i don’t know the specs well and have skimmed your post but, lazily throwing this out there, you do now there are only so many PCIe lanes the more stuff you add and fill slots up the slower they run and in come cases not all slots or CERTAIN slots can’t all be populated or rather if they are something won’t function.
ie if slot 1 AND 2 are populated slot 3 runs at slower speed or disables, this common for motherboards in gneeral at least in years past
In its standard configuration with all three slots populated, slot 6 and 4 share 16 lanes, getting 8 each through the MUX, and slot 5 has a dedicated 4 lanes. If slot 4 is left unpopulated, slot 6 gets all 16 lanes.
After going back and forth with Asrock support for a couple of weeks, it apparently turned out to be an issue of compatibility between the BIOS and CPU. The 1000 series CPUs aren’t officially supported, and my 1600AF was an edge case being a 1000 series CPU, but built on the Zen+ architecture, typically found on 2000 series CPUs. Their BIOS team gave me a custom BIOS, and all seems to be working now, although I’m still setting up and testing.
yeah man the 1600AF was hot shit flying off shelves when people found out AMD ran out chips start selling zen+ under the AF revision i actually used on on my asus x570 board until i could snag a 5600x
I know this is now an ‘older’ board and a long thread, but for anyone curious about the IOMMU groups:
Lite-On M.2 drives are in slot #4, bifurcated to x4x4
10Gb Ethernet card is in slot #5 @x4
Quadro P2000 is in slot #6 @x8
Along with the PCIe slot, one pair of USB ports (and what appears to be the SATA controller for the PCIe 3.0 x2 M.2 slot) are broken out into their own IOMMU groups, but otherwise, everything is fed through the chipset on a single IOMMU group.
I tried the latest firmware today (4.20) and could no longer boot from USB to ESXi whenever I changed the “PCIe 6/4 width” setting to anything other than x16. Reverting to 3.50 made it boot again normally.
I did check “Preserve BIOS settings” during the upgrade and haven’t tried resetting to factory defaults, so maybe there’s some way to make it work still.
Has anyone tried using a Athlon 220GE or 3000G with this motherboard? I’m thinking about trying to drop down to a cheaper, less powerful processor but I don’t want to spend money on it if it’s not going to work.
CPU support list shows there supported so they should work.
Dunno if the graphics from the CPU will work but this is a server board so it has the ASpeed AST2500 for basic video out anyway.
I don’t see any Athlon CPU’s on that list. I’m thinking they should work and there will be less lanes available but I wanted to check.
For anyone wanting to tweak your memory a bit with this motherboard, I found out from a German Youtube video that the timings in the BIOS are in hexadecimal so you you need to convert them from normal decimal format. I am now running a Samsung B die kit at 3200MHz 18-18-18-18-36-58 and trfc of 375. I tried to keep the settings fairly high for stability. I’m also not sure what voltage the memory is running at since it will not show proper voltages in Windows though I did try to tweak that as high as I could in BIOS as well. I could run these much faster with more voltage. These settings are a lot better than the default timings if you dont have a memory set with JEDEC.
I’m about to pull the trigger on this board, but I have few questions:
I will probably use it with a Ryzen 2700X. I see in the support page that 2nd gen CPUs should be used with BIOS 3.50 and not the latest. Is that really the case? Anybody using the latest with a 2nd gen?
ECC memory support. I assume this board only support Unbuffered ECC (according to the QVL). Depending on availability and price I would like to go with 4x16GB or 4x32GB. is 3200 ECC supported? Can I get a recommendation on what kits are known to work well with this board (QVL is very limited)
I will be using ESXi 6.7 (lest update). My plan is to passthrough 2 of the PCIe slots (GPU and HBA) as well as the 2 ASmedia SATA ports. Is that doable? Does the CPU choice (2nd gen) affects this?
Thanks!
That is not what it says. The table header says Since BIOS.
Ah, 2 different tables… yay 
Stick with 3.50. Direct from Asrock Support: “BIOS P4.20 or up no longer support 2000 series CPU.”
Basically, 4.20 drops support for older 1000 (which were never officially supported anyway) and 2000 series CPUs, and replaces them with 5000 series CPUs.
Mine just arrived with P4.20
And is working with Ryzen 1600 
Is that one of the OG 14nm ones or one of the newer 12nm ones? Could just depend on that because the 12nm variant was introduced when Ryzen 3000 was already out, and I would bet that they have different hardware identifiers.
I am using the X470D4U with Ryzen 2700X and once I had it running with the mentioned BIOS I just sticked to ASRocks recommendation.
What are you hoping to gain with the latest BIOS?
As for the ECC memory I used 4 sticks listed on the ASRock page to a total of 64 gigs. I did not try any other sticks as stability was my goal, not top performance.
I have 2 systems running like this without any issues.
Passthrough also worked for me with slots PCIE4 and PCIE5 for an HBA and 10Gbit NIC - but with ESXI 7.0
edit :
right after I wrote my post I noticed the IPMI GUI not displaying system components (System Information - System Inventory) , after that i proceeded to flash the newest version of BMC firmware. Here are my results
BMC 03.02.00 - BMC gets the IP address but is not accessible
since I had no IPMI GUI I had to revert to the DOS method of flashing
BMC 02.32.00 - BMC accessible but still no information in the Inventory
BMC 02.20.00 - the original version I had - accessible, no Inventory information
BMC 01.90.00 - everything fine lol
First gen 14nm. Only memory is running 1866 instead of 2133.
Will try with 2600x on weekend.
CPU 0:
vendor_id = "AuthenticAMD"
version information (1/eax):
processor type = primary processor (0)
family = 0xf (15)
model = 0x1 (1)
stepping id = 0x1 (1)
extended family = 0x8 (8)
extended model = 0x0 (0)
(family synth) = 0x17 (23)
(model synth) = 0x1 (1)
(simple synth) = AMD (unknown type) (Summit Ridge/Naples ZP-B1) [Zen], 14
huh, odd then 
Maybe it’s just one of those situations where they don’t guarantee anything but it can still work…
Single or dual rank memory?