MP5,1: Return of the microcodes (MP51.0089.B00) **OBSOLETE**

[tsialex]

Comparatively speaking, I still consider you to be in a Pro skillset category.

Since you know the exact chip that Apple used in these fragile little things, how many write cycles are they good for? I guess I should start avoiding the changes made by the startup disk manager, and only use the option boot picker. This may save me some write cycles.

I wish Apple used a bootloader like GRUB. It’s ugly, but at least it won’t brick your machine.

10,000 cycles was the endurance from the SPI flash of my dual MP5,1. I do not know what's the flash used on my single one, but I know it's not the same. My single, was produced on December 2009, my dual on July 2010 (a refurb store B08 one).

I really don't think that you need to save writes on the SPI flash, I suspect that it's a combination of factors that cause this, like high temperature, shutting down when the SPI flash is being write, etc.

I think it's more sane to keep a saved dump of your BootROM, keep your Mac Pro clean and with low temperatures - My Mac Pro was filthy when I removed my logic board and I frequently used compressed air to clean it. To this day, I have not seen a lot of reports of corrupted BootROMs on Mac Pros, like I've seen with MacBooks. It's really robust system, but Apple did not think that we were using 1080's on 40º days with a lot of dust bunnies on ours Mac Pros when they projected the 5,1. Don't let me start with the cursing of the 6,1 ones…

Try to not force shutdown, remove the power supply cable when your Mac is on, etc. Keep more than one backup of the BootROM dump.
[doublepost=1532415673][/doublepost]On the summer, winter now on the Southern Hemisphere, I'll put a thermopar on the flash and convert some videos to x265/run some benchs and check the SPI flash temperature. My Northbridge Tdiode is now on 65º, go to 75º/76º easily on the summer.

 

[crjackson2134]

Well, I have 2 backup copies, but I don’t know how that would help if my only Mac is dead.

My Mac innards are fairly clean. I do PMI on a regular basis, and keep the machine in a cool environment. I would guess my machine would have a decent amount of write cycles, it was assembled in 2013 when bought new. It would be nice to have a safetynet plan though.

 

[tsialex]

Well, I have 2 backup copies, but I don’t know how that would help if my only Mac is dead.

My Mac innards are fairly clean. I do PMI on a regular basis, and keep the machine in a cool environment. I would guess my machine would have a decent amount of write cycles, it was assembled in 2013 when bought new. It would be nice to have a safetynet plan though.

I have a PC motherboard made by Intel (DH61CR) that corrupted the SPI flash when updating the firmware, I have tried to use just the Pi+Pomona before and failed. Since I wasn't using it, went to low priority pile of things to repair. I was planning to remove it from the board like I did with the Mac Pro, but I'll test with the 3V3 on VDD from a ATX power supply this time.

 

[Synchro3]

Wow, could it be that it didn’t shutdown because it already failed during the process and locked there until you forced a power off? Hard to understand how cutting power alone would brick the thing.

Could be possible it failed before the shutdown.

@Cecco @Synchro3 @h9826790 what firmware version were you all running when your cMP failed?
Might it have been the same for all three of you?

Also, were you all 4,1–>5,1 cMPs?

A Mac Pro 4,1->5,1 with Firmware MP51.0087B00 (the one with no microcode)

I've repaired a lot of Macs from family/friends over the years and one of the things I encounter is corrupted BootROMs. Apple don't use a battery backed RAM to store configurations like on PCs, all settings are saved on the SPI Flash, so I suspect things got corrupted sometimes. Change your Wi-Fi password, Back to My Mac update the SPI Flash. Change your boot disk, SPI flash updated. Go to Windows, SPI Flash updated. Clear your NVRAM, SPI flash updated.

The SPI flash is not a NAND like we use today on SSDs, the endurance is a lot lower, some have 4,000, others 10,000 writes, military ones with 100,000. Working on the limit of the temperature, maybe some 4k sector writes (you have to write a sector to change one bit) get garbled, and other information is corrupted on the writing.

This is very interesting. I still have my old failed board, but I don't have the skills to change the SPI flash.

 

[Squuiid]

A Mac Pro 4,1->5,1 with Firmware MP51.0087B00 (the one with no microcode)

I have a feeling @h9826790 was the same. Not sure about @Cecco however.

Coincidence? Perhaps, perhaps not.

 

[tsialex]

I have a feeling @h9826790 was the same. Not sure about @Cecco however.

Coincidence? Perhaps, perhaps not.

Not to confirm your suspects since I did the damage, but my failed Mac Pro was on 0087 too.

 

[Squuiid]

Not to confirm your suspects since I did the damage, but my failed Mac Pro was on 0087 too.

@Cecco was also on a 4,1-->5,1. Not sure what firmware he had. @Synchro3 has now confirmed he was also.
That makes three (four if we count yours) cMPs that bricked in a very short and close period of time and all likely running a firmware version that was a) not for the specific machine and b) later pulled by Apple.

Again, nothing concrete here but it's starting to not look so much like coincidence.
@h9826790 was convinced that he did nothing wrong and is a very experienced cMP user. @Cecco the same.
There have been zero 2010 models bricking in this way in recent history as far as I can tell.

I have a 2009 5,1 and a 2010 and fortunately both are now on 0089 without any problems. Lucky perhaps? My 2009 was upgraded from 0084 to 0089, skipping 0087 entirely.

 

[Cecco]

Well didn't check the Firmware version before it died, but as I was on 10.13.5 and IMO 0089 only came with the 10.13.6 combo updater I assume I was on 0087.
Updated to High Sierra just about 3 weeks ago and had to do the GPU switch to be able to flash the mandatory firmware update for High Sierra.

So maybe 0087 is more prone to brick a Mac Pro. In that case Apple would be responsible for the damage in the first place by releasing a faulty firmware. But I hardly believe they refund the costs.

@tsialex: Concerning flashing the SPI without desoldering. Why not just connect the mobo (without any other components) to the MacPro power supply instead of a bench PSU? Would definitely have enough power to drive the mobo and probably prevent other components on the mobo that run on 5V (are there any ?), to get any unwanted backcharge through the 3.3 V line.

 

[Squuiid]

So maybe 0087 is more prone to brick a Mac Pro.

Only when on a 4,1-->5,1 perhaps?

In that case Apple would be responsible for the damage in the first place by releasing a faulty firmware. But I hardly believe they refund the costs.

Tricky, in that all of these bricked cMPs were 2009 4,1-->5,1 and not technically supported

 

[tsialex]

Concerning flashing the SPI without desoldering. Why not just connect the mobo (without any other components) to the MacPro power supply instead of a bench PSU? Would definitely have enough power to drive the mobo and probably prevent other components on the mobo that run on 5V (are there any), to get any unwanted backcharge through the 3.3 V line.

That's not gonna work, if anything else is running, you not gonna write the flash since the lines will be settled and things are gonna be accessed/blocked (#WP) when you're trying to flash it.

You have to only compensate for the drain caused by other components on the 3V3 line that feeds VDD and not power anything else.

I don't have the schematics for the Mac Pro, but maybe if we can get it, we'll find some resistor/cap that isolates the 3V3 line when removed or a test point/via to feed 3V3 to the logic board like the PP3V3_SUS on MacBooks.

 

[Squuiid]

Hmm, I have a 2009 5,1 here at work that is on 0087. Going to try and upgrade it to 0089 now and hope it goes well! Will report back.

UPDATE. It went fine. 2009 5,1 now on 0089, no problem. Glad to be off of 0087 for what it's worth.

 

[h9826790]

The more troublesome part is to get the board out of the Mac and remove the SPI flash, the programming part is simple.

I never did the on board flashing on a Mac Pro, but did several times on Mac minis upgrading from 1,1 to 2,1 firmware. With a 3V3 bench power supply to compensate the board components drain, like I suggested to @Cecco, you will get it done without having to use a hot air rework station, like I did on my Mac Pro.

Remove the board is easy for me. Remove the chip may be a harder job (I've never done that before).

@Cecco @Synchro3 @h9826790 what firmware version were you all running when your cMP failed?
Might it have been the same for all three of you?

Also, were you all 4,1–>5,1 cMPs?

Can't remember, but that was in mid May 2018. The firmware version was the latest official firmware by that time.

And yes, that's a 4,1 flashed to 5,1. The flash was done years ago.

 

[tsialex]

Can't remember, but that was in mid May 2018. The firmware version was the latest official firmware by that time.

10.13.4 was released on March 29, with 0087.

Remove the board is easy for me. Remove the chip may be a harder job (I've never done that before).

Wait for other people test if the corrupted BootROM is causing this, then you can try with a Pi+Pomona SOIC clip and feeding 3V3 on the VDD line, doing this will be the easiest and most cost effective.

You can use any Pi, even the Pi Zero if you get the version with the GPIO terminals already soldered, US$ 14, or solder the necessary ones yourself, US$ 5 then.

 

[Squuiid]

Can't remember, but that was in mid May 2018. The firmware version was the latest official firmware by that time.

And yes, that's a 4,1 flashed to 5,1. The flash was done years ago.

That makes 3 out of 3. All on 2009 5,1 and almost certainly running 0087.

Those of you with 2009 5,1 cMPs and still on 0087 I would move off of it and over to 0089 asap.

 

[Cecco]

I don't have the schematics for the Mac Pro, but maybe if we can get it, we'll find some resistor/cap that isolates the 3V3 line when removed or a test point/via to feed 3V3 to the logic board like the PP3V3_SUS on MacBooks.

Never seen schematics for the Mac Pro. Don't think they are out in the wild. So it will be likely trial and error with the onboard flashing.

 

[Synchro3]

That makes 3 out of 3. All on 2009 5,1 and almost certainly running 0087.

Those of you with 2009 5,1 cMPs and still on 0087 I would move off of it and over to 0089 asap.

Yes, I upgraded the new board directly to 0089, skipping 0087.

I still have the old board.

 

[eksu]

That makes 3 out of 3. All on 2009 5,1 and almost certainly running 0087.

Those of you with 2009 5,1 cMPs and still on 0087 I would move off of it and over to 0089 asap.

Glad I checked this thread — I’m on a ‘09 5,1 running Mojave and am still on 0087. (I figured I could use the performance gains more than the security on my mostly for gaming Mac Pro).

In light of the recent Intel ME HTTP parser vulnerability (that isn’t getting patched on any core series chips before Ivy Bridge), has anyone ran me_cleaner or disabled the management engine on their machine?

 

[tsialex]

Glad I checked this thread — I’m on a ‘09 5,1 running Mojave and am still on 0087.

In light of the recent Intel ME HTTP parser vulnerability (that isn’t getting patched on any core series chips before Ivy Bridge), has anyone ran me_cleaner or disabled the management engine on their machine?

If I'm not wrong, Mac Pro 4,1 and 5,1 Xeons don't have ME activated, VPro on Intel lingo, but 6,1 has.

You can dump the BootROM from one Mac Pro 4,1/5,1 and flash it to another, it'll work except when iCloud/Facetime/iMessage validates the UUID. On a Mac with ME activated it will not work.

 

[tsialex]

Off-topic:

Apple has just released
macOS High Sierra 10.13.6 Supplemental Update for MacBook Pro (2018). Seems that all youtube whine bring results fast with Apple.

 

[bookemdano]

Off-topic:

Apple has just released
macOS High Sierra 10.13.6 Supplemental Update for MacBook Pro (2018). Seems that all youtube whine bring results fast with Apple.

Especially with their newest flagship product!

 

[h9826790]

Finally has some time to perform the 10.14 PB3 update.

Microcode version 31 for W3690 with 0089.B00 BootROM.

It's definitely an OS level injection, because the microcode version in 10.13.6 on the same machine still 30.

 

[bookemdano]

Finally has some time to perform the 10.14 PB3 update.

Microcode version 31 for W3690 with 0089.B00 BootROM.
View attachment 772572
It's definitely an OS level injection, because the microcode version in 10.13.6 on the same machine still 30.

Excuse me if someone has already tested this and I missed it, but does the 10.14 PB3/DB4 OS microcode injection still occur successfully if one is on an old firmware revision (0087, 0085 or even older)?

 

[tsialex]

Excuse me if someone has already tested this and I missed it, but does the 10.14 PB3/DB4 OS microcode injection still occur successfully if one is on an old firmware revision (0087, 0085 or even older)?

Yes, works on 0085, I've checked before, others too.

 

[Squuiid]

Finally has some time to perform the 10.14 PB3 update.

Microcode version 31 for W3690 with 0089.B00 BootROM.
View attachment 772572
It's definitely an OS level injection, because the microcode version in 10.13.6 on the same machine still 30.

Hmm, I'm still on 30 with DB4.

machdep.cpu.microcode_version: 30

 

[JedNZ]

I’ve got two cMP - a (single) hexa-core and (dual) 12-core, each with X5680s, and both on HS 10.13.5 with FW 0085.B00. Is it safe/advised I upgrade both to 10.13.6 using the MAS full installer and FW 0089.B00, or wait?