Resolved Mac Pro (Mid-2012) - no video and a very weird mainboard/processor board issue

[Valdaquendë]

I have a Mid-2012 cMP (5,1, A1289 - 2629) that has a very strange problem (or maybe I just don't have enough information). It powers up but won't boot or drive the display. I have followed the Apple Service Manual's diagnostic procedures for this device but cannot narrow the problem down further than "backplane and/or processor board". I have already tried swapping out the GPU, power supply and backplane, so far to no avail. I hope that someone can enlighten me or suggest further steps.

When the system is fully assembled and powered up:
• System boots to power
• System chimes properly (and resets PRAM upon Cmd-Opt-P-R)
• Power button LED turns on and remains steady
• No RAM Diagnostic LEDs illuminate
• System fans and GPU fans start
• Hard drive does NOT spin up
• Keyboard “Caps Lock” LED does NOT light up when "Caps Lock" key is depressed
• GPU fans spin up but no video signal is sent
• System shuts down immediately when the power button is pressed; the normal shut-down delay (for housekeeping, etc.) does NOT take place.
• As mentioned, I have tried swapping out the PSU, GPU and backplane; the symptoms remain the same.

I performed the Minimum Configuration Tests as specified in the Apple Mac Pro Mid 2010 Mid 2012 Technician Guide Service Manual, pages 33-37, and found the following:

1) The backplane (when installed by itself) seems to test OK; with the backplane-only installed and power applied:
- CPU_A LED illuminates (Red for .5 sec)
- CPU_B LED illuminates (Red for .5 sec)
- 5V STBY illuminates upon pressing the DIAG button
- Upon removing power, both CPU LED's flash momentarily (as they are supposed to)

2) When the processor board (stripped of CPUs, heat sinks and RAM) is subsequently installed, the system does NOT function as expected. Initially, it is supposed to behave as shown above. Then the processor board is supposed to be activated by having the technician jump the SYS_PWR solder pads on the backplane, at which point:
- The processor cage fans should start turning slowly
- One RAM LED on the processor board should illuminate
- Upon pressing the DIAG button:
--- PSU_PWROK LED should illuminate
--- 5V STBY LED should illuminate
--- EFI_DONE LED should illuminate
--- GPU_DONE should illuminate

In my case, when power is applied WITHOUT jumping the SYS_PWR solder pads, the system behaves as expected but only for 4 seconds. Then system behaves as follows:
- At 4 seconds, BOTH RAM error LEDs on the processor board illuminate
- At 9 seconds, the processor cage fans begin turning slowly
- All appropriate LEDs (as described above) illuminate as they should

Again, this happens without jumping the SYS_PWR solder pads on the backplane. It is as if the backplane is sending power to the processor board without the essential step of jumping the SYS_PWR pads (and is also illuminating both RAM LEDs). I have actually ordered and installed a replacement backplane (tested and warranted to be working by a reputable vendor that I've used many times and having been careful to obtain the exact same model backplane) only to have it behave the exact same way. I have swapped out the power supply and the GPU for known good ones.

At first blush, the processor board looks to be the problem but if the backplane is not supposed to power up the processor board until SYS_PWR is activated (but is doing so), that may indicate some obscure backplane issue not mentioned in the documentation. At the same time, two backplanes in a row have manifested this behaviour; one of them a tested, warranted component. One that I didn't need, if that is not the problem, since the original one acted the very same way.

Any insight or advice would be gratefully accepted and carefully considered. I hate to add a processor board to the growing list of replaced components if that is not the problem. Whatever your answers, my sincere thanks and appreciation of the time and energy you spent reading and considering this problem.

 

[tsialex]

What's the status of EFI_DONE when you press the DIAG button with a complete CPU tray?

Btw, it's a lot easier to test each suspect component of the defective Mac Pro one at a time with a known working Mac Pro than trying to decipher the signals on board without having very deep knowledge of Mac Pro hardware, diagnostic tools and Apple interfaces.

 

[Valdaquendë]

Thanks for the swift reply, tsialex! I will check that out poste-haste. I dismantled the processor board per the Minimum Configuration Test's directions but will reassemble it, test and report back today.

I have a rather different question; perhaps you or someone can guide me here. Long ago, I was given to understand that, when ordering a backplane, it is important to make sure that the product numbers on the replacement board match those of the original. In the current case, they were:
639-0461 (on the left side of the processor board connector)
631-1427 (on the right side of the processor board connector)
I have always (blindly) followed that advice but I have often wondered (and have done some fruitless Googling): Is there some guide to understanding what these part numbers connote with regard to compatibility with other models or components (like processor boards)? I ask, in part, because the listings I am seeing for processor boards have part numbers (mostly '639-0460'); the incremental difference seems to imply that they work together. I just wondered how to confirm what works with what.

You are, of course, correct with regard to testing. Swapping out is the perfect technique if you've got two. And I'd love it if I did.

 

[tsialex]

Thanks for the swift reply, tsialex! I will check that out poste-haste. I dismantled the processor board per the Minimum Configuration Test's directions but will reassemble it, test and report back today.

I have a rather different question; perhaps you or someone can guide me here. Long ago, I was given to understand that, when ordering a backplane, it is important to make sure that the product numbers on the replacement board match those of the original. In the current case, they were:
639-0461 (on the left side of the processor board connector)
631-1427 (on the right side of the processor board connector)
I have always (blindly) followed that advice but I have often wondered (and have done some fruitless Googling): Is there some guide to understanding what these part numbers connote with regard to compatibility with other models or components (like processor boards)?

You are, of course, correct with regard to testing. Swapping out is the perfect technique if you've got two. And I'd love it if I did.

You can use any mid-2010 or mid-2012 CPU tray, or backplane, the only requirement is the same SMC 1.39f11 - you can't use early-2009 backplane/CPU tray (SMC 1.39f5) with a mid-2010/mid-2012 CPU tray/backplane (SMC 1.39f11):

Switching CPU tray from Mac Pro 2010 to Mac Pro 2012

SMC: The best way to know if your Mac Pro is a real mid-2010/mid-2012 is checking the SMC version with >AboutThisMac>SystemReport>HardwareOverview>SMCVersion(system). If the System SMC version is 1.39f5, it's an early-2009 Mac Pro. If the System SMC version is 1.39f11, it's a real mid-2010...

forums.macrumors.com

Correct part numbers for the CPY tray on the post linked above.

Btw, a mid-2012 Mac Pro don't have a processor board, seems you are talking about the backplane - the order number for a mid-2010/mid-2012 backplane is 661-5706.

Backplanes:

Year model:	Part Number:	Part Number:
early-2009	661-4996	661-5444
mid-2010 / mid-2012	661-5706

 

[Valdaquendë]

OK; I am so glad you pointed me to this information, tsialex!. The tables and other information will invaluable in future. Thank you many times. I intended to respond this morning but it's been one very busy day.

I'm finally done working for clients and will reassemble the processor board, install it and check the EFI_DONE LED while pressing DIAG as you suggest. I'll report the result.

I'm not clear about the remark that the Mid-2012 Mac Pro does not have a processor board. This one does. Its case has the serial number which This Everymac.com page shows as being a 2012 (Westmere) Mac Pro. Apple's Tech Spec Page for the Mid-2012 Mac Pro shows its standard configuration to be a match for this one; it has two 2.4GHz 6-Core Intel Xeon E5645 processors as also shown in the excellent Mac Pro CPU Compatibility List you linked to. Its backplane numbers are:
Part Number: 820-2337-A
Left side of processor board connector: 639-0461
Right side of processor board connector: 631-1427

The backplane has a connector for a processor board and a processor board is installed with the 2 2.4GHz CPUs. Its numbers:
Part Number: 820-2742-A
Barcode Label Part Number: 639-0640

This system came to me in an un-bootable state (which I'm now trying to cure) so of course I can't determine the SMC number. I would imagine that in such a situation it would be reasonable to presume that they are the same as long as the year, serial and barcode numbers indicate it. The same would be true of any processor board I acquired, I guess.

As a matter of curiosity, I wondered whether the fact that the backplane's number and the processor board's number (639-0461 and 639-0640 respectively) being one number apart was indicative of their compatibility in any way. Just curious.

Thanks again, tsialex, and please don't hesitate to correct me in any particular. As Emil Faber, founder of Faber College, famously said, "Knowledge is good."

 

[Valdaquendë]

tsialex - I assembled the processor board, installing the CPUs, heat sinks and RAM. Please note that I did not install the CMOS battery, which the Minimum Configuration Test has you remove. I can repeat these tests with the battery installed if this is helpful.

I found that, when I applied power WITHOUT the DIAG switch:
- CPU_A LED illuminates (Red for .5 sec)
- CPU_B LED illuminates (Red for .5 sec)
- After about 4 seconds, the power supply emits a click and the RAM LEDs briefly flash
- After about 9 seconds, the fans begin turning slowly
- If I press the DIAG switch after these events take place, the CPU_A, CPU_B and SYS_PG LEDs illuminate as long as the DIAG switch is held down.
- After about 30 seconds the system chime is heard
- When power is unplugged the CPU and RAM LEDs flash briefly
- At no point does the EPI_DONE LED illuminate

When I apply power WITH the DIAG switch pressed a second or so after applying the power:
- CPU_A LED illuminates (Red for .5 sec)
- CPU_B LED illuminates (Red for .5 sec)
- PSU_PWROK, SYS_PWR and SYS_PG all illuminate as long as the DIAG switch is held down
- After about 5 seconds:
--> the EPI_DONE LED illuminates briefly
--> the GPU_OK LED illuminates briefly (twice)
- After about 15 seconds:
--> the EPI_DONE LED illuminates briefly
--> the GPU_OK LED illuminates briefly (twice)
- After about 30 seconds the system chime is heard
- When power is unplugged the CPU and RAM LEDs flash briefly

I don't know if this is helpful or not; happy to perform any further diagnostics you consider advisable. Thanks again for your help.

 

[zoltm]

@tsialex @Valdaquendë I am reading this thread with great interest. I am also in a very similar position where my trusted and very stable cMP 2009 5,1 failed to reboot after a freeze. That's the only cMP I have access to, ability to swapping out components for trouble shooting are severely limited.

I will look at ideas here to see if I can at least getting my cMP chime again, mine can't even chime! The LED status of my cMP are very similar to this one that @Valdaquendë described above!

 

[tsialex]

tsialex - I assembled the processor board, installing the CPUs, heat sinks and RAM. Please note that I did not install the CMOS battery, which the Minimum Configuration Test has you remove. I can repeat these tests with the battery installed if this is helpful.

I found that, when I applied power WITHOUT the DIAG switch:
- CPU_A LED illuminates (Red for .5 sec)
- CPU_B LED illuminates (Red for .5 sec)
- After about 4 seconds, the power supply emits a click and the RAM LEDs briefly flash
- After about 9 seconds, the fans begin turning slowly
- If I press the DIAG switch after these events take place, the CPU_A, CPU_B and SYS_PG LEDs illuminate as long as the DIAG switch is held down.
- After about 30 seconds the system chime is heard
- When power is unplugged the CPU and RAM LEDs flash briefly
- At no point does the EPI_DONE LED illuminate

When I apply power WITH the DIAG switch pressed a second or so after applying the power:
- CPU_A LED illuminates (Red for .5 sec)
- CPU_B LED illuminates (Red for .5 sec)
- PSU_PWROK, SYS_PWR and SYS_PG all illuminate as long as the DIAG switch is held down
- After about 5 seconds:
--> the EPI_DONE LED illuminates briefly
--> the GPU_OK LED illuminates briefly (twice)
- After about 15 seconds:
--> the EPI_DONE LED illuminates briefly
--> the GPU_OK LED illuminates briefly (twice)
- After about 30 seconds the system chime is heard
- When power is unplugged the CPU and RAM LEDs flash briefly

I don't know if this is helpful or not; happy to perform any further diagnostics you consider advisable. Thanks again for your help.

I don't understand why you did follow the minimal steps of the Apple Technician Guide, do you have access to a LITTLE FRANK interface? If you have, you just need to connect to the Apple server running the diagnostic software to know exactly what is wrong with your Mac Pro. If you don't have access, don't make sense to follow it.

Anyway, before anything, please confirm that your CPU tray is really working with a known working Mac Pro, seems you have more than one problem since EFI_DONE should be fully lit when you press the DIAG button, not flashing or turning off. If the CPU tray is not reliable, you won't have reliable results for the DIAG button. Also, test your Apple OEM GPU and be sure that is working.

Be sure that your CPU tray is working and when you start testing again install the RTC battery, be sure that it's voltage no less than 3.0V.

 

[Valdaquendë]

tsialex - First, let me be clear; I am not an Apple-Employed or Apple-Certified Technician, though I have worked on and performed troubleshooting and repairs on various Macs for about 25 years, now. I am an independent technician with roughly 30 years' experience with Macs and PC's and have gotten what knowledge I have from service manuals, forums like this one, practical experience, etc. I do not have access to a LITTLE FRANK interface nor the server that it interfaces with; in fact, though I had noticed "BIG FRANK" and LITTLE FRANK" on Mac Pro backplanes in the past, I was unable to find or ascertain any reliable information (other than speculative remarks) about them until I read your posts in this forum. Thank you for that info.

I do not have another such Mac Pro that I can swap out parts with.

This system came to me in a non-bootable state; its former owner said it had Sierra installed and had been working until a power outage; after that it would not boot. I swapped the PSU for a known good one to make sure that was not the problem; it wasn't. In following the steps in the Apple Technician's Guide's Symptom Charts Overview (Pg. 39), I got as far as Step 2 (Powers on, System Chime, No video).

Per its directions, I checked the Memory Diagnostics LEDs (Pgs 23-24); none lit. Going to the "Video Quick Check" chart (Pg. 53), I reseated all DIMMs just to be thorough. Attaching a known-good display, I got no video. Nor could I hear the hard drive spin up. I went to the "No Video Deep Dive".

The video card in the MP was a third party card (XFX AMD Radeon RX 580 8GB) that had, according to the owner, been working. The AMD Radeon 580 is known to work in MP's of this model when they are running Sierra (I have successfully installed them in two other MP 2010/2012's). I couldn't be sure that the existing card was OK, so I ordered a Radeon Rx PULSE 580, a card also known to work in this model with Sierra. It did not work either.

Concluding that the card was probably not the issue, I reset the RTC and verified that the front panel LED was illuminated and not flashing; however the Caps Lock key did not illuminate when pressed; I tried three different wired Apple keyboards to make sure that this was the case. I concluded that this meant that there was some more fundamental cause; following the "No Video Deep Dive" didn't seem to make any sense, as the required conditions were not met. Nor did it address the "no drive spinup". So I turned to the Minimum Configuration Tests to try to identify the problem on a component-by-component basis and reached the point described in my original post. Then, thinking that the backplane was probably the issue, I ordered and installed a tested and warranted replacement, which acted in the same way as the original. And here I am.

BTW, I installed a known-good battery and found that when power is applied:
- CPU_A LED illuminates (Red for .5 sec)
- CPU_B LED illuminates (Red for .5 sec)
The system is otherwise inert.
If the DIAG button is pressed, only the 5V_STBY LED illuminates
If I disconnect power, press the RTC Reset button and apply power, it behaves as described in my last posting.

I had had the front panel board out (per the MCT) when reporting the above; I then reinstalled it and retested. With front panel board installed, pressing the power button and immediately pressing the DIAG button:
The PSU_PWROK, 5V_STBY and SYS_PG illuminate (and stay on as long as the DIAG switch is held down)
--> the EPI_DONE LED illuminates briefly
--> the GPU_OK LED illuminates briefly (twice)
After 20 seconds or so the System Chime is heard

Please correct me if I am wrong or add anything you consider significant; again, I really appreciate your help and the invaluable information you have pointed me to but from your remark ("If the CPU tray is not reliable, you won't have reliable results for the DIAG button."), it seems reasonable to conclude that the CPU tray is at fault or at least suspect. As I cannot avail myself of LITTLE FRANK or swapping out with another cMP, the remaining step would appear to be replacing the CPU tray. I expect I'll need to obtain a dual-processor board with part number: 820-2742-A and barcode label part number: 639-0640. Even though my understanding is that Apple's reuse and switching SMC's in CPU trays was limited to single-CPU trays, I'll make sure that the vendor warrants the part as a genuine 2010/2012 and will make sure that it is returnable.

It's embarrassing to find myself in this spot, though I feared it might be so. I very much appreciate your help and hope that I haven't taxed your time, patience or sagacity too severely. You are clearly a pillar, in this forum, and I would not willingly have abused your time or attention. Any advice or suggestions will be appreciated and heeded to the best of my ability to do so.

 

[tsialex]

tsialex - First, let me be clear; I am not an Apple-Employed or Apple-Certified Technician, though I have worked on and performed troubleshooting and repairs on various Macs for about 25 years, now. I am an independent technician with roughly 30 years' experience with Macs and PC's and have gotten what knowledge I have from service manuals, forums like this one, practical experience, etc. I do not have access to a LITTLE FRANK interface nor the server that it interfaces with; in fact, though I had noticed "BIG FRANK" and LITTLE FRANK" on Mac Pro backplanes in the past, I was unable to find or ascertain any reliable information (other than speculative remarks) about them until I read your posts in this forum. Thank you for that info.

I do not have another such Mac Pro that I can swap out parts with.

This system came to me in a non-bootable state; its former owner said it had Sierra installed and had been working until a power outage; after that it would not boot. I swapped the PSU for a known good one to make sure that was not the problem; it wasn't. In following the steps in the Apple Technician's Guide's Symptom Charts Overview (Pg. 39), I got as far as Step 2 (Powers on, System Chime, No video).

Per its directions, I checked the Memory Diagnostics LEDs (Pgs 23-24); none lit. Going to the "Video Quick Check" chart (Pg. 53), I reseated all DIMMs just to be thorough. Attaching a known-good display, I got no video. Nor could I hear the hard drive spin up. I went to the "No Video Deep Dive".

The video card in the MP was a third party card (XFX AMD Radeon RX 580 8GB) that had, according to the owner, been working. The AMD Radeon 580 is known to work in MP's of this model when they are running Sierra (I have successfully installed them in two other MP 2010/2012's). I couldn't be sure that the existing card was OK, so I ordered a Radeon Rx PULSE 580, a card also known to work in this model with Sierra. It did not work either.

Concluding that the card was probably not the issue,

Don't get me wrong, but how did you conclude it? Both cards don't even have pre-boot configuration support when installed on a Mac Pro. Get yourself an AppleOEM GPU or you won't conclude the diagnostic process, a cheap Apple GT120 from early-2009s will do it fine for diagnostic purposes.

I reset the RTC and verified that the front panel LED was illuminated and not flashing; however the Caps Lock key did not illuminate when pressed; I tried three different wired Apple keyboards to make sure that this was the case. I concluded that this meant that there was some more fundamental cause; following the "No Video Deep Dive" didn't seem to make any sense, as the required conditions were not met. Nor did it address the "no drive spinup". So I turned to the Minimum Configuration Tests to try to identify the problem and reached the point described in my original post. Then, thinking that the backplane was probably the issue, I ordered and installed a tested and warranted replacement, which acted in the same way as the original. And here I am.

Please correct me if I am wrong or add anything you consider significant; again, I really appreciate your help and the invaluable information you have pointed me to but from your remark ("If the CPU tray is not reliable, you won't have reliable results for the DIAG button."), it seems reasonable to conclude that the CPU tray is at fault or at least suspect. As I cannot avail myself of LITTLE FRANK or swapping out with another cMP, the remaining step would appear to be replacing the CPU tray. I expect I'll need to obtain a dual-processor board with part number: 820-2742-A and barcode label part number: 639-0640. Even though my understanding is that Apple's reuse and switching SMC's in CPU trays was limited to single-CPU trays, I'll make sure that the vendor warrants the part as a genuine 2010/2012 and will make sure that it is returnable.

It's embarrassing to find myself in this spot, though I feared it might be so. I very much appreciate your help and hope that I haven't taxed your time, patience or sagacity too severely. You are clearly a pillar, in this forum, and I would not willingly have abused your time or attention. Any advice or suggestions will be appreciated and heeded to the best of my ability to do so.

Borrow/buy a working CPU tray from a mid-2010 or mid-2012, any model of CPU tray single or dual CPU from mid-2010 or mid-2012 will work with a mid-2012 backplane. Forget the copyright part numbers on the CPU tray, read my post I linked to you before, use the part numbers from the last page of the Apple Technician Guide to search/buy one.

If you support Mac Pros early-2009 to mid-2012, you need an AppleOEM GPU at hand (any AppleOEM from early-2009 to mid-2010 will work, models are GT120/HD4870/HD5770/HD5870), you will need to borrow/buy one.

After you get a working CPU tray and an AppleOEM GPU you can start to really diagnose your Mac Pro without Apple internal tools.

 

[Valdaquendë]

You were right to correct me; my conclusion was based upon the evidence at hand; as you correctly pointed out, it can be as flawed as the holes in that evidence. Thanks.

I will get a dual-CPU tray, since, if this one is at fault, I'll then have the correct replacement at hand. Will get a #661-5708. I'll also dig up an Apple OEM GPU.

Thanks again, tsialex; I very much appreciate both the information and the education. I'll report back with results.

 

[tsialex]

You were right to correct me; my conclusion was based upon the evidence at hand; as you correctly pointed out, it can be as flawed as the holes in that evidence. Thanks.

I will get a dual-CPU tray, since, if this one is at fault, I'll then have the correct replacement at hand. Will get a #661-5708. I'll also dig up an Apple OEM GPU.

Thanks again, tsialex; I very much appreciate both the information and the education. I'll report back with results.

Dual CPU trays for mid-2010/mid-2012 are very expensive, since your current one could be working/repairable, maybe a single CPU tray or even a complete working single CPU Mac Pro will be cheaper or a more reasonable expense. Since you also have to buy an Apple OEM GPU for having pre-boot configuration support (aka boot screens), take a look at locally sold mid-2010/mid-2012 Mac Pros before starting on another buying spree.

Even a cheap enough early-2009 will be extremely helpful to diagnose your failed one, the SMC firmware of the early-2009 backplane/CPU tray will be a mismatch with your mid-2012 ones and the fans will run at full RPM/full time, but it will work fine for testing all the mid-2012 components.

 

[Valdaquendë]

Thanks; good advice. I actually found, locally, a single-CPU tray (mid-2010 - 5,1 - Westmere with hex-core 3.33GHz); based on your remarks, this should work. I will try to get it this afternoon. As soon as I can find an OEM card I'll perform the tests and report back.

Question (seems painfully obvious, but want a clear understanding): Any CPU (or matched set of CPUs) specified in the Apple Technical Guide can be installed in the appropriate tray to alter a system's speed without any other modifications, correct?

 

[tsialex]

Thanks; good advice. I actually found, locally, a single-CPU tray (mid-2010 - 5,1 - Westmere with hex-core 3.33GHz);

If the price is right, go for it.

based on your remarks, this should work.

Correct.

I will try to get it this afternoon. As soon as I can find an OEM card I'll perform the tests and report back.

Question (seems painfully obvious, but want a clear understanding): Any CPU (or matched set of CPUs) specified in the Apple Technical Guide can be installed in the appropriate tray to alter a system's speed without any other modifications, correct?

From factory, any CPU that works with a MacPro5,1 will be recognized without anything else than a triple NVRAM reset (you only need to do it when you change the memory controller from 1066 to 1333MHz, like when upgrading from a Nehalem to a Westmere Xeon), see below for all the supported Nehalem and Westmere Xeons that work:

Mac Pro CPU Compatibility List

Bokkow and I have created a Mac Pro CPU compatibility list. The intent is to cover all relevant and compatible CPU upgrades for all Mac Pros. These are asked about over and over and over again, so I thought it would be nice to have it all in one place. Constructive criticism, corrections to...

forums.macrumors.com

 

[tsialex]

BTW there are no matched pairs for Nehalem and Westmere Xeon generations, Intel didn't manufactured multiple/different S-Specs for a specific model for these two generations. All retail models have the same S-Spec for a determined model/frequency, check the documentation and you can see for yourself. All retail X5690 are exactly the same, for example.

The only way that you won't get a matched pair is if you combine an engineering sample/QS and a retail one.

 

[Valdaquendë]

I now have the CPU tray I mentioned (single-CPU tray, mid-2010 - 5,1 - Westmere with hex-core 3.33GHz) and have ordered an Apple ATI Radeon HD5870, which should be here in a week. Will report results at that time.

Many thanks, tsialex!

 

[Valdaquendë]

tsialex - I received the GPU yesterday. I reinstalled the original backplane, installed the swap-out processor tray and the OEM GPU, reset PRAM and attempted to boot. The system booted (with video) to ye olde blinking question mark disk. I replaced the drive with a known-good SSD (OSX 10.13 - Sierra) and it booted properly.

I ran the system for some time and have now installed an SSD with Sierra as the system's boot drive, in part because I need a clean boot device going forward and in part to put it through some paces. So far, it has run for hours (and been through several tests) without apparent problem.

Clearly, the diagnostic path you suggested was the proper one; thank you. If there are any diagnostic steps that good practice would dictate before going forward, please don't hesitate to point them out. If there aren't, I'd like to outline what I consider to be the way forward and ask for any corrections/suggestions you may care to offer.

It seems clear that the original CPU tray is faulty, as its behaviour was aberrant even when stripped of CPUs, RAM, etc.; I'll need to replace it. If my understanding of the remarks on the Mac Pro CPU Compatibility List is correct, I can use the current processor tray to test the functionality of the CPUs and determine whether I need a fully-equipped tray, as opposed to a "tray sans-CPUs." I'll also install the GPU that came with the system to test whether it is OK. The Radeon 580 family runs properly on systems from 10.11 to 10.13, if I recall correctly, though I have seen reports of XFX 580's (which this one is) having problems with 10.14 and up.

I ran TechTool ProToGo 11 on the system and it found no errors; wanting to test the RAM, I loaded 1/2 of the 8 4GB modules (which is all a single-CPU tray can hold), ran tests, swapped them out for the other 4 and ran the tests again. All tested OK. I've noted no other unexpected phenomena. I'd like to be thorough, making sure there are no hidden issues; if you can suggest any other advisable diagnostics, I will follow through.

======================

Also, I'd like to ask a question or two regarding the numbers on the Mac Pro backplane; I have done some pretty extensive searching but have not been able to find a cogent explanation. If there is some horribly-obvious source of this info, sorry to say that I missed it.

There are two sets of numbers on the Mac Pro backplane; the set on the left side of the CPU tray connector has a barcode label with a serial number (asterisks on both sides) and another number to the right of it (639-0461, in my case). On the right side of the connector, there is another barcode label with a serial number (with asterisks) and a number to the right of it (631-1427, in this case).

I had read, on a forum, somewhere long ago, that it was important, when obtaining a replacement backplane, to ensure that, along with the part number, both of these barcode part numbers match the original's. I have scrupulously followed this advice ever since but I don't know why there are two or what they (particularly the right-hand number) indicate. I'd really like to understand the significance of both of these backplane numbers and their implications with regard to compatibility. If the Tech Guide's part number is found nowhere on the board (be it backplane or CPU board), how can one tell which part is compatible with which? Take, for example, a part that was purchased at an auction or acquired as a piece-part. Reading your posting on the subject, it sounds like you really need to install the part in an otherwise-bootable environment and check the SMC number. Have I got that right?

Thanks again for your sage advice and counsel.

 

[tsialex]

Yes, you can test the Xeons from the dual CPU tray.

Btw, before starting to test anything again you should check the state of the BootROM image, your NVRAM volume VSS stores can be full of KPs and a failed garbage collection will certainly brick your Mac Pro.

OpenCore on the Mac Pro

Thank you @tsialex as always! Your knowledge overwhelms me! I have fixed my problem just reflashing BootROM using the reconstructed one that you made it for me. Now everything is ok again

forums.macrumors.com

You will probably need a BootROM reconstruction.

 

[tsialex]

Also, I'd like to ask a question or two regarding the numbers on the Mac Pro backplane; I have done some pretty extensive searching but have not been able to find a cogent explanation. If there is some horribly-obvious source of this info, sorry to say that I missed it.

There are two sets of numbers on the Mac Pro backplane; the set on the left side of the CPU tray connector has a barcode label with a serial number (asterisks on both sides) and another number to the right of it (639-0461, in my case). On the right side of the connector, there is another barcode label with a serial number (with asterisks) and a number to the right of it (631-1427, in this case).

I had read, on a forum, somewhere long ago, that it was important, when obtaining a replacement backplane, to ensure that, along with the part number, both of these barcode part numbers match the original's. I have scrupulously followed this advice ever since but I don't know why there are two or what they (particularly the right-hand number) indicate. I'd really like to understand the significance of both of these backplane numbers and their implications with regard to compatibility. If the Tech Guide's part number is found nowhere on the board (be it backplane or CPU board), how can one tell which part is compatible with which? Take, for example, a part that was purchased at an auction or acquired as a piece-part. Reading your posting on the subject, it sounds like you really need to install the part in an otherwise-bootable environment and check the SMC number. Have I got that right?

Thanks again for your sage advice and counsel.

Back when new, would made sense to buy a replacement backplane from the exact same model of Mac Pro, so you get the same HWC and etc and AboutMyMac would show the correct model for the case and etc. Today to do this is completely non-sense and it's basically a way to segment the market and sell more expensive backplanes to people that don't know better.

Mid-2010/mid-2012 backplanes are fully interchangeable, don't matter if the Mac was originally a single CPU and you want to install the backplane to a dual CPU Mac Pro and vice-versa.

You wouldn't use an early-2009 backplane since the SMC version won't match. You can clearly know what's an early-2009 from a mid-2010/mid2012 from the MLB label SSN suffix, see my post about that.

Even if you have a mid-2012 and install a mid-2010 used backplane and now you have a mid-2010 with AboutMyMac, a BootROM reconstruction service can match the backplane with the mid-2012 case correctly, exactly like an Apple Service Center would do with an Apple refurbished backplane.

Btw, after 9/11/12 years only a never used backplane won't have lot's of trash/superseded settings inside the NVRAM volume. A BootROM reconstruction should be part of the refurbishment process and once you have the never booted BootROM image for the Mac Pro, you can flash it again whenever needed.

 

[Valdaquendë]

Thanks for this excellent advice; although I have worked on close to a dozen cMPs, I have never run into a bricked one or some of the other problems that point to NVRAM issues (though I have routinely reset NVRAM on cMPs every few months using a 3-chime regime as a matter of maintenance) and I was entirely unaware of these issues. Thank you for expanding my mind on this point. I can see there is a lot to learn and I'm eager to do so. I read several pages of posts on either side of the linked one to try to better acquaint and orient myself on this topic and can see I'll need to read more.

I will download and run ROMTool by DosDude1 and follow the other directions in your excellent and informative post. I'm stuck doing other tasks today but will follow up on this tomorrow. I am really looking forward to learning/understanding the examination, reconstruction, backup and flashing of BootROM.

The system I am working on now has a clean install of MacOS 10.12 (Sierra) but I will be attentive, in future, to your admonition that all security apps (AVG, Malwarebytes, etc.) should be removed using AppCleaner before installing and running ROMTool. I'll report the results of the dump tomorrow.

Thanks also for your words of wisdom regarding compatibility; it really clarified my understanding.

 

[Valdaquendë]

tsialex - OK; I downloaded ROMTool and UEFITool. I checked my EEPROM version (25L3205D), disabled SIP, loaded and ran ROMTool and opened the dump with UEFITool. The following image shows the state of the first VSS Store's (24576) free space:

I also noted that there were at least 38 instances of "Invalid" entries in the first VSS store; this is a bit of a new world to me and I don't know if this is significant. Then I performed a deep NVRAM reset (5-chimes) and repeated the steps above. The free space was exactly the same (24576 - not surprising, perhaps, since I had reset PRAM many times during previous work on this system). I noted, however, that the number of "Invalid" entries in the first VSS Store dropped from 38 to 15 in number.

In your posting, you pointed to healthy first-VSS store free spaces of:
45000 - 40000 for Single CPU MP with 3 DIMMs
35000 - 30000 for Dual CPU MP with 8 DIMMs

This system originally had, as you will recall, dual Xeon E5645 CPUs and 8 DIMMs. Right now, it is running with the substitute CPU tray which has a single CPU (Xeon 6-core) and 4 DIMMs. I don't know if the switch in CPUs is a relevent factor in evaluating the ROM. In either case, though, the free space is inadequate.

After reading the portion of the post regarding overwritten secondary VSS stores, I checked the dump for the second VSS Store; it exists and its free space size is 65448. I'm not sure if this is helpful information or not.



I guess my basic question is what can I do to restore/reconstruct the ROM, at least to the degree possible. You mentioned BootROM reconstruction. How can this be accomplished? Once done, I will ABSOLUTLEY follow your advice and create a dump file that can then be periodically flashed back onto the ROM; that struck me as extraordinarily good counsel.

Later today, or tomorrow morning, I will test the original CPUs to determine whether I need a DP tray with or without CPUs.

Again, tsialex, thank you so much for sharing your knowledge, time and energy. It is very much appreciated.

 

[tsialex]

tsialex - OK; I downloaded ROMTool and UEFITool. I checked my EEPROM version (25L3205D)

Weird, it's extremely rare a mid-2012 to have MXIC 25L3205D. Maybe it's a very early build date?

, disabled SIP, loaded and ran ROMTool and opened the dump with UEFITool. The following image shows the state of the first VSS Store's (24576) free space:

View attachment 1871545

I also noted that there were at least 38 instances of "Invalid" entries in the first VSS store; this is a bit of a new world to me and I don't know if this is significant. Then I performed a deep NVRAM reset (5-chimes) and repeated the steps above. The free space was exactly the same (24576 - not surprising, perhaps, since I had reset PRAM many times during previous work on this system). I noted, however, that the number of "Invalid" entries in the first VSS Store dropped from 38 to 15 in number.

In your posting, you pointed to healthy first-VSS store free spaces of:
45000 - 40000 for Single CPU MP with 3 DIMMs
35000 - 30000 for Dual CPU MP with 8 DIMMs

This system originally had, as you will recall, dual Xeon E5645 CPUs and 8 DIMMs. Right now, it is running with the substitute CPU tray which has a single CPU (Xeon 6-core) and 4 DIMMs. I don't know if the switch in CPUs is a relevent factor in evaluating the ROM. In either case, though, the free space is inadequate.

After reading the portion of the post regarding overwritten secondary VSS stores, I checked the dump for the second VSS Store; it exists and its free space size is 65448. I'm not sure if this is helpful information or not.

Seems your 1st VSS store have lot's of trash and like I wrote and you probably noticed, this is a sign of failed (or at least incomplete) garbage collection. The 2nd VSS should be a copy of all working entries of the 1st VSS and not empty.

When empty, it's usually a sure sign of failed GC.

View attachment 1871669

I guess my basic question is what can I do to restore/reconstruct the ROM, at least to the degree possible. You mentioned BootROM reconstruction. How can this be accomplished? Once done, I will ABSOLUTLEY follow your advice and create a dump file that can then be periodically flashed back onto the ROM; that struck me as extraordinarily good counsel.

Again, tsialex, thank you so much for sharing your knowledge, time and energy. It is very much appreciated.

Check your PMs, I've sent you the instructions on what/how to get everything needed for a BootROM reconstruction.

 

[Valdaquendë]

tsialex - It looks like my request to remove the serial numbers (as you suggested) fell on deaf ears. I will remember to omit such information in future projects, though.

This week has been incredibly busy and I haven't gotten to flashing the BootROM but I hope to this weekend. I thank you again. I also gave a cursory read to your article on BootROM reconstruction; I want to go over this carefully when time permits and experiment with the tools until I understand them and can perform the operations correctly. Thanks for so much excellent information.

I tested the CPUs by installing them (one at a time, of course, as the swap-out processor tray is a single-CPU tray) and booting up. In both cases, the system booted properly and seemed to run perfectly normally. I ran TechTool Pro on the system with each CPU installed; it found no fault, so I am presuming that the CPUs are OK. If there are other diagnostics I should consider, please don't hesitate to advise me.

So the processor board appears to be the defective component here. I'll be looking for a replacement dual-processor CPU board this weekend and will update this thread with results.

Have a great weekend.

 

[tsialex]

tsialex - It looks like my request to remove the serial numbers (as you suggested) fell on deaf ears. I will remember to omit such information in future projects, though.

It's removed, at least I can't see it anymore, I even searched the page for the SSN and MLB.

This week has been incredibly busy and I haven't gotten to flashing the BootROM but I hope to this weekend. I thank you again. I also gave a cursory read to your article on BootROM reconstruction; I want to go over this carefully when time permits and experiment with the tools until I understand them and can perform the operations correctly. Thanks for so much excellent information.

I'd flash the never booted image before anything. Get your Mac Pro working correctly before anything.

I tested the CPUs by installing them (one at a time, of course, as the swap-out processor tray is a single-CPU tray) and booting up. In both cases, the system booted properly and seemed to run perfectly normally. I ran TechTool Pro on the system with each CPU installed; it found no fault, so I am presuming that the CPUs are OK. If there are other diagnostics I should consider, please don't hesitate to advise me.

Run AHT and ASD with the AppleOEM GPU.

AHT, you'll need 022-4831-A.dmg:

GitHub - upekkha/AppleHardwareTest: Apple Hardware Test Download Links

Apple Hardware Test Download Links. Contribute to upekkha/AppleHardwareTest development by creating an account on GitHub.

github.com

ASD, you'll need 3S149:

Information Architect Blog

As an information architect, these are a few tools, articles, and miscellanea that I find useful.

www.loremipsum.info

So the processor board appears to be the defective component here. I'll be looking for a replacement dual-processor CPU board this weekend and will update this thread with results.

Have a great weekend.

Did you inspected the CPU tray? Corrosion, bent pins, missing/damaged capacitors?

I'd do a very through microscope inspection before spending so much money on a mid-2010/mid-2012 CPU tray.

Pay attention to the underside, see if there are any damage near the northbrige push pins, people trying to replace the push-pins or re-apply thermal paste frequently damage the SMD components right next the holes for the push pins. I repaired several boards that were user damaged by pliers that rip inductors/capacitors/resistors.

 

[Valdaquendë]

Thanks; I have a USB-connect microscope and will inspect the board very closely and post photos of anything that looks suspect. Saturdays are "bookkeeping day" (ugh) in my office but I'll try to get to that this afternoon.

By "push pins", I assume you are referring to the pins that hold down the Northbridge heatsink?