Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Intel isn't wrong. Please tell me how do you obtain the T-case temperature in real world. There is no sensor on this parameter.
My Mac Pro isn't special. I quote few posts, all their Mac + X5690 combo can run at above 80C without any problem.
I never change the NB paste, because I don't have to.
You are also making assumption that re-paste NB fix the issue, but I can't see the logic behind.
3 random cases, same procedure to fix it. There should be one (or more) actions within the whole procedure fix the issue. I didn't argue with this point. But how can you sure it's the NB paste? As I said, you were force to re-seat the CPU as well. How do to eliminate it wasn't poor contact on the pin? Please tell me how you climate all other factors, but narrow down to only NB paste does matter.
Yes, I know. That why I want to find out the truth, because I may be affected as well (touch wood, by 4,1 still stable so far). And I really appreciated that you share your experience (which I don't have), but it's so hard to me to follow your logic, and how to achieve this conclusion.
Also, please study what's T-case max, and please tell me how to observe that in real world.
As I said, my W3690's T-case max is 68 (directly from Intel as well). Please tell me why my W3690 can run at 82C T-diode (with screenshot as prove) without shutdown, or even thermal throttling (by using the command "
pmset -g thermlog" in terminal)
If we blindly recommend the other re-paste the NB with no valid reason behind, that's not helping the others, but may by creating problems.
In fact, pastrychef just bring up a good pint. If you ever fix that problem on a 5,1 as well. Then you can eliminate the "re-seat" theory, because the clip on the 5,1 make you no need to re-seat the CPU when re-paste the NB.
Last edited:
Sofar only 4.1's because for upgrading the 5.1's are way to expensive here.
I would like to take a 5.1 octo apart to see if there is any difference.
I noticed the actual NB sensor is as small as a delidded CPU, there will also be a problem if you are not precise in adjusting the tension on each side of the NB heatsink! Probably why Apple used springs.
I think you quote THE most important part of my statement!
most will not crash because they are not used as intense.
Of course I tested these MP's after upgrading,...usually a minimum of two days.
I also swap the tray into my own Mac, to work with it all day and see if anything is out of the ordinary.
I ran Valley en Heaven for an hour on extreme specs....no problem.
Ran Geekbench 3-4 times ...I'm just not able to push these Macs as the guy who did the rendering.
I also think that as with most older 4.1's the paste is dried up and the heatsinkk loosely on the NB, will probably mean too that the reading of the sensor cannot be trusted.(sensor must be in the heatsink)
Another thing...when I visited the guy at work and saw the Mac crashing every 10 minutes, I swapped the tray with the newly upgraded 12core I brought...same problem.
Swapped the trays to the newly upgraded machine....same thing happened.
I did notice taking out the processor tray after crashing, that the side of CPU A was warmer than body temp while the CPU B side was cool.
That actually led me to do the NB fix....nothing to loose.
I would like to take a 5.1 octo apart to see if there is any difference.
I noticed the actual NB sensor is as small as a delidded CPU, there will also be a problem if you are not precise in adjusting the tension on each side of the NB heatsink! Probably why Apple used springs.
I think you quote THE most important part of my statement!
most will not crash because they are not used as intense.
Of course I tested these MP's after upgrading,...usually a minimum of two days.
I also swap the tray into my own Mac, to work with it all day and see if anything is out of the ordinary.
I ran Valley en Heaven for an hour on extreme specs....no problem.
Ran Geekbench 3-4 times ...I'm just not able to push these Macs as the guy who did the rendering.
I also think that as with most older 4.1's the paste is dried up and the heatsinkk loosely on the NB, will probably mean too that the reading of the sensor cannot be trusted.(sensor must be in the heatsink)
Another thing...when I visited the guy at work and saw the Mac crashing every 10 minutes, I swapped the tray with the newly upgraded 12core I brought...same problem.
Swapped the trays to the newly upgraded machine....same thing happened.
I did notice taking out the processor tray after crashing, that the side of CPU A was warmer than body temp while the CPU B side was cool.
That actually led me to do the NB fix....nothing to loose.
Sorry for my poor English. So, do you mean
1) You have a Mac Pro suffer from crash. (Mac Pro A)
2) You take the tray out, and then put in in another Mac Pro. (Mac Pro B)
3) The Mac Pro B now suffer from crash.
4) So you can sure it's somewhere in the CPU tray cause the trouble?
Anyway, I sometimes run Handbrake 24/7 on my 4,1 for a week. That's 100% stress for >100 hours. I am sure that's long enough to crash the machine if time / loading is the trigger. And many users here use that Mac Pro for proving heavy duty as well. That's why we want / need the Mac Pro.
Besides that, I agree that the sensors may be wrong in some situations. And I have absolutely no idea where it is.
In my experience with my 5,1, I have found that ambient temperatures to be extremely important and graphics card can make a huge difference in the temps of CPU A. Also, even a small increase in the booster fans can lower temperatures of the CPUs and Northbridge greatly.
Yes, I know. That why I want to find out the truth, because I may be affected as well (touch wood, by 4,1 still stable so far). And I really appreciated that you share your experience (which I don't have), but it's so hard to me to follow your logic, and how to achieve this conclusion.
Well, read all I wrote again.
I'm all about thinking logical, they call me McGuyver because I usually find a fix.
That often means going where others didn't go, thinking outside the box, because of what you state too..."how can this be logical...so I will ignore it!"
What could you possibly loose by trying it yourself? Or you don't trust yourself fixing something without an iFixit guide?
Also, please study what's T-case max, and please tell me how to observe that in real world.
Let me tell you that when you ran 2 or 3 different Apps to check what's going on with your Mac, you should have noticed they all use their own language.
You tell me what the T-diode is!?
Its like the temp of the NB, I look at the IOH diode, no clue what it means...but I know its the NB.
As I said, my W3690's T-case max is 68 (directly from Intel as well). Please tell me why my W3690 can run at 82C T-diode (with screenshot as prove) without shutdown, or even thermal throttling (by using the command "
pmset -g thermlog" in terminal)
Correct.
Again, we are fed with numbers...without really knowing what the diodes do/mean.
So tells me there is a difference between that T-diode and the actual core temperature of the CPU.
What you see in the screenshot is like a minute before it crashed...also the crash reporter told me it was a CPU failure....funny btw,...the Mac automatically rebooted within 10 seconds....go figure.
If we blindly recommend the other re-paste the NB with no valid reason behind, that's not helping the others, but may by creating problems.
I totally 100% disagree!
What problem could you create!!
You can **** up the procedure,.. you will have to detach the motherboard from the tray...leave the cpu's in the socket please and use some tape to make sure they don't fall out during the procedure!
Why??? You can very easily bent the pins of the socket.
So again,...I think that it should become a standard procedure when upgrading a MP....only something to gain.
In fact, pastrychef just bring up a good pint. If you ever fix that problem on a 5,1 as well. Then you can eliminate the "re-seat" theory, because the clip on the 5,1 make you no need to re-seat the CPU when re-paste the NB.
You have zero experience with this procedure, so what makes you think you can have an educated opinion about this!!
[doublepost=1463660757][/doublepost]Sorry for my poor English. So, do you mean
Hey man, Dutch is my native tongue!
1) You have a Mac Pro suffer from crash. (Mac Pro A)
2) You take the tray out, and then put in in another Mac Pro. (Mac Pro B)
3) The Mac Pro B now suffer from crash.
4) So you can sure it's somewhere in the CPU tray cause the trouble?
MP 1 crashes,
I take tray of MP 2 and use in MP 1...crash.
I take tray of MP 1 and stick in MP2...crash
We try MP2 with tray that belongs in it...crash
Anyway, I sometimes run Handbrake 24/7 on my 4,1 for a week. That's 100% stress for >100 hours. I am sure that's long enough to crash the machine if time / loading is the trigger. And many users here use that Mac Pro for proving heavy duty as well. That's why we want / need the Mac Pro.
Thanks for the tip, I should have figured that out myself!
It was kind of a shock to me, to see that thse guys run an App that is OpenGL (cinema 4D) but they wanted a GTX980...but when I was there I learned they cannot use GPU support in Cinema 4D (I don't know why) and were rendering 100% on the CPU's.
Besides that, I agree that the sensors may be wrong in some situations. And I have absolutely no idea where it is.
Me neither! Plus we don't really know what these 'diodes' are.
BTW using the same logical approach I managed to find a problem with another 4.1 octo,...a sensor gave a zero reading (forgot which one)... I had the guy bring me the processor tray. When I tested it in my MP I found out I did have a reading on that sensor, so the problem had to be in the backplane board (and it was)
[/QUOTE]
Well, read all I wrote again.
I'm all about thinking logical, they call me McGuyver because I usually find a fix.
That often means going where others didn't go, thinking outside the box, because of what you state too..."how can this be logical...so I will ignore it!"
What could you possibly loose by trying it yourself? Or you don't trust yourself fixing something without an iFixit guide?
Also, please study what's T-case max, and please tell me how to observe that in real world.
Let me tell you that when you ran 2 or 3 different Apps to check what's going on with your Mac, you should have noticed they all use their own language.
You tell me what the T-diode is!?
Its like the temp of the NB, I look at the IOH diode, no clue what it means...but I know its the NB.
As I said, my W3690's T-case max is 68 (directly from Intel as well). Please tell me why my W3690 can run at 82C T-diode (with screenshot as prove) without shutdown, or even thermal throttling (by using the command "
pmset -g thermlog" in terminal)
Correct.
Again, we are fed with numbers...without really knowing what the diodes do/mean.
So tells me there is a difference between that T-diode and the actual core temperature of the CPU.
What you see in the screenshot is like a minute before it crashed...also the crash reporter told me it was a CPU failure....funny btw,...the Mac automatically rebooted within 10 seconds....go figure.
If we blindly recommend the other re-paste the NB with no valid reason behind, that's not helping the others, but may by creating problems.
I totally 100% disagree!
What problem could you create!!
You can **** up the procedure,.. you will have to detach the motherboard from the tray...leave the cpu's in the socket please and use some tape to make sure they don't fall out during the procedure!
Why??? You can very easily bent the pins of the socket.
So again,...I think that it should become a standard procedure when upgrading a MP....only something to gain.
In fact, pastrychef just bring up a good pint. If you ever fix that problem on a 5,1 as well. Then you can eliminate the "re-seat" theory, because the clip on the 5,1 make you no need to re-seat the CPU when re-paste the NB.
You have zero experience with this procedure, so what makes you think you can have an educated opinion about this!!
[doublepost=1463660757][/doublepost]Sorry for my poor English. So, do you mean
Hey man, Dutch is my native tongue!
1) You have a Mac Pro suffer from crash. (Mac Pro A)
2) You take the tray out, and then put in in another Mac Pro. (Mac Pro B)
3) The Mac Pro B now suffer from crash.
4) So you can sure it's somewhere in the CPU tray cause the trouble?
MP 1 crashes,
I take tray of MP 2 and use in MP 1...crash.
I take tray of MP 1 and stick in MP2...crash
We try MP2 with tray that belongs in it...crash
Anyway, I sometimes run Handbrake 24/7 on my 4,1 for a week. That's 100% stress for >100 hours. I am sure that's long enough to crash the machine if time / loading is the trigger. And many users here use that Mac Pro for proving heavy duty as well. That's why we want / need the Mac Pro.
Thanks for the tip, I should have figured that out myself!
It was kind of a shock to me, to see that thse guys run an App that is OpenGL (cinema 4D) but they wanted a GTX980...but when I was there I learned they cannot use GPU support in Cinema 4D (I don't know why) and were rendering 100% on the CPU's.
Besides that, I agree that the sensors may be wrong in some situations. And I have absolutely no idea where it is.
Me neither! Plus we don't really know what these 'diodes' are.
BTW using the same logical approach I managed to find a problem with another 4.1 octo,...a sensor gave a zero reading (forgot which one)... I had the guy bring me the processor tray. When I tested it in my MP I found out I did have a reading on that sensor, so the problem had to be in the backplane board (and it was)
[/QUOTE]
There is something call logic. I don't have to try by myself, but I know if I jump out from my windows (I live at 43/F), I will die. That's it. It's logic, common sense, not necessary to know everything from our own experience.
May be you are completely right, the problem is from the NB. However, what I can't agree is that must because of the NB thermal paste. You've mention that the heatsink need to be installed with right tension. Will that be the reason? But not necessary the NB thermal paste?
Haha, what a co-incident. My English name during 2nd school is exactly MacGyver. Hahaha.
Lateral thinking is good, but the process should not contradict the logical thinking.
I emphasis I believe that "the whole re-paste procedure fix the issue", but I cannot just accept "the NB paste is the root cause". There is not enough data to eliminate all other possible factors at this moment.
About why I said "that may create more trouble". I just think that being a responsible person, we should advice the others they need to have a plan B in case they break the rivet during the procedure. That's a well know weak part in the 4,1 (after years of thermal stress). I am not saying there is a high chance that they will bend the pin, and kill the Mac etc. I am not that extreme.
Some rivets proved to be brittle, but most were still quite strong, I had to cut them of the board (safer than actually applying force )
Without you doing it yourself, you will not be able to see the logic.
As the NB is by far the hottest component on the Mac, it must be the reason after 6-7 years the paste is dry and offers no adhesive anymore.
Yesterday someone else reported (after reading my NB fix) that he could move the NB heatsink with no effort (push sideways in de middle).
Mac suffers panics and crashes...so soon I will do the same thing there and I already know will fix his issues.
Without you doing it yourself, you will not be able to see the logic.
As the NB is by far the hottest component on the Mac, it must be the reason after 6-7 years the paste is dry and offers no adhesive anymore.
Yesterday someone else reported (after reading my NB fix) that he could move the NB heatsink with no effort (push sideways in de middle).
Mac suffers panics and crashes...so soon I will do the same thing there and I already know will fix his issues.
Thanks for the sharing. May I know what's the temperature after re-paste? I am not trying to argue anything, but just want to know the number. I just want to know if the heatsink is always about 10-15C cooler than the NB even just after apply the GC extreme. I am thinking if there is anyway to improve the NB heatsink efficiency.
I just asked the guy to sent me a new screenshot.
Let me tell you the guy was convinced it was the PSU, he bought another one and didn't even react to my question for iStat numbers. (French chauvinism?)
Only when he found out the new PSU did not solve anything,...I received the screenshot.
MP no 2 did exactly the same...and when I applied the fix it ran like sunshine too....you will have to re-calibrate you logics I'm afraid.
Let me tell you the guy was convinced it was the PSU, he bought another one and didn't even react to my question for iStat numbers. (French chauvinism?)
Only when he found out the new PSU did not solve anything,...I received the screenshot.
MP no 2 did exactly the same...and when I applied the fix it ran like sunshine too....you will have to re-calibrate you logics I'm afraid.
I just asked the guy to sent me a new screenshot.
Let me tell you the guy was convinced it was the PSU, he bought another one and didn't even react to my question for iStat numbers. (French chauvinism?)
Only when he found out the new PSU did not solve anything,...I received the screenshot.
MP no 2 did exactly the same...and when I applied the fix it ran like sunshine too....you will have to re-calibrate you logics I'm afraid.
I am more than happy to learn the new thing, but the new concept must not contradict to the correct logic.
I cannot, I emphasis "I CANNOT" say that your "NB paste" is wrong, but I just cannot eliminate other factors as well. May be you know some more details, which deep in your mind and hard to share with us, and those experience / data allow you to have this conclusion.
The only thing so far I can sure you are wrong is that X5690 will not shut down when the T-diode shows 80C (e.g. by iStat). I am sure you cannot denied this real world fact as well. What Intel says is one thing, but what we observe is another. Again, T-case is not T-diode, IMO, these 2 info are not contradicting each others (I mean T-case max at 78C, and the CPU still running stabling when T-diode at 80C).
Mine runs at 52c
The 12 core did 66c before the fix.
But I know a guy where the NB runs at 80c all the time
Looks like these figures can't be trusted.
[doublepost=1463662856][/doublepost]
Sorry to say that you didn't get it at all,...I said already, read it again...I just explained to you that the CPU core temp and your 80c quote are 2 totally different numbers.
You want to be right or what?
The 12 core did 66c before the fix.
But I know a guy where the NB runs at 80c all the time
Looks like these figures can't be trusted.
[doublepost=1463662856][/doublepost]
Sorry to say that you didn't get it at all,...I said already, read it again...I just explained to you that the CPU core temp and your 80c quote are 2 totally different numbers.
You want to be right or what?
It really depends on ambient. But it seems after long idle (not just after cool boot, or after stress), with native fan speed, assuming the heatsink / fan is clean. The temperature is around 30C above the system ambient.
Low CPU loading sometimes stress NB more than high CPU loading, because that will increase the NB loading (temperature), but will not increase the fan speed to increase cooling effect. (may go up to around 40C above system ambient).
I tried, but still did't quite get it yet. Sorry for my poor English.
Anyway, since both me and OP was talking about the temperature that's shown by the temperature monitoring software, that T-case max is quite irrelevant. And it seems confuse me a bit.
On 15 min, room temp about 60 to 65F
Looks very reasonable to me. 36C above system ambient with idle fan speed is quite average.
It's still climbing. Levels off in about an hour of light use. IOH Tdiode seem normal?