Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

Candidoo

macrumors newbie
Original poster
Jan 31, 2022
9
2
Hi EveryOne,

After an issue about a serious shutdown, i've reed some posts about my potential issue.
Then I do my best to find the knowledge about to do the best to find the Issue with your recommendations.
The mac is a CMP 4.1 --> 5.1 --> 144.0.0.0.0

First thing , I read the posts to upgrade the mac pro from the last firmware to put a nvme with the RIITOP M2TPCE + Sabrent Rocket 2280 M.2 PCIe in Slot 3, it works but i did not have time to format the Disk on time (it's FAT 32 maybe) because i've got a serious shutdown sunday , I've try to resolve the issue till now, and now i'm on the way to lesson your expertise.

My configuration is : CMP 2009 firmware 144.0.0.0.0 - 2 X 3.33 delidded (X5680) - 96GB 1333 - Nvidia GTX 760 flashed by mac_cards (Ebay Spain) - SSD crucial 1To + WD black 1To + WD purple 2To

After that I decide to Put the Nvme + the RIITOP adapter in the slot, it was recognised, coool but as external ! Maybe it's possible to do something to see it as internal to use as a boot drive, i saw a post to do that, is it right ?
ok
The Mac shutdown for the first in my life like that, he shutdown and ok, what 's going on !
-
I push the on/off button, It restart like as normal, at the middle of the started bar it stop, one time, the second time, it start do the same, at the third start he shutdown like , he is like dead but he's not ! i'm not shure he's alive at this moment of my life.
So
I download the service manual and begin to take out everything and clean everything and it smell something strange like a small flavour of burned electric cable or something like that.

Ok let's begin to find what's wrong on this fabulous machine i need to continue my life with "making music my life".

ok I do all the procedure in the service manual, and I find something interesting

Thanks to read me till here right now,

So after the procedure,
backplane board seems to be ok, Power supply seems to be ok, CPU A ok, CPU B ok, Heat sink CPU A ok,
but heatsink CPU B !! Problem !!,
It smell near the temperature testor like burned??

When I put it back with the CPU B, the machine won't start, and it smell so I stop the powersupply
I don't want to continue to burn the heatsink or something else.

Without the CPU B + heatsink B,
CPU A + 16 GB RAM slot 1 + GPU ati radeon HD5770 + SSD MX crucial 1To with 10.13.6
The Mac come back to Life and the fans goes full speed,

I'm so happy during all the procedure a small victory after a smal victory on the diagnostic.
I would like to thanks a lot the forum to that.
Ok ,
please :

- Is it possible that something is wrong on the Dual tray to put the Heatsink B like dead or goes in trouble , is the sensor can be dead and it burn ?
Or
- Is the dual tray is totally ok ? because It start like normal with only one CPU but not with 2 CPU ?
- IS it possible that, maybe because I didnot screws as much as enough hard the CPU HeatSink B, and it put something wrong on CPU B heatsink, hypothetic, i don't know ?

by the end , Is it ok to replace the heatsink CPU B only and see if it goes to normal life ?


Thanks by advance for your advice,
Candide
 

Attachments

  • IMG_20220202_203107_589~2.jpg
    IMG_20220202_203107_589~2.jpg
    537 KB · Views: 376
While everything is possible, I don't think that your problem is the CPU tray B heatsink sensor. Unless is a very mild failure, to the point of being only internal, it's not there that you got the smell. When the sensor fails, extremely rare, it carbonize itself.

Another thing, if it was the temperature sensor, your Mac Pro would boot with both CPUs, but with fans at fail-safe mode (maximum RPM).
 
While everything is possible, I don't think that your problem is the CPU tray B heatsink sensor. Unless is a very mild failure, to the point of being only internal, it's not there that you got the smell. When the sensor fails, extremely rare, it carbonize itself.

Another thing, if it was the temperature sensor, your Mac Pro would boot with both CPUs, but with fans at fail-safe mode (maximum RPM).
Hi Alex,

Thanks for your reply
You were right !

I decide to re-install CPU B + Heatsink B and see what's happen a bit more !, and ...
It's smell more and more and , The MAc won't start!

I take out the dual cpu tray , both CPU and Heatsink and after investigation ,effectively,
it's not the heatsink B
Maybe QG902 smell a lot in this place

what to do now ?


So ....


To be sure, the install of Mojave on an HD and after the install of the Nvme Sabrent with the Riitop adapter on the PCIe can't be the source of this issue, isn't it ?

If I change the dual cpu tray by an other one, is it a good thing to do ?

Is the issue can come back ? I mean , can the issue source could be somewhere else than from the CPU dual card ?

Thx by advance for your advices

Candide
 

Attachments

  • IMG_20220203_145707_829.jpg
    IMG_20220203_145707_829.jpg
    583.8 KB · Views: 181
  • IMG_20220203_173733_260.jpg
    IMG_20220203_173733_260.jpg
    432.4 KB · Views: 165
Last edited:
QG902 is definitively short-circuited, probably QG901 too. Your CPU socket B is dead, that is the PWM/power circuit for it.

You gonna need a replacement tray. While dual CPU trays are expensive, single CPU are not. You can get a single CPU tray and use it until you find a dual CPU tray on the cheap.
 
Hi Alex,

Thx for your reply ,

I saw that on that conversation about CPU compatibility
* 4,1 dual-processor CPU trays made before May 2009 are problematic with 130W Xeons.

Is it possible than these CPU of 130W X5680 3.33 used too much power
on this dual cpu board to burn these QG 902 & 1 ?,

The number of the 2009 dual tray is 630-9402
maybe this dual CPU board 4.1 come from this first generation before may 2009 ?

really don't know
If I knew that before....so

from now

yes you're right need to find one single CPU tray on ebay cheapest than a dual CPU tray!,
and i found a MP 2012 with single CPU
with a very good price,
so I'm gonna try it on sunday noon, will see ;-)
 

Attachments

  • IMG_20220204_150506_154.jpg
    IMG_20220204_150506_154.jpg
    687.6 KB · Views: 282
  • IMG_20220204_150519_987.jpg
    IMG_20220204_150519_987.jpg
    768 KB · Views: 187
Hi Alex,

Thx for your reply ,

I saw that on that conversation about CPU compatibility
* 4,1 dual-processor CPU trays made before May 2009 are problematic with 130W Xeons.

Is it possible than these CPU of 130W X5680 3.33 used too much power
on this dual cpu board to burn these QG 902 & 1 ?,

The number of the 2009 dual tray is 630-9402
maybe this dual CPU board 4.1 come from this first generation before may 2009 ?

really don't know
If I knew that before....so

from now

yes you're right need to find one single CPU tray on ebay cheapest than a dual CPU tray!,
and i found a MP 2012 with single CPU
with a very good price,
so I'm gonna try it on sunday noon, will see ;-)
Since you already have a complete dual CPU tray, you can buy just the dual tray PCB itself, without any accessories like heatsinks/the aluminum tray/etc.

It's a lot cheaper:


Btw, Rev. A CPU trays usually don't even boot with X5680/X5690, I'm surprised that yours even worked.
Screen Shot 2022-02-04 at 18.36.15.png

From the CPU tray serial, 13 week of 2009 makes it a Rev. A.
 
Btw, whenever I need an early-2009 CPU tray, I ask for the serial number and I only buy if it's from end of July onwards to be absolutely conservative and sure that I'm not getting a Rev.A.

For example:

J591300Z01LUC - your tray is from 13th week of 2009, the week from 23 to 29 of March.
J5929xxxxxxxx - example 29th week of 2009 (3rd digit is the year, 4th and 5th are the week number).
 
Since you already have a complete dual CPU tray, you can buy just the dual tray PCB itself, without any accessories like heatsinks/the aluminum tray/etc.

It's a lot cheaper:


Btw, Rev. A CPU trays usually don't even boot with X5680/X5690, I'm surprised that yours even worked.
View attachment 1954559
From the CPU tray serial, 13 week of 2009 makes it a Rev. A.
I was completely novice on doing the CPU replacement
and even know nothing about the speed limit on Xeons on this Dual CPU tray 2009 Rev A. especially,

I 've reed theses threads a bit too late!!
by the way I feel a bit like a kid who's made a mistake grr!

Thanks a lot i for dvwarehouse,
don't know if they are shipping to France
but it will cost a bit more with importation taxes

my firmware is now 144.0.0.0.0 on the backplane board,

May I replace the CPU tray (2009 only) by a single or dual CPU tray (2009 only)
without any reflash of the bootROM firmware, is it transparent for the CPU tray ?
thx by advance for your reply
 
Last edited:
I was completely novice on doing the CPU replacement
and even know nothing about the speed limit on Xeons on this Dual CPU tray 2009 Rev A. especially,

I 've reed theses threads a bit too late!!
by the way I feel a bit like a kid who's made a mistake grr!

Thanks a lot i for dvwarehouse,
don't know if they are shipping to France
but it will cost a bit more with importation taxes

my firmware is now 144.0.0.0.0 on the backplane board,

May I replace the CPU tray (2009 only) by a single or dual CPU tray (2009 only)
without any reflash of the bootROM firmware, is it transparent for the CPU tray ?
thx by advance for your reply
It's not the clock that is the problem, but the power draw.

BootROM is on the backplane. CPU trays don't have BootROM, just SMC and the SMC is not upgradeable.

You can use a dual or single CPU tray, don't matter. Only thing that matter is that for an early-2009 CPU tray, you need an early-2009 backplane and vice-versa. SMC must be the same, 1.39f5.
 
It's not the clock that is the problem, but the power draw.

BootROM is on the backplane. CPU trays don't have BootROM, just SMC and the SMC is not upgradeable.

You can use a dual or single CPU tray, don't matter. Only thing that matter is that for an early-2009 CPU tray, you need an early-2009 backplane and vice-versa. SMC must be the same, 1.39f5.
Hi Alex,

Thank you again for your answer
This information is very useful
when we don't really know the exact difference
between different versions of CMP 4.1 2009

which can be problematic with upgrade with 130w Xeon
You can crash your dual CPU board like I did!


I'll recap to make sure that the lesson it's well understanding.


The next step to replace the CPU tray

My backlane is an early-2009 too, right?


I need to replace my dual tray CPU Rev. A. (early-2009) damaged by:

- either by another dual CPU tray REV A. early-2009 (exactly the same tray with exactly the same SMC 1.39f5
(In this case I need to replace the Xeon X5680 - 130W with less powerful CPUs, for example Xeon of 95W with the capsule delidded)

Or by

- a simple CPU Rev A tray. early-2009 (Single CPU tray cheaper but it's ok, it will works fine, with the same SMC 1.39f5
(idem I can't use a X5680 - 130W, in a Rev A. single CPU, I need to install a less power CPU 95 W is ok)

In all of theses 2 cases
I do not change my backlane early-2009 (System + Tray (1.39f5))

or i can also :

- Continue to use my current early-2009 backlane (System (1.39f5))

with

- another single or double CPU tray from 2009 that accepts 130W xeon , so that is NOT an early 2009 Rev A.

but necessarily with the same SMC 1.39f5

Is it correct please?

thx by advance for your advices

Candide
 

Attachments

  • IMG_20220205_144856_306.jpg
    IMG_20220205_144856_306.jpg
    1.1 MB · Views: 147
Last edited:
This is simple:

An early-2009 backplane needs to be matched to a an early-2009 CPU tray, single or dual. SMC is 1.39f5.

A mid-2010/mid-2012 backplane needs to be matched with a mid-2010/mid-2012 CPU tray, single or dual. SMC is 1.39f11.

Following the SMC rule, you can do whatever you want, upgrade a single CPU Mac Pro to dual or downgrade a dual CPU to single.

Another thing, only early-2009 dual CPU trays require a pair of de-lidded Xeons, early-2009 single CPU require a lidded/normal Xeon. Also, don't use 130W Xeons with Rev. A CPU trays, use a 95W one.
 
Your backplane is an early-2009. Remove any picture or mention of the SN for the backplane, with the MLB SN, someone can clone your Mac Pro. Never write it publicly.
 
  • Like
Reactions: Candidoo
Your backplane is an early-2009. Remove any picture or mention of the SN for the backplane, with the MLB SN, someone can clone your Mac Pro. Never write it publicly.
Thanks a lot Alex for all these advices
and for this fabulous forum
 
Btw, whenever I need an early-2009 CPU tray, I ask for the serial number and I only buy if it's from end of July onwards to be absolutely conservative and sure that I'm not getting a Rev.A.

For example:

J591300Z01LUC - your tray is from 13th week of 2009, the week from 23 to 29 of March.
J5929xxxxxxxx - example 29th week of 2009 (3rd digit is the year, 4th and 5th are the week number).

I’m glad I ran into this thread. Just dusted a MP 4,1 at original form 8c 2.26ghz and was about to order a pair of X5690s but the tray is the same 630-9402 with a J5920xxx.. serial, making it a late May 09 board? Where can these be checked which one is Rev A and which is Rev B?

Years ago I upgraded one 4,1 with X5690s but luckily it had a different tray number 639-1060 and the serial is C0701xxx.. which makes me puzzled when was it made. The serial is long and has alphabets in it, maybe something for short. Both of these MPs came with identifier B07 before flashing which means they were refurbished at some point?

Also the tray on the newly aqcuired MP has a connector that is missing from my work horse MP and I haven’t seen it in online photos either, I wonder what is it about? Photos attached.

So the X5670 option is the top of the line for Rev A if I’m not mistaking as they are 95W? In other words a 12-core 2.93ghz setup would be the max with the Rev A.
 

Attachments

  • 48FADF13-621C-4165-BD2E-27E4CA103E83.jpeg
    48FADF13-621C-4165-BD2E-27E4CA103E83.jpeg
    321.6 KB · Views: 100
  • D3EEA822-7665-460D-9B31-AC373A4CDBFF.jpeg
    D3EEA822-7665-460D-9B31-AC373A4CDBFF.jpeg
    514.5 KB · Views: 137
  • Like
Reactions: Candidoo
I’m glad I ran into this thread. Just dusted a MP 4,1 at original form 8c 2.26ghz and was about to order a pair of X5690s but the tray is the same 630-9402 with a J5920xxx.. serial, making it a late May 09 board? Where can these be checked which one is Rev A and which is Rev B?

Apple internal documentation, not accessible without GSX access.

Years ago I upgraded one 4,1 with X5690s but luckily it had a different tray number 639-1060 and the serial is C0701xxx..

It's a 17-digits SSN, from factory refurbished boards.

which makes me puzzled when was it made. The serial is long and has alphabets in it, maybe something for short. Both of these MPs came with identifier B07 before flashing which means they were refurbished at some point?

Yes.

Also the tray on the newly aqcuired MP has a connector that is missing from my work horse MP and I haven’t seen it in online photos either, I wonder what is it about? Photos attached.

BIGFRANK connector, not really common, but other people here had CPU trays with the connector factory installed before.

So the X5670 option is the top of the line for Rev A if I’m not mistaking as they are 95W? In other words a 12-core 2.93ghz setup would be the max with the Rev A.

X5675 is the fastest 95W Xeon. Use this table for reference:


 
  • Like
Reactions: Candidoo
BIGFRANK connector, not really common, but other people here had CPU trays with the connector factory installed before.
Was thinking it’s probably something like this. The other spots on the boards look like their connectors were desoldered. Maybe they ran diagnostics at the refurb and someone was in a hurry or at work after a long weekend and decided to leave the connector there (or just in case if more diagnostics need to be made). Interesting stuff :)


X5675 is the fastest 95W Xeon.
Thanks for sorting this out, I missed that one on a wiki list of Xeons.

And thanks for the quick response! :)
 
QG902 is definitively short-circuited, probably QG901 too. Your CPU socket B is dead, that is the PWM/power circuit for it.

You gonna need a replacement tray. While dual CPU trays are expensive, single CPU are not. You can get a single CPU tray and use it until you find a dual CPU tray on the cheap.
Has anyone tried to replace these QG901 and QG902 parts on fried trays? Got me wondering if there’s a way to spot which week is the one when they changed this part of the chip set for better power handling, and if it’s possible to swap the weak components to the sturdier ones. They are ridiculously small though and I’m not versed with SMD soldering.
 
More info here:


Btw, different factories over the world have different dates for the start of rev.B production. I try to buy early-2009 CPU trays only after July because of this. From previous experience with BootROMs, the Cork factory usually is the slowest one to implement any changes.
Ahh so there’s no definitive week for it. Gotta take a look of both boards and see if there’s a visual way to differ between the components instead as I’m now quite sure the J5920 board wouldn’t be able to handle the 130W CPUs.
 
Has anyone tried to replace these QG901 and QG902 parts on fried trays? Got me wondering if there’s a way to spot which week is the one when they changed this part of the chip set for better power handling, and if it’s possible to swap the weak components to the sturdier ones. They are ridiculously small though and I’m not versed with SMD soldering.
No one really knows what changed between revisions - the MacPro electric schematic never leaked from Apple.

Btw, unless you have several defective CPU trays to scavenge components, have all the equipment at hand, experience/training, and lot's of free time to kill, it's not in any way economical to repair a defective CPU tray since you can get a replacement bare PCB for $129 + shipping.
 
I’m glad I ran into this thread. Just dusted a MP 4,1 at original form 8c 2.26ghz and was about to order a pair of X5690s but the tray is the same 630-9402 with a J5920xxx.. serial, making it a late May 09 board? Where can these be checked which one is Rev A and which is Rev B?

Years ago I upgraded one 4,1 with X5690s but luckily it had a different tray number 639-1060 and the serial is C0701xxx.. which makes me puzzled when was it made. The serial is long and has alphabets in it, maybe something for short. Both of these MPs came with identifier B07 before flashing which means they were refurbished at some point?

Also the tray on the newly aqcuired MP has a connector that is missing from my work horse MP and I haven’t seen it in online photos either, I wonder what is it about? Photos attached.

So the X5670 option is the top of the line for Rev A if I’m not mistaking as they are 95W? In other words a 12-core 2.93ghz setup would be the max with the Rev A.
Hi everyone,

I'm glad that this thread helps ;-)

I've found this CPU board on ebay.fr today from Italy ,

"Apple Mac Pro 825-7113-A 2009 Dual Socket Logic board 820-2336-A 639-1060" on the ebay.fr search bar

As I'm living in France so it's much easier to receive the cpu tray.

the number is like your 4.1 upgrade : 639-1060 with the serial C07016301XXX

Does it means it's a REV A. refurbished ?

May I use it in replacement of my defected board with my 2 X5680 ?

Or should I have to sold my 2 X5680 for 2X5670 ?

Thx by advance for your reply
Candide
 
Last edited:
  • Like
Reactions: joelkalsi
the number is like your 4.1 upgrade : 639-1060 with the serial C07016301XXX

Does it means it's a REV A. refurbished ?

May I use it in replacement of my defected board with my 2 X5680 ?
Sounds like it's a refurb board and should work okay with 130W CPUs but diy at own risk of course. I've had my refurb board run with X5690s fine for 6 years or so.

Or should I have to sold my 2 X5680 for 2X5670 ?
No, you should sell the 2x5680 for 2x5690 ;)

Edit : X5675 is 95W and works with the first "weaker" boards too and they are 3.06GHz so only 13% less processing power than with dual X5690s if I'm not mistaking. But the difference between X5670 and X5675 is not that much so.. depends on how much the cpus cost in the bay.
 
  • Like
Reactions: Candidoo
The speed differences would probably be negligible for those 6-core Westmeres. For a casual user to even notice it one would need to make serious attempts to measure it in real life performance.
(cpubenchmark.net; table includes the differences in percents too, please see attached).

1663520737993.png

Just grabbed one more cMP myself. The last one for me I hope, huh. It's a 2010 single with one X5680. BootROM seems to be at 138.0.0.0. (144.0.0.0 being the latest).

I think I need to disassemle it and clean out properly. Check the north bridge heatsink and so on. Upgrade the bootROM of course.

I guess there is nothing to do to prevent those tiny components (like QG902) from frying and dying, other than to keep it all clean and cool enough.
 
  • Like
Reactions: Candidoo
The speed differences would probably be negligible for those 6-core Westmeres. For a casual user to even notice it one would need to make serious attempts to measure it in real life performance.
(cpubenchmark.net; table includes the differences in percents too, please see attached).

View attachment 2070977
Just grabbed one more cMP myself. The last one for me I hope, huh. It's a 2010 single with one X5680. BootROM seems to be at 138.0.0.0. (144.0.0.0 being the latest).

I think I need to disassemle it and clean out properly. Check the north bridge heatsink and so on. Upgrade the bootROM of course.

I guess there is nothing to do to prevent those tiny components (like QG902) from frying and dying, other than to keep it all clean and cool enough.

Yea it would likely be of interest only when using CPU intensive tasks. I produce music so I maxed that production cMP back then with dual X5690s and the music production software Logic uses threads pretty evenly. The thing there is I always mix against a master bus which is kind of like having a set of effects on the output stereo track in real time to maximise loudness and polish the mix -- it's a "trick" that I implemented as part of the work pipeline because I was often having to go back to the "final" mix to fix issues here and there when trying to get maximum loudness (..yes I know, so pssst.) so now it's always "what you hear is what you get" and it makes for better finalised songs among sounding "production ready" already when working on a song.

I know some guys used to do this back in the analogue times with Finalizers etc and they got results I used to yearn after, but I think most home studio producers still today prefer to "fix it in the master". The thing with Logic's processing load is that master bus seems to get loaded on a single CPU thread and sometimes it re-shuffles the loads depending what you insert and delete (plugins, tracks) and sometimes it collides with a bus track that is also heavy on plugins and it caused stuttering.

I think they fixed it on Logic at some update or I might have started to use other kind of plugins or just haven't gone so ballsy with bus tracks, because it hasn't done the stutters for a couple of years now. Other than that the 12c 3.46ghz is a monster for music production and I don't know if I barely ever go past half of the processing power available, although I tend to use a lot more plugins and tracks and soft synths than the next guy 😄 Some say it's going overboard and "keep it simple" but the devil's in the details imo.

For video production a cMP with less CPU but more GPU would still do great I think, depending a bit on the platform at hand (Final Cut uses quite a lot of CPU but I guess Resolve is more on the GPU side, maybe a Pixlas mod and a RX 6900 XT is the peak for it).
 
  • Like
Reactions: Candidoo
Yea it would likely be of interest only when using CPU intensive tasks. I produce music so I maxed that production cMP back then with dual X5690s and the music production software Logic uses threads pretty evenly. The thing there is I always mix against a master bus which is kind of like having a set of effects on the output stereo track in real time to maximise loudness and polish the mix -- it's a "trick" that I implemented as part of the work pipeline because I was often having to go back to the "final" mix to fix issues here and there when trying to get maximum loudness (..yes I know, so pssst.) so now it's always "what you hear is what you get" and it makes for better finalised songs among sounding "production ready" already when working on a song.

I know some guys used to do this back in the analogue times with Finalizers etc and they got results I used to yearn after, but I think most home studio producers still today prefer to "fix it in the master". The thing with Logic's processing load is that master bus seems to get loaded on a single CPU thread and sometimes it re-shuffles the loads depending what you insert and delete (plugins, tracks) and sometimes it collides with a bus track that is also heavy on plugins and it caused stuttering.

I think they fixed it on Logic at some update or I might have started to use other kind of plugins or just haven't gone so ballsy with bus tracks, because it hasn't done the stutters for a couple of years now. Other than that the 12c 3.46ghz is a monster for music production and I don't know if I barely ever go past half of the processing power available, although I tend to use a lot more plugins and tracks and soft synths than the next guy 😄 Some say it's going overboard and "keep it simple" but the devil's in the details imo.

For video production a cMP with less CPU but more GPU would still do great I think, depending a bit on the platform at hand (Final Cut uses quite a lot of CPU but I guess Resolve is more on the GPU side, maybe a Pixlas mod and a RX 6900 XT is the peak for it).

It's the same thing on my side, I was looking to mount a good machine with two 5680 (because a little cheaper and very powerful) on a cMP 2009 to make electronic music and put the necessary plugins to do the mixing and the master on Ableton Live.

cMPs are good machines for producing electronic music ;-) AND thanks to the MAc RUmors forum for all these explanations, a good number of people who sell are not aware of this specificity about the 2009 first generation bi-cpu cards.

Well, thank you for your feedback !!
 
  • Like
Reactions: joelkalsi
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.