Well, actually, the specifications should not matter in answering these questions...
... but I see that it happens on SB850 and SB950 southbridges.
And nobody—as far as I know—has ever been able to come up with a definitive solution, let alone a reasonable explanation as to why this happens.
I just simply want to know the root cause of this problem, so that a definitive solution can be found, either in the form of a working fix or an acceptable workaround, or an acknowledgement that, ultimately, it cannot be fixed.
I have been battling with it for over a week, and it is really crazy that other builds, based on an A88X or an X470 work like a charm (seriously; the FM2+ chipsets were said to be particularly finicky, yet they work like a charm, e.g. here with an AMD Athlon X4 845 in a MSI A88X-G41 PC Mate V2 with a Palit GeForce GTX 1050Ti KalmX, etc.), but this one is like an insurmountable obstacle:
GA-890FXA-UD5 rev. 3.1, BIOS version FEf,
KFA2 GeForce GTX 960 EXOC 4GB;
4 x 8 GiB Netlist NVD1G7235107H-D12T PC3-12800E;
LG GGC-H20L + Sony NEC Optiarc AD-7203A on an 80-wire ribbon cable to the JMicron JMB363 PATA connector;
Western Digital VelociRaptors WD1000DHTZ + WD6000HLHX connected to the SB850 6Gb/s SATA ports through the AHCI protocol;
Creative Labs SB0880 Sound Blaster X-Fi Titanium;
some add-in PCI-Express card with the Texas Instruments XIO2213B chip with 3 IEEE-1394b Firewire-800 ports;
some Mitsumi 3.5 inch 1.44MB floppy drive on a round cable comfortably tucked away.
At other times, a QLogic QLE2564 8Gb/s Fibre Channel host bus adapter or an QLogic QLA2462 PCI-X 4 Gb/s Fibre Channel host bus adapter or an Adaptec 39320A-R Ultra-320 SCSI Raid card, an SAS 9311-4i4e 12Gb/s host bus adapter (mostly with a Toshiba PX04SLB200 2 TB Read Intensive SSD), some NVMe Optane SSDs, etc.—all of these are mixed and matched depending on the application and storage needs and free slots on the mainboard.
These were all disconnected during the installation attempts. Likewise, all the non-essential onboard devices, such as the COM port and the LPT ports and the additional storage controllers as well as the Renesas USB 3.0 controller were disabled through the BIOS setup.
As you see, nothing extraordinary, a pretty normal configuration of an average legacy BIOS-based computer (with some EFI elements that I wish were not there: http://www.rodsbooks.com/gb-hybrid-efi/
Unfortunately, USB 2.0 controllers could not be disabled without disabling USB ports completely, because, as you can see, all the errors come from the three odd-numbered Enhanced Host Controllers (USB1, USB3, USB5) on the mainboard.
Probably most importantly, I used a PS/2 keyboard but a USB mouse (a Logitech RX-250 USB+PS/2 mouse) because this mainboard has only 1 PS/2 connector. This has no effect, as I also tried to run the installation with the mouse disconnected.
I also used various USB 2.0 and USB 3.0 sticks in USB 2.0 (EHCI) ports... to no avail.
I have not tried them in USB 3.0 ports, but I believe that it would be pointless, as the installer spews these errors for all the USB 2.0 ports, regardless of whether any devices are connected to them or not.
I was also contemplating doing it the old-school way, creating an installer image which could then be recorded on an optical medium, such as a DVD+DL or a BD-RE, so that I could disable the USB ports completely... but then again, I wonder how it would be with the onboard JMB363 PATA controller... the LG GGC-H20L BluRay-burner/HD-DVD player is connected through a JMicron JM20330 bridge adapter, so I could possibly take that out and connect it directly to the SB850 southbridge and set the last ports to IDE emulation mode... but still, I wonder how the installer would like it overall.
Due to fact that the optical drives are located in extremely cramped conditions (which is the main reason why the floppy drive is still connected
) and I would need to take it all apart, I would consider this to be my last-ditch attempt.
But that still would not be a solution and it would give no answers to the 2 above-mentioned questions.
Windows Server 2016 runs without a glitch on it. When I switch off AMD C1E support, Oracle Solaris, Gentoo Linux and DragonFly BSD run just as fine. Well, Oracle Solaris is not run very often and DragonFly BSD does report resetting some AHCI controllers at bootup, but it runs fine afterwards.
It is just MacOS X that chokes on this hardware so strangely and inexplicably.