to 97c and even caused a shutdown in the mid 80's. Reapplying the thermal paste didn't fix the issue. Not sure if the thermal pad at the base of the heatsink could be a culprit.
[doublepost=1469150632][/doublepost]Something tells me the fans are not responding the way they should. Even when it gets in to the mid 80's running geek bench, they are only spinning up to less than half their max rpm.

Yes CPU A always runs hotter than CPU B but not by 20 degree celsius and that gap will narrow even more under full CPU load (from looking at the pictures correct me if I'm wrong). I think your CPU A heatsink is either not mating correctly with the board and cpu itself or perhaps the lidless CPU mating surface isn't even. Does CPU B heat up as quickly as A? On my 2010 oMP it takes about 4 to 5 minutes to get into the 60 degrees celsius on both CPU's while running Prime95. Nothing taxes an intel CPU as well as Prime95. The fact that your CPU is reaching 60+ degrees in such a short period of time doesn't seem normal to me. It should take a while before that huge copper and aluminum heatsink starts heating up. Maybe try swapping sockets for CPU A and CPU B to see if the hot temps follow the cpu or stay with CPU A? That might narrow it down to a heatsink / uneven mating surface or a problem CPU. If you do swap the CPU's, after you remove and clean them. Set them top side down on a level surface and shine a flashlight behind it to see if you can see light shining through. If the CPU is relatively even not much light should be coming through from underneath the CPU. Good Luck! Keep us posted.

Edit: I did a test with my oMP 2010 with dual X5690 CPU's. The CPU's started out at an idle temp around at 35 C for CPU A and 31 C for CPU B. It took about almost 3 minutes before my CPU's were in the 60's. Then it seemed to reach max temp around 3 minutes 45 seconds. This is running Prime95 on the maximum heat torture test.

 

[wiski15b]

Yes CPU A always runs hotter than CPU B but not by 20 degree celsius and that gap will narrow even more under full CPU load (from looking at the pictures correct me if I'm wrong). I think your CPU A heatsink is either not mating correctly with the board and cpu itself or perhaps the lidless CPU mating surface isn't even. Does CPU B heat up as quickly as A? On my 2010 oMP it takes about 4 to 5 minutes to get into the 60 degrees celsius on both CPU's while running Prime95. Nothing taxes an intel CPU as well as Prime95. The fact that your CPU is reaching 60+ degrees in such a short period of time doesn't seem normal to me. It should take a while before that huge copper and aluminum heatsink starts heating up. Maybe try swapping sockets for CPU A and CPU B to see if the hot temps follow the cpu or stay with CPU A? That might narrow it down to a heatsink / uneven mating surface or a problem CPU. If you do swap the CPU's, after you remove and clean them. Set them top side down on a level surface and shine a flashlight behind it to see if you can see light shining through. If the CPU is relatively even not much light should be coming through from underneath the CPU. Good Luck! Keep us posted.

Edit: I did a test with my oMP 2010 with dual X5690 CPU's. The CPU's started out at an idle temp around at 35 C for CPU A and 31 C for CPU B. It took about almost 3 minutes before my CPU's were in the 60's. Then it seemed to reach max temp around 3 minutes 45 seconds. This is running Prime95 on the maximum heat torture test.

View attachment 641278

CPU B does heat up a bit slower, but not much slower than A. However, what gets me is its max temp. Under stress test, CPU B maxed at about 78c, whereas CPU A got up to 97c as stated earlier in the thread. Im gonna try swapping them tomorrow and see if its the CPU or Heatsink, ill post back when I see what happens. Interesting that your "relative to ProcHot" is a lot lower than mine as well. Mine is sees my max temp as 103c and yours says about 87c.
[doublepost=1469163719][/doublepost]Well I just swapped the CPU's and it looks to be something to do with the heatsink. CPU A got up to 97c again whereas CPU B only got to about 77c. Interestingly, when I start the stress test, CPU A jumps in temp about 10c almost immediately, leaving a 20c gap between the two for the remainder of the test. CPU B just climbs steadily as you would expect.

 

[Andy2k]

CPU B does heat up a bit slower, but not much slower than A. However, what gets me is its max temp. Under stress test, CPU B maxed at about 78c, whereas CPU A got up to 97c as stated earlier in the thread. Im gonna try swapping them tomorrow and see if its the CPU or Heatsink, ill post back when I see what happens. Interesting that your "relative to ProcHot" is a lot lower than mine as well. Mine is sees my max temp as 103c and yours says about 87c.
[doublepost=1469163719][/doublepost]Well I just swapped the CPU's and it looks to be something to do with the heatsink. CPU A got up to 97c again whereas CPU B only got to about 77c. Interestingly, when I start the stress test, CPU A jumps in temp about 10c almost immediately, leaving a 20c gap between the two for the remainder of the test. CPU B just climbs steadily as you would expect.

I agree with you.

CPU B does heat up a bit slower, but not much slower than A. However, what gets me is its max temp. Under stress test, CPU B maxed at about 78c, whereas CPU A got up to 97c as stated earlier in the thread. Im gonna try swapping them tomorrow and see if its the CPU or Heatsink, ill post back when I see what happens. Interesting that your "relative to ProcHot" is a lot lower than mine as well. Mine is sees my max temp as 103c and yours says about 87c.
[doublepost=1469163719][/doublepost]Well I just swapped the CPU's and it looks to be something to do with the heatsink. CPU A got up to 97c again whereas CPU B only got to about 77c. Interestingly, when I start the stress test, CPU A jumps in temp about 10c almost immediately, leaving a 20c gap between the two for the remainder of the test. CPU B just climbs steadily as you would expect.

Good work, I think you have identified the problem. It sounds like some kind of a problem with the heat-sink or the way it's seating on the CPU. That almost instant jump of 10 C might be due to an air pocket in the thermal compound caused by the heat-sink issue. Did it look like an air pocket in the thermal compound on the CPU or heatsink when you removed it the last time?

Not sure if this would work or not but it might be worth a shot. Maybe you could buy a few sheets of carbon copy paper from a copy shop. What I was thinking was removing CPU A and cleaning up the compound. Very carefully cut a small piece of the paper to completely cover the surface area of the CPU. Now tighten down the heatsink, like you normally would, with the paper sandwiched between the heat-sink and CPU (taking the place of the thermal compound obviously not powering it up). Then remove the heatsink. What I was thinking was if there were any imperfections or irregularities with the heatsink. It might show up on the paper? Like if only a part of the heatsink is actually touching the CPU. Not sure if it will work but if you could get a couple/few sheets of it for cheap, it would be worth a shot. Keep me posted.

 

[h9826790]

"relative to ProcHot" is NOT the CPU temperature, that's the temperature that you AWAY from the CPU overheat state. That means, the lower the better.

 

[JesterJJZ]

In Macs Fan Control whats the difference between CPU Diode and CPU Core From PCECI? The latter is hotter for both CPUs on my machine.

 

[nigelbb]

Integrated heatsink/cap removed from the cpu. 2009 Mac Pro's were shipped this way.

Dual CPU 2009 Mac Pro 4,1 systems use de-lidded CPUs. The single CPU systems do not.

 

[h9826790]

In Macs Fan Control whats the difference between CPU Diode and CPU Core From PCECI? The latter is hotter for both CPUs on my machine.

I have no idea about that at this moment. I can only assume there is something wrong in the software calibration.

As you can see, the temperature go up to 89C under stress, which is 12C above the CPU diode.

However, when at idle, That's just 43C, a bit cooler then the CPU diode.

Since no matter idle or loaded.All the heat should still come from the same source, and getting out via the same path. So I cannot think about why there is a temperature that can be cooler then the diode's temperature in one situation, but warmer in the other.

Unless it's one of the core's temperature. Since the diode's temperature can be increased by other cores's temperature, so that one of the core is slightly cooler then the average diode temp is possible. And of course, that core will warm up quickly when under stress.

 

[pastrychef]

You have to remember, the dual CPU 4,1s and 5,1s were never designed to be used with 130W CPUs. The hottest CPUs Apple ever shipped in them were 95W CPUs. Single CPU 4,1s and 5,1s did ship with 130W CPUs but they also had larger heatsinks. It's not too much of a problem unless you really start doing CPU intensive tasks. If/when you do, you should crank up the fans a bit since by default, Apple never anticipated the CPUs to be running so hot.

In my experience, the key fan to adjust is the exhaust fan. By cranking up the exhaust fan, Boost A and Boost B also spun faster. Of course, this will come at the expense of a bit more noise. I usually kept the default fan settings and only made adjustments when I was about to some CPU intensive tasks.

 

[wiski15b]

You have to remember, the dual CPU 4,1s and 5,1s were never designed to be used with 130W CPUs. The hottest CPUs Apple ever shipped in them were 95W CPUs. Single CPU 4,1s and 5,1s did ship with 130W CPUs but they also had larger heatsinks. It's not too much of a problem unless you really start doing CPU intensive tasks. If/when you do, you should crank up the fans a bit since by default, Apple never anticipated the CPUs to be running so hot.

In my experience, the key fan to adjust is the exhaust fan. By cranking up the exhaust fan, Boost A and Boost B also spun faster. Of course, this will come at the expense of a bit more noise. I usually kept the default fan settings and only made adjustments when I was about to some CPU intensive tasks.

Are the intake and exhaust fans based on the system ambient sensor by default?
[doublepost=1469210167][/doublepost]

Dual CPU 2009 Mac Pro 4,1 systems use de-lidded CPUs. The single CPU systems do not.

Right, thanks for clarifying.
[doublepost=1469210637][/doublepost]

I agree with you.


Good work, I think you have identified the problem. It sounds like some kind of a problem with the heat-sink or the way it's seating on the CPU. That almost instant jump of 10 C might be due to an air pocket in the thermal compound caused by the heat-sink issue. Did it look like an air pocket in the thermal compound on the CPU or heatsink when you removed it the last time?

Not sure if this would work or not but it might be worth a shot. Maybe you could buy a few sheets of carbon copy paper from a copy shop. What I was thinking was removing CPU A and cleaning up the compound. Very carefully cut a small piece of the paper to completely cover the surface area of the CPU. Now tighten down the heatsink, like you normally would, with the paper sandwiched between the heat-sink and CPU (taking the place of the thermal compound obviously not powering it up). Then remove the heatsink. What I was thinking was if there were any imperfections or irregularities with the heatsink. It might show up on the paper? Like if only a part of the heatsink is actually touching the CPU. Not sure if it will work but if you could get a couple/few sheets of it for cheap, it would be worth a shot. Keep me posted.

No air pockets, and it looks like it was seated nice and level.

Combined heat from the close position of the Northbridge?? I mean, the Northbridge isn't running dangerously hot but regularly 73 to 77c at idle and/or under stress.....I'm sure this would have some effect on CPU A temps, right?

 

[SoyCapitanSoyCapitan]

Nobody mentioning power spikes on the GPU. They happen for microseconds but they are high enough under sustained load to shut down an ageing system that only has true support for two 6 pin GPU connectors.

 

[pastrychef]

Are the intake and exhaust fans based on the system ambient sensor by default?

I don't think so. But that's just a guess...

 

[h9826790]

Are the intake and exhaust fans based on the system ambient sensor by default?

AFAIK, their speed are based on the CPU temperature, but not ambient.

 

[wiski15b]

AFAIK, their speed are based on the CPU temperature, but not ambient.

Heatsink temps?

 

[h9826790]

Heatsink temps?

No, should be the CPU diode temperature. There were some reports about cool heatsink temperature but high fan speed (because of poor heatsink installation / thermal paste application etc).

However, Apple in fact may change the fan profile on OSX update. Both the minimum speed and the characteristic. And since I use my own fan control software to handle all the CPU fans. I really not sure how exactly the native fan profile react now.

https://forums.macrumors.com/thread...ol-now-version-2.1340705/page-6#post-20982276

In the past, the native fan profile will keep my CPU run at mid 80. I personally don't like that temperature. Also, I don't like that it spin up the booster fan way more then the others, which increase the fan noise. So, I now make my own fan profile, which auto react base on the CPU diode temperature (and will react to the NB temperature as well, if that's too high), and make all fans spin up a bit, but not mainly rely on booster fan.

And now, even with ambient 35C, the booster fan rarely go over 1400RPM, because both the intake and exhaust will spin up to about 1000RPM, which helps the cooling. This 1400/1000 combination will keep my CPU stay at around 78C with very low fan noise.

The software is on the above link, you can try that if you want to (but better read the thread about how to remove it before you try). And my Mac is the single CPU model, so less problem from heat either. For this software to work, those CPU fans must in auto mode in MacsFanControl.

For your info, Almost all fan control software (including the one that on the link) is controlling the minimum fan speed (by increase the minimum speed, the system will spin up the fan to match the min). Which gives you extra protection. If the fan control software fail, the fan will still react to the system, because the system is actually command the fan to spin up by increasing the target fan speed. And target fan speed has higher priority than the min fan speed.

However, MacsFanControl is actually altering the target fan speed, that's why it's so powerful and can fix the high PCIe fan speed issue. If you use that on the CPU, all intake, exhaust, booster fans can no longer spin up by the system (regardless how hot the CPU is). Again, target fan speed has higher priority, even if the target is lower than then minimum.

So, be very careful if you use MacsFanControl to cool your CPU. You must stay on the safe side, otherwise, zero protection from the system. And will totally rely on CPU throttling, system shutdown, etc as the firewall.

 

[poematik13]

I don't think (by default) resolve is gpu intensive on render, I think it's heavily dependent on cpu.

Quite the opposite. Resolve is incredibly GPU dependent. Nearly every render process in that app is GPU accelerated. It's causing the Titan X to run at full steam and your power supply can't keep up.

He can't. He is now suspended for some reason, and can't post anything.

Because he was constantly derailing threads. Good riddance.

 

[wiski15b]

Quite the opposite. Resolve is incredibly GPU dependent. Nearly every render process in that app is GPU accelerated. It's causing the Titan X to run at full steam and your power supply can't keep up.



Because he was constantly derailing threads. Good riddance.

Titan X isn't showing heavy load when doing final render, the CPU's process shows intense load however. I also had no issues running furmark stress tests on the gpu. My CPU however, crashed on stress tests first try, then repeatedly climbed to 97c. I swapped the cpu's and the problem stayed with heatsink A, not the CPU. I'm pretty sure resolve uses the gpu heavily for most color grading and noise reduction, but converting files on import or export is still very CPU intensive. From the trends, it looks to be at least a 3:2 ratio favoring the CPU.

 

[SoyCapitanSoyCapitan]

Titan X isn't showing heavy load when doing final render,

What are you using to analyse the load on the GPU? I am not aware of any reliable OSX app for analysing GPU activity and GPU temps.

That is a 250W card and as you know it's power envelope is higher than the cMP allows for (regardless of what forum salesboys are saying).

The power spikes of the Titan X can exceed 350W. Those spikes last a fraction of a second in normal scenarios but under a render load the spikes are much more frequent. The system is detecting the consumption and thinks the PCIE bus could be overloaded. The computer can shut down for safety reasons.

 

[pastrychef]

A hot GPU can also contribute to added CPU temps. Particularly the GPU coolers that don't exhaust the heat out of the case.

 

[h9826790]

Sorry, I am not an expert in this area and not very clear about the terms. However, my understanding is that rendering is done mainly by GPU, but encoding (export) is done by CPU, am I correct?

 

[orph]

if your titan is from macvideo cards might be worth sending him an email, i think he supply's support.
if your GPU go's past the power limit your mac pro will shutdown to stop a overload or something.
(might be also worth trying a ram test)

iv seen some reports of problems with the new update to resolve on the FB user group, have you updated to 12.5 recently might be worth downgrading to the last version that was stable on your system.
also if it's always the same clip that has problems then trans coding that clip might fix it if it's borked.