Yes. The issue is related to the PCIe system of the Mac Pro. Different devices may have different effects, perhaps even positive effects. I think that's why some users are seeing a certain level of consistency.
This is weird, 100% success from a spinner, hangs installing on an ssd (both sata)...
I'm only testing SSDs, I'll take a look at a HDD - it wouldn't be too crazy that with the relaxed timings for a HDD the success rate improves.
If you were running OpenCore, you will still have the boot picker. I don't get the issue?
I alternated between Mojave and 11.3 3x, then the 4th time 11.3 hung, so that's a bust.
I'll be very interested to hear your results. My HDD is a CCC clone of my 11.2.3 primary system drive, that has been upgraded with every 11.3 beta through GM, and is now running 11.4b1. I never had an upgrade fail, and never had a boot hang, with (doing simple math) at LEAST 80x reboots throughout the 11.3 beta process. That's with the PCIe SSD installed (and eSAS HBA installed) in slots 3&4 (as I don't have easy physical access to that machine).
I have an 11.2.3 Parallels VM, and it won't tell me there's an update to 11.3, likely because the host is running 11.2.3. It's my theory, not researched, that Parallels won't let you upgrade beyond the host OS version. To test that, I created a new VM with an 11.2.3 installer. It too will not show there's an update to 11.3, so I cannot test 11.3 VM's on my primary system. Maybe I'll install a Parallels trial on my 11.4 test machine, install 11.2.3, and try upgrading to 11.3. That will take some (computer) time - not much of mine.
I run HighPoint 7101-A on cMP 3,1 with boot NVMe 970 Evo and mine DOES work.
Understand about the PCIe system. Just that this is first report I an seeing of issues when nothing is on the PCIe. I think previous reports had stated that it works fine when the PCIe slots are empty except for the GPU.
Correct. Looks like a race condition and some devices may slow something(s) down so that other condition(s) are more likely to be ready before being required or accelerate something(s) so that they are more likely to be ready before being required.
Actually, there is a point to ponder there as slower can be an advantage in a race condition (which seems to be the issue) if it means that Required Condition "A" is always ready before Dependent Condition "B". Good chance solution will be to slow something or the other in the 5,1 down.
EDIT 1: It was @JohnD:
forums.macrumors.com
Sorry to say this, but your theory is wrong. My Parallels Big Sur 11.2.3 virtual machine DID offer me the 11.3 update. All the while, my host operating system is 11.2.3. As I said, the VM update installation failed twice, but that is another issue. My third attempt consisted in downloading the 11.3 installer from the App Store. Then, I fed that installer to Parallels Desktop in order to create a new virtual machine. When that installation froze up, I restarted the new, half-installed virtual machine and the installation completed successfully this time.
Seems to support the switch hypothesis but yeah, we are in groping in the dark mode.
That should be "Any insufficiently comprehensible kernel panics are indistinguishable from magic.
It is not only on cMPs: https://forums.macrumors.com/thread...ter-because-of-a-problem-never-stops.2284418/
I think our best lead so far are machines that present some consistency. Working backwards from those machines to determine the cause of that consistency will give us valuable data.
Right, but that message itself does not halt the boot process nor it causes any panics during boot. In fast I have seen it on my fully BS supported MBP11,3 without any previous panic or restart. Just a normal boot or restart.
