MP 1,1-5,1 Mac Pro 4,1-5,1 stuck on gray screen

[Shiunbird]

Hello everyone,

I've been running a very stable setup for years. My 4,1-5,1 Mac Pro has a MacVidCards Vega 64, 64GB of ECC RAM in triple channel and boots to opencore on nvme. There I sometimes choose to boot to Windows 10 on SATA. The nvme has Mojave.

Everything was working fine for a long while. A few days ago, it only gave me a straight gray screen after bong. I turned it off and on and it booted fine.

This happened today again but turning it off and on still got me stuck on the gray screen.

I replaced the Vega 64 with a flash RX 580 with the same result. My RX 560 stays blank.

So I tried removing the nvme to see if it would go straight to Windows, but I don't think it will just boot like that (I changed my windows to EFI after I started using coreboot). I tried also keeping only the nvme (it blinks a bit before gray screen comes up, but then stops). I tried different outputs on the cards (always one of the monitors goes gray).

Any tips? I don't have another Mac at the moment to build a mojave bootable USB. I tried booting to EFI Debian USB, no improvement.

Thanks, as always.

 

[tsialex]

From what you wrote above, you have a 99,5% of chance of BootROM corruption.

Try doing a multiple/consecutive NVRAM reset. Install a wired KB, power up, press CMD-Option-P-R and keep it pressed until you hear at least five chimes. This is extremely common with cross-flashed early-2009 Mac Pros, because of the differences between the MP4,1 factory installed NVRAM volume and what the MP5,1 firmware expect the NVRAM to be.

Also, for the other 0,5%, disconnect all SATA devices and test one by one. A defective SATA device can in rare cases cause the same issue.

Four things to make it easier for you to diagnose your Mac Pro:

1. Double check the voltage of the BR2032 RTC battery with a voltmeter/multimeter, if the battery is below 3.00V you need to replace it. When the RTC battery have a low voltage, the RTC time related counters will be stuck or with invalid values and the Mac Pro behaves completely crazily - it's the first thing you should check.
2. After checking the RTC battery voltage, remove all DIMMs installed except one that should be the DIMM installed in the DIMM slot 1, this makes it easier for the PEI initialize the system with a corrupt NVRAM or defective SPI flash memory.
3. If you have a defective SPI flash memory with dead NAND cells where the NVRAM volume is stored inside the Mac Pro BootROM, disconnect the mains power cable from the PSU, remove the BR2032 RTC battery, wait at least a minute, connect the mains power cable to the PSU again, power on and check if you can boot - with the battery removed the PEI will bypass the NVRAM VSS stores and power on with a fail safe mode. You can't work with the Mac Pro like that, but you will be certain about what is the defect.
4. Defective SATA devices also include the ODD drives, disconnect everything.

 

[tsialex]

Btw, coreboot is not the same thing of OpenCore. I bet that you are mistaking one by the other since, AFAIK, coreboot does not support Mac Pros.

 

[Shiunbird]

All right - that's interesting.

Thank you so much!!!

I did God knows how many CMD OPT P R and I tried first with RAM on slot 5 and on slot 1. Not even a gray screen. It booted straight into windows bypassing the nvme and opencore.

From here, what's the best course of action? Can I re-add my RAM? I need Windows this weekend, so I won't mess around too much.

Interesting how it goes bad over time and so on.

PS: I just tried adding my RAM back and it boots to gray. Should it bong with bad RAM?

 

[Shiunbird]

OpenCore

I use coreboot on a thinkpad so I brainfarted.

 

[Dayo]

I bet that you are mistaking one by the other.

Seems the OP meant OpenCore and apparently tried to set things up to boot UEFI Windows directly after having issues. Most likely now has Windows Certs in the NVRAM ... more so after resetting NVRAM and getting booted directly into Windows.

 

[tsialex]

I did God knows how many CMD OPT P R and I tried first with RAM on slot 5 and on slot 1. Not even a gray screen. It booted straight into windows bypassing the nvme and opencore.

This is not good, besides the cross-flashing mess that you already had, now you have Windows SecureBoot signing inside the VSS store and a even more messed up circular log.

Never diagnose a Mac Pro with a Windows UEFI install disk present, you can inadvertently boot the UEFI Windows install without OpenCore protection and have a even more messier problem to solve.

From here, what's the best course of action? Can I re-add my RAM?

No, you didn't solved anything yet - your Mac Pro booted again because of the minimal config.

Like I wrote from the start, you probably have a SPI flash memory with defective/stuck NAND sectors, or at least a corrupt NVRAM volume.

Don't forget to check the RTC battery voltage.

Should it bong with bad RAM?

Depends, there are different RAM failure modes and ECC sometimes mask some failures.

Your problem seems to be more related to the MemoryConfig entries inside the VSS store and not a defective DIMM itself - obviously you can have a defective DIMM, but right now it's much more probable that it's a NVRAM volume corruption or SPI flash memory failure than a defetive DIMM.

Since you can boot again with the minimal RAM installed, remove your Windows disk and try to boot macOS to dump your BootROM image - you will need a urgent BootROM reconstruction service.

 

[Shiunbird]

Wow - thanks for linking me to that thread. I spent some time through it but I'm not nearly done.
I run a bunch of legacy hardware and my 4,1 is actually my daily driver - still runs FS2020 at 30fps with excellent graphics, so why bother upgrading? Plenty RAM for the VMs I need for work.

So I'm usually very careful at taking preventative steps, but I honestly had NEVER heard of the issue. I got this Mac Pro used 4 years ago from a UK reseller, with 1 year warranty, and I got a MacVidCards RX 580, then the Vega 64. But I wanted H264 acceleration, so I got coreboot and converted my Windows installation. I work on Windows nowadays (employer demand), so I boot Mac OS only for my personal things.

Had I known, I'd have take a dump after getting the machine and followed whatever preventative steps required (still pretty lost at the thread, and after a week of porting bash to powershell, pretty brain dead).

The Windows disk is out, I booted into Mojave (via opencore) on the nvme and the system is up and fine. What are the necessary steps for me to confirm that I just have corrupt NVRAM and not a flash failure? I saw some screenshots on the thread but got very confused. Should I still take the rom dump or it is messed up because of my Windows boot?

During my troubleshooting, I followed the standard steps before coming here - remove disks, change GPU, remove extra cards, but it never crossed my mind that the mess would be so deep. Especially since I haven't had a software update or any settings change in mojave in years, and I don't even have a wifi/bt card or anything in the sort of settings changes.

I'll leave my computer on just in case until I get your reply.

Thank you so much! I hope I dodged a bullet.

PS: It wasn't a windows install disk, but my daily windows installation, in case it makes any difference. The system just booted straight to it after clearing pram, maybe because it is in the sata bus and maybe it favours sata over nvme?

 

[tsialex]

Wow - thanks for linking me to that thread. I spent some time through it but I'm not nearly done.
I run a bunch of legacy hardware and my 4,1 is actually my daily driver - still runs FS2020 at 30fps with excellent graphics, so why bother upgrading? Plenty RAM for the VMs I need for work.

So I'm usually very careful at taking preventative steps, but I honestly had NEVER heard of the issue. I got this Mac Pro used 4 years ago from a UK reseller, with 1 year warranty, and I got a MacVidCards RX 580, then the Vega 64. But I wanted H264 acceleration, so I got coreboot and converted my Windows installation. I work on Windows nowadays (employer demand), so I boot Mac OS only for my personal things.

Had I known, I'd have take a dump after getting the machine and followed whatever preventative steps required (still pretty lost at the thread, and after a week of porting bash to powershell, pretty brain dead).

The Windows disk is out, I booted into Mojave (via opencore) on the nvme and the system is up and fine. What are the necessary steps for me to confirm that I just have corrupt NVRAM and not a flash failure?

This seems very complicated to someone that is not following the BootROM issues of MacPro5,1, but the easiest way to confirm NAND flash failure is to inspect the BootROM image dump and look at the VSS stores area.

Usually if your dump have a block of 0xFFs right in the middle of the circular log, the NAND cells failed.

I saw some screenshots on the thread but got very confused. Should I still take the rom dump

Yes, you will need to dump the BootROM image.

or it is messed up because of my Windows boot?

Windows UEFI SecureBoot signing already signed your BootROM, this is done the moment you booted your UEFI Windows for the first time you booted without OC.

During my troubleshooting, I followed the standard steps before coming here - remove disks, change GPU, remove extra cards, but it never crossed my mind that the mess would be so deep. Especially since I haven't had a software update or any settings change in mojave in years, and I don't even have a wifi/bt card or anything in the sort of settings changes.

I'll leave my computer on just in case until I get your reply.

Thank you so much! I hope I dodged a bullet.

You still can boot, so much easier to repair than a brick where you have to desolder the SPI flash memory to start the repair. I'm gonna send you a PM with instructions, follow it to the letter and I'll diagnose your BootROM image dump.

 

[Shiunbird]

Thanks!
On Sunday eve, I will uninstall OC and follow your instructions.

My Mac had been working so reliably that I haven't browsed anything about it - not much time for the Internets lately. =(