MP 1,1-5,1 MacPro5,1: BootROM thread | 144.0.0.0.0

Synchro3 · Jan 23, 2019

I think the beginning of the firmware issue is always booting Windows in EFI mode.

I have to correct my statement in this post: https://forums.macrumors.com/threads/mp5-1-bootrom-thread-140-0-0-0-0.2132317/page-75#post-26661511

Because later I remembered I tried years ago indeed booting once a Windows 7 installer DVD in EFI mode (the Mac crashed and I did never install Windows in EFI mode). That alone was enough to corrupt my ROM, I guess.

MrCoBalt · Jan 23, 2019

tsialex said:
I did a BootROM reconstruction a month ago for the early2009 Mac Pro, so probably no NVRAM problems. I'll ask about Windows.

tsialex said:
…another is a November 2009 made MP4,1, I have the serials.

Pretty sure this is referring to my system (a work spare, not mission critical at least!)

Windows has never been installed on this system; it was upgraded using the netkas 4,1->5,1 method a couple years back and was chugging along nicely (no upgrades other than an RX 560).

I upgraded it myself to 10.14.2 w/ 140.0.0.0.0 BootROM via the official 10.14.2 installer, then dumped the ROM and sent to Alex for review/rebuild. I flashed the rebuilt ROM to the system around Dec 10th with no issues, and everything was dandy; The system was on 24/7 but mainly unused since that date.

Sometime on around 1 AM on the morning of Jan 19th (according to our system monitoring) the system silently crashed and when I was next in the office my cursory troubleshooting revealed that on reboot there's no chime and the "GPU OK" diag light constantly flashes (the "EFI OK" light stays solid green though). Removing the RX 560 results in the same behaviour.

The system is sitting totally unplugged, GPU removed, etc. beside me but I haven't had a chance to put it back together and see if "having a rest" has made it any happier………

Reindeer_Games · Jan 23, 2019

Synchro3 said:
I think the beginning of the firmware issue is always booting Windows in EFI mode..

MrCoBalt said:
Windows has never been installed on this system...

My system only uses legacy Bootcamp booting so no EFI use on my end ever on my machine since Jan 2015 for sure when it was first upgraded from 4,1-5,1. No crashes-but scary behavior when Mojave gets booted to it. I have noticed it's fine if I hold Option to boot every time-I believe it may be because so many use third party boot option that auto trigger these boot menu's and there may be a bug related to when the Apple Boot-loader is not being loaded/auto-loaded (just a thought). When not holding Option but letting the system auto-boot I've experienced :

1) No or low tone at boot.
2) Wrong OS drive booting when changing OS's on warm boot.
3) No change to another OS drive when selecting a separate OS drive on warm boot.
4) Recovery loading on occasional cold boot to NVMe OS (this is still present, but far less frequency).
5) Odd fan spin-up's for no apparent load or temp's that should be requiring it.

Sometimes multiple NVRAM's as well as SMC clears were required before I got fan and tone behavior to re-appear. I also had an old power-strip fail that made me think my MP had failed for a few minutes after coming home two-weeks ago. I tore down and cleaned PSU just for good measure.

I'm capping the TDP of the GPU w/ Afterburner for now at 65% in Bootcamp under User Mode just as a precaution and and considering using Kernal Mode on it to keep Mojave in check (not really sure if that would work anyhow), but I've been issue free by simply avoiding Mojave. My issues have only been symptoms so far though, thankfully. Maybe something in there might help MrCoBalt get it going again-hopefully its booting to recovery and you just aren't able to view it with the RX 560.

I will note the Service Manual says the GPU OK light should illuminate-not flash, when using the backplane LED test circuit.

Physalia · Jan 23, 2019

tsialex said:
I’ll PM instructions.

tsialex,

Another newbie here, any chance you could please send these instructions to me as well?

5,1 mid 2012
Mojave & 140.0.0.0.0

Follow-up questions if you don’t mind:

1) Is there a forum here that can check the bootROM for me or can you point me in the right direction on how to interpret (check) it.
2) I believe I know the answer, but just to make sure, using W10 as a VM with ‘Parallels’ plus all updates from W10 shouldn’t have an impact on my cMP, right? I can also expect the VM to run as always (or better) with the
SSD7101-A setup. Do you foresee any issues?

Thanks for the help and knowledge you share.

tsialex · Jan 24, 2019

Physalia said:
tsialex,

Another newbie here, any chance you could please send these instructions to me as well?

5,1 mid 2012
Mojave & 140.0.0.0.0

Follow-up questions if you don’t mind:

1) Is there a forum here that can check the bootROM for me or can you point me in the right direction on how to interpret (check) it.
2) I believe I know the answer, but just to make sure, using W10 as a VM with ‘Parallels’ plus all updates from W10 shouldn’t have an impact on my cMP, right? I can also expect the VM to run as always (or better) with the
SSD7101-A setup. Do you foresee any issues?

Thanks for the help and knowledge you share.

Instructions sent by PM.

When you run a VM, the virtualised Windows 10 don't access or sign the real BootROM, so no problem with SecureBoot certificates.

You can't make a real Windows install into PCIe blades unless you use VMware Fusion to create it using a non-standard install process. For a virtualised one, makes no difference.

Squuiid · Jan 24, 2019

tsialex said:
You can't make a real Windows install into PCIe blades unless you use VMware Fusion to create it using a non-standard install process.

Huh? Since when?

tsialex · Jan 24, 2019

Squuiid said:
Huh? Since when?

Ok, let me rephrase: You can't install BootCamp/CSM into a PCIe blade.

Physalia · Jan 24, 2019

tsialex said:
Instructions sent by PM.

When you run a VM, the virtualised Windows 10 don't access or sign the real BootROM, so no problem with SecureBoot certificates.

You can't make a real Windows install into PCIe blades unless you use VMware Fusion to create it using a non-standard install process. For a virtualised one, makes no difference.

Thanks tsialex! I’ll be on it this weekend.

Squuiid · Jan 24, 2019

tsialex said:
Ok, let me rephrase: You can't install BootCamp/CSM into a PCIe blade.

Ah, I did not know that but that actually makes sense.
Another reason to install in EFI mode

.

cococheaf · Jan 24, 2019

tsialex said:
Ok, let me rephrase: You can't install BootCamp/CSM into a PCIe blade.

Alexandre, I don't want to question your knowledge, but the statement is not quite correct. it's true that booting nativlely from nvme disks in csm mode doesn't work. but I had installed windows 10 1709 in CSM mode on a 970evo, using a workaround.
booting was possible because i installed the windows bootloader on a USB stick (via windows setup dvd). every thing worked quite well, the only thing that didn't work was the boot disk selection via the system settings and no graphics display with nvidia cards that had an MVC flash.

tsialex · Jan 24, 2019

cococheaf said:
Alexandre, I don't want to question your knowledge, but the statement is not quite correct. it's true that booting nativlely from nvme disks in csm mode doesn't work. but I had installed windows 10 1709 in CSM mode on a 970evo, using a workaround.
booting was possible because i installed the windows bootloader on a USB stick (via windows setup dvd). every thing worked quite well, the only thing that didn't work was the boot disk selection via the system settings and no graphics display with nvidia cards that had an MVC flash.

Translated with www.DeepL.com/Translator

Let's be reasonable, you can't do a normal install of BootCamp/CSM mode into a PCIe blade.

You can workaround the external drive limitation with some exotic installs, but you can't do it with a plain install.

tsialex · Jan 24, 2019

Apple released 10.14.4 DP1. Same 140.0.0.0.0 BootROM.

Code:

$IBIOSI$ MP51.88Z.F000.B00.1809191555
Apple ROM Version
  Model:        MP51
  EFI Version:  140.0.0.0.0
  Date:         Wed Sep 19 15:49:52 2018
  Build Type:   Release

kdeda · Jan 24, 2019

tsialex said:
20181030: 10.14.1

Today Apple released 10.14.1 final, it's the same 140.0.0.0.0 BootROM released with 10.14.1 DP3. (10.14.2 still has the same 140.0.0.0.0 BootROM)

The new Mac Pro BootROM with NVMe boot support is now released for everyone, but remember that Mac Pro 5,1 don't install firmwares automatically with updates like every other Mac.

Code:

$IBIOSI$ MP51.88Z.F000.B00.1809191555 Apple ROM Version Model: MP51 EFI Version: 140.0.0.0.0 Date: Wed Sep 19 15:49:52 2018 Build Type: Release

You have three ways to install 140.0.0.0.0 (updated to 10.14.2):

install 10.14.2 from scratch into another drive,

download and open Mac App Store full installer for 10.14.2 (~6GB download), the installer will ask you to perform the upgrade,

manually trigger the install with 10.14.2 RecoveryHDMetaDmg.pkg. Instructions in this post, use the current RecoveryHDMetaDmg.pkg (~450MB download).

Note 1:

You can't use the createinstallmedia pen drive to upgrade the firmware, you have to do it from macOS.

Note 2:

If you have a Metal supported GPU and you open the full Mac App Store installer and the installer tells you that your GPU don't have Metal, please report this bug to Apple. It's a bug with the detection of Metal support that the macOS installer has with some HD 79xx and RX-580 GPUs.

Please report to it to Apple: Submitting Bugs and Feedback

Note 3:

If you can't install 140.0.0.0.0 because of the Metal detection bug with some HD 79xx and RX-580 GPUs, boot 10.13.6 and run the full Mac App Store 10.14.2 installer from there. Worked for most people.

140.0.0.0.0 and previous macOS releases note:

Since a lot of people ask this question: firmware 140.0.0.0.0 can boot even 10.6.8, but only if your GPU has drivers for it. With RX-4xx/RX-5xx, you are limited to 10.12.6/10.13/10.14, per example.

NVMe and previous macOS support:

Apple started supporting NVMe protocol with Sierra, but only for blades with 4 Kbytes per sector. This seems counterintuitive since 512 Kbytes per sector are older, but Apple only supported their own NVMe drives and those are 4 Kbytes per sector. With High Sierra, Apple supports both 4 Kbytes and 512 bytes per sector blades.

So, to use your NVMe drive you are limited to Sierra and newer macOS versions for drives with 4 Kbytes per sector and only to High Sierra and Mojave for NVMe drives with 512 bytes per sector.

Most blades on the market are 512 bytes per sector, all Samsung OEM and consumer blades are 512 bytes per sector and usually only NVMe drives made for the datacenter market are 4 Kbytes per sector with the exception of some rare Toshiba and Intel blades that are 4 Kbytes too.

20181008: BootROM 140.0.0.0.0

Apple released 10.14.1 DP3 and 140.0.0.0.0 Mac Pro BootROM.

NVMe EFI module added and it's different than the one from MP61,

PCIe drives still recognised as external,

Same 29 & 31 microcodes from May,

No boot screens for UEFI cards.

Code:

$IBIOSI$ MP51.88Z.F000.B00.1809191555 Apple ROM Version Model: MP51 EFI Version: 140.0.0.0.0 Date: Wed Sep 19 15:49:52 2018 Build Type: Release

NVMe EFI module already injected:

View attachment 793233

First post about 140.0.0.0.0 #1455. You can read posts #1554 and #1569 to know how to install 140.0.0.0.0 manually or just wait for the final release of 10.14.1.

20180925: BootROM 139.0.0.0.0

Apple released 10.14.1 DP1 and 139.0.0.0.0 Mac Pro BootROM.

No boot screens for UEFI cards,

no NVMe EFI module added,

no FileVault2,

same microcodes,

still safe to insert the NVMe EFI module.

Code:

$IBIOSI$ MP51.88Z.F000.B00.1808171030 Apple ROM Version Model: MP51 EFI Version: 139.0.0.0.0 Date: Fri Aug 17 10:24:21 2018 Build Type: Release

First post about 139.0.0.0.0 #1242

20180813: BootROM 138.0.0.0.0
New BootROM 138.0.0.0.0 for Mac Pro 5,1 with updated microcodes (0x106A5 29 and 0x206C2 31), support for 5GT/s link speed on macOS/Windows without R17 mod, enabled HDMI audio on RX480 and new/improved firmware flasher on Mojave DP7/PB6 (18A365a) full installer.

Now seems you don't need a Mac EFI GPU to flash it, but the Mojave installer checks for a Metal supported one before upgrading the BootROM. @Squuiid confirmed flashing with a RX580, sadly without boot screens/boot selector.

Note:

NVIDIA GPUs have a default mode of 2.5GT/s, only when in 3D use they go to 5GT/s, it's a power saving feature. Only eVGA GTX 680 Mac Edition and GPUs flashed with the same firmware show 5GT/s all the time into System Information.

Attention:

The new Firmware Flasher don't have the usual progress bar, just a grey screen on Mac EFI GPUs, or a black one with a PC GPU, and the usual DVD tray opening. Since the new flashing process don't have progress indicator/completion and seems to take longer to complete and shutdown the Mac Pro, you have to wait until the shutdown/restart.

View attachment 775772

Updated microcodes:

View attachment 775777

5GT/s link speed with a AMD HD 7870 without R17 mod:

View attachment 775822

BIOS version:

The complete BIOS version for 138.0.0.0.0 is MP51.88Z.F000.B00.1807300628

NVMe support:

Apple did not add the NVMe EFI module with 138.0.0.0.0, but the injection method is the same as was done with High Sierra - no changes.

What I have to do to upgrade the firmware:

If your Mac Pro have a BootROM earlier than MP51.0089.B00, to upgrade to 138.0.0.0.0 you will have to do:

Download the full Mac App Store installer for 10.13.6, open it and then do as the installer says to upgrade to MP51.0089.B00. After the firmware upgrade reboot, you can close the installer app.

Download the full Mac App Store installer for Mojave, the most recent one, open it and then do as the installer says to upgrade to 138.0.0.0.0. After the firmware upgrade reboot, you can close the installer app.

Mac Pro 5,1 recent firmware releases:

10.13 DP5 with MP51.0083.B00 with initial APFS support
10.13 DP6 with MP51.0084.B00 with APFS support
10.13.4 with MP51.085.B00 (Mojave DP1/DP2/PB1/DP3/PB2 too)
10.13.5 with MP51.087.B00 missing the Intel Xeon microcodes
10.13.6 with MP51.089.B00 updating to the Spectre mitigated microcodes on the April 2 Microcode Update Guidance.
10.14 DP7/PB6 with 138.0.0.0.0 with 5GT/s and new microcodes
10.14.1 DP1 with 139.0.0.0.0
10.14.1 DP3 with 140.0.0.0.0 with NVMe support

First of all Thank you so much for the tips. I have extend the life to my dear mac pro 5.1 from 2010.
I had been using a raid 0 with 4 Samsung951 m2 modules in amfeltec
They got hot and after a few years the write speed was pretty much gone. (800mb/s, down from like 4.5Gbs)
I had cloned HighSierra into a raid 0 with 2 Samsung 970Pro (under amfeltec) and it would not reboot.
Did some research and found this, To get the new firmware, i simply followed step 2. Updated the firmware and my raid 0 boot volume showed up and here i am with what will be a sweet 5GBs read write file system.
Thank you again.

download and open Mac App Store full installer for 10.14.2 (~6GB download), the installer will ask you to perform the upgrade,

cococheaf · Jan 25, 2019

tsialex said:
View attachment 817714 View attachment 817715 View attachment 817716 View attachment 817717 View attachment 817718 View attachment 817719 View attachment 817720 View attachment 817721 View attachment 817722 View attachment 817723

Let's be reasonable, you can't do a normal install of BootCamp/CSM mode into a PCIe blade.

You can workaround the external drive limitation with some exotic installs, but you can't do it with a plain install.

true.

MrCoBalt · Jan 25, 2019

MrCoBalt said:
Sometime on around 1 AM on the morning of Jan 19th (according to our system monitoring) the system silently crashed and when I was next in the office my cursory troubleshooting revealed that on reboot there's no chime and the "GPU OK" diag light constantly flashes (the "EFI OK" light stays solid green though). Removing the RX 560 results in the same behaviour.

The system is sitting totally unplugged, GPU removed, etc. beside me but I haven't had a chance to put it back together and see if "having a rest" has made it any happier………

So an update on this: I brought in my personal known-good single-CPU 4,1 (previously flashed with Alex's rebuilt 140.0.0.0.0 and running nicely) and swapped some components around.

Long story short: not the backplane ROM, but rather appears to be a popped rivet on the northbridge of the work system's dual CPU tray. Regardless of backplane/GPU/RAM/etc in use the system shows the exact same behaviour (the flashing "GPU OK" light and failure to boot) when the dual-CPU tray is used.

This led me to find this Apple discussion thread which definitely sounds like my issue. I will have to see if the board is salvageable when I have a bit more time…

flehman · Jan 25, 2019

MrCoBalt said:
So an update on this: I brought in my personal known-good single-CPU 4,1 (previously flashed with Alex's rebuilt 140.0.0.0.0 and running nicely) and swapped some components around.

Long story short: not the backplane ROM, but rather appears to be a popped rivet on the northbridge of the work system's dual CPU tray. Regardless of backplane/GPU/RAM/etc in use the system shows the exact same behaviour (the flashing "GPU OK" light and failure to boot) when the dual-CPU tray is used.

This led me to find this Apple discussion thread which definitely sounds like my issue. I will have to see if the board is salvageable when I have a bit more time…

If it’s Northbridge rivets you could remove the CPU tray and verify that visually. Mine broke late last year in my 4,1->5,1 and the pieces were still laying on the tray when I removed it and you could see the heatsink was heaved upward on one side. I was actually sitting at my computer when the rivet went and it was LOUD, made me jump out of my chair when it happened.

Anyway, if that is the case it is easy to verify and fix - check out the huge thread elsewhere in this forum on sourcing replacement rivets.

marioliv66@ · Jan 26, 2019

hello, I just published this post, would there be a report with version 140.0.0.0.0?

https://forums.macrumors.com/thread...hci-invisible-since-ssd-sata-10-12-6.2166971/

tsialex · Jan 29, 2019

Today I was doing a BootROM clean-up for a user here and totally accidentally found that the BootROM stores what was the original EFI version, from factory, at the address 0x3F0064.

The BIOS version from the image below corresponds to MP51.007F.B03:

Checked various dumps and all of them have the same info at the same address. It's a totally useless information, but interesting.

Another thing that I found, the process of MP4,1 to MP5,1 conversion overwrites the factory EFI version to the one from MP51.007F.B00, so every MP4,1 that was upgraded via Netkas method lost the original EFI version.

One useful application of this info, for me at least, is to check if the BootROM was previously reconstructed using UEFITool method of extracting NVRAM + LBSN_BD sector. If the original EFI version is newer than MP51.007F.B03, it's a reconstructed one.

Kon_Kipa · Jan 29, 2019

tsialex said:
Today I was doing a BootROM clean-up for a user here and totally accidentally found that the BootROM stores what was the original EFI version, from factory, at the address 0x3F0064.

The BIOS version from the image below corresponds to MP51.007F.B03:

View attachment 818885

Checked various dumps and all of them have the same info at the same address. It's a totally useless information, but interesting.

Another thing that I found, the process of MP4,1 to MP5,1 conversion overwrites the factory EFI version to the one from MP51.007F.B00, so every MP4,1 that was upgraded via Netkas method lost the original EFI version.

One useful application of this info, for me at least, is to check if the BootROM was previously reconstructed using UEFITool method of extracting NVRAM + LBSN_BD sector. If the original EFI version is newer than MP51.007F.B03, it's a reconstructed one.

so if i am going to upgrade my 4,1 --> 5,1 using Netkas upgrading tool will my machine then be able to be upgraded to 140.0.0.0 using your dual step method?

tsialex · Jan 29, 2019

Kon_Kipa said:
so if i am going to upgrade my 4,1 --> 5,1 using Netkas upgrading tool will my machine then be able to be upgraded to 140.0.0.0 using your dual step method?

Yes.

Kon_Kipa · Jan 29, 2019

tsialex said:
Yes.

so the 4,1 although physically is a 2009 model will essentially become a 5,1 2010/2012 model (meaning all the hardware - apart from original graphics, ram & CPU's) will act & function as any original 5,1

tsialex · Jan 29, 2019

Kon_Kipa said:
so the 4,1 although physically is a 2009 model will essentially become a 5,1 2010/2012 model (meaning all the hardware - apart from original graphics, ram & CPU's) will act & function as any original 5,1

From the software point of view, yes, but not from the hardware. SMC version can't be upgraded and you can't use mismatched trays/backplanes.

2009 - 1.39f5
2010/2012 - 1.39f11

Kon_Kipa · Jan 29, 2019

tsialex said:
From the software point of view, yes, but not from the hardware. SMC version can't be upgraded and you can't use mismatched trays/backplanes.

2009 - 1.39f5
2010/2012 - 1.39f11

Yup, i've worked out that if you use the 2009 tray in a 2010/2012 the fans run at full speed...

I never bothered to find out what would happen if you put a tray from a 2010/2012 into a 2009 model (but i'm sure it would be something similar)
[doublepost=1548798536][/doublepost]

tsialex said:
From the software point of view, yes, but not from the hardware. SMC version can't be upgraded and you can't use mismatched trays/backplanes.

2009 - 1.39f5
2010/2012 - 1.39f11

are you able to change using a soldering station an SMC from a 2010/2012 model onto a 2009 model MLB? - what i am asking is will it work if one was to do so?

Also is it worth the effort to do so? meaning are there any advantages?

tsialex · Jan 29, 2019

Kon_Kipa said:
are you able to change using a soldering station an SMC from a 2010/2012 model onto a 2009 model MLB? - what i am asking is will it work if one was to do so?

Also is it worth the effort to do so? meaning are there any advantages?

If you have all the equipment needed to replace that enormous TQFP-144, it's possible and already been done. It will probably cost more of supplies and equipment time than an eBay replacement board, tho.

Kon_Kipa · Jan 29, 2019

tsialex said:
If you have all the equipment needed to replace that enormous LQFP, it's possible and already been done. It will probably cost more of supplies and equipment time than an eBay replacement board, tho.

for the most i would say yes... but i mean i've got guys that are to me "geniuses" in that department...

So not factoring in the cost, would there be any advantages?

MP 1,1-5,1 MacPro5,1: BootROM thread | 144.0.0.0.0

macrumors 68000

macrumors newbie

macrumors 6502

macrumors newbie

Contributor

macrumors 68000

Contributor

macrumors newbie

macrumors 68000

macrumors regular

Contributor

Contributor

macrumors newbie

macrumors regular

macrumors newbie

macrumors 6502

macrumors member

Contributor

macrumors member

Contributor

macrumors member

Contributor

macrumors member

Contributor

macrumors member

Our Staff