ASRock Rack B650D4U MicroATX for Ryzen 7000 processors

[carbide]

Check the BIOS if it has an option to enable or disable ECC functionality.

Yes, under Advanced -> AMD CBS -> UMC Common Options -> DDR Options -> DDR RAS -> DDR ECC Configuration. Options are Auto, Enabled, and Disabled.

Check the BIOS if it has an option like "Disable Memory Error Injection" which can be set to "False."

Also yes, one level above the previous. Options are Auto, True, and False.

Please check the Overclocking section of the BIOS, see if there's a way to cap max power usage, e.g. at 105W, 90W, etc. These CPUs underclock well, so this could make your 12 cores or a 16 core more useable WRT thermals.

This is in two different places, one is Advanced -> AMD Overclocking -> Precision Boost Overdrive where you can put in manual entries for PPT, TDC, and EDC the other is under Advanced -> AMD CBS -> SMU Common Options where you have "TDP Control", "PPT Control", "Thermal Control", and a few others. Not sure which setting takes priority over the others.

Also, to answer my own question from Patrick's article, there is an option for bifurcation of the PCIe x16 slot. This is found in Advanced -> AMD PBS -> PCIe/GFX Lanes Configuration. Options are x16, x8/x8, x8/x4/x4, and x4/x4/x4/x4

EDIT: To keep others from beating their head against this issue as well, there is a BIOS setting that has to be used to keep system output using the BMC VGA port (and the HTML5 console if you're planning to use that) after leaving the BIOS:
Advanced -> AMD PBS -> Graphics Features -> Primary Video Adapter: "Ext Graphics (PEG)"

 

[SapphironZA]

Does it need the new 6.2 kernel for Proxmox 7.3/4 to function properly?

 

[carbide]

Does it need the new 6.2 kernel for Proxmox 7.3/4 to function properly?

I’m not sure. I’ve successfully booted it with VMware 7.0U3, Win10 22H2, the latest version of Slax, and Proxmox 7.4 with whatever its default kernel is. I haven’t had any boot-stopping driver issues with anything so far.

 

[marcoi]

Are you running the beta bios? any issues with 10gb nics under ESXI?

 

[carbide]

Are you running the beta bios? any issues with 10gb nics under ESXI?

Yes, and none that I’ve seen. It detected them immediately and I have one set up for vmotion/provisioning and the other as an uplink for a dswitch.
Spoilers for the writeup I plan to write, but that beta bios is almost a requirement for reliable booting. Could be a quirk of my build though, planning to run a few more tests tonight.

Does it need the new 6.2 kernel for Proxmox 7.3/4 to function properly?

Also, I did double-check this last night. The 5.15 kernel was the default for Proxmox 7.4 and it booted fine with the exception of an ACPI driver error that didn’t appear to affect anything (I could still use the server and create/run VMs at least). I installed the 6.1 opt in kernel and the error went away; didn’t notice that there was a 6.2 kernel available. That’ll teach me to check forum posts more closely.

 

[marcoi]

the manual mentions OCuLink, but i don't really see them on the images of the board that are online. Can you confirm if they are present or not?
also are you running 2 or 4 memory slots? I'm planning to run 128GB but not sure if the memory will run at 4800 speed or 3600 speed if bios limited when running 4 slots.

 

[RolloZ170]

the manual mentions OCuLink, but i don't really see them on the images of the board that are online

Note
-M.2 share lanes with OCuLink1,
M2_1 and OCuLink1 are BOM option
-4 SATA 7-pin share lanes with OCuLink2,
4 SATA 7-pin support possible by depopulating OCuLink2 and other BOM change

 

[carbide]

the manual mentions OCuLink, but i don't really see them on the images of the board that are online. Can you confirm if they are present or not?
also are you running 2 or 4 memory slots? I'm planning to run 128GB but not sure if the memory will run at 4800 speed or 3600 speed if bios limited when running 4 slots.

OCulink, as Rollo said, are options that are not part of the off the shelf board, though there are two pads for them that aren’t too hard to find. One is near the bottom between the chipset and the usb3 header, the other is in front of the SATA ports. Also, I’m only running it with two 32GB sticks. I’d heard of issues with the platform running four sticks so I decided not to risk it. Hopefully good support for the 24/48GB sticks happens sooner rather than later.

 

[codec]

1. Check the BIOS if it has an option to enable or disable ECC functionality.
2. Check the BIOS if it has an option like "Disable Memory Error Injection" which can be set to "False."
3. Run the beta version of Memtest86 and check to see if ECC is supported and if error injection is functioning: https://www.passmark.com/temp/memtest86-pro-usb-10.2.1027.zip I would be happy to see a screenshot if error injection is functioning. (I have a Gigabyte B650 board and it does not support error injection.)
4. Here are instructions to confirm if your board supports error injection on Linux: APEI Error INJection — The Linux Kernel documentation
   1. Check to see if dmesg does show anything like this:

      ACPI: EINJ 0x000000007370A000 000150 (v01 INTEL           00000001 INTL 00000001)

   2. Check if "/sys/firmware/acpi/tables" contains an EINJ file.
   3. Check if BIOS has options for APEI or WHEA.

Echo what was already said about 1 and 2. Looks like the beta of Memtest86 was pulled and I don't have access to the Pro version so can't test that way. Running edac-util I get an EINJ table not found error, but haven't looked further than that.

Please check the Overclocking section of the BIOS, see if there's a way to cap max power usage, e.g. at 105W, 90W, etc. These CPUs underclock well, so this could make your 12 cores or a 16 core more useable WRT thermals.

There is both an Eco Mode enable/disable, which locks it to 65W, and a free-form TDP limit option.

Does it need the new 6.2 kernel for Proxmox 7.3/4 to function properly?

Echoing the previous comment. Boots fine on default 5.15. I upgraded to 6.2 and have no issues either.

also are you running 2 or 4 memory slots? I'm planning to run 128GB but not sure if the memory will run at 4800 speed or 3600 speed if bios limited when running 4 slots.

Straight from the spec page, when 2 slots are filled per channel it will run at 3600.

 

[carbide]

Ooh, another early adopter has entered the ring? I’m guessing you’re running the beta BIOS too, since eco mode didn’t show up on mine until after I’d updated it.

I’m not sure what you intend to use yours for, but have you tested any gpu/pci passthru on it? I’ve gotten some interesting results so far on my attempts. I need to write up something more formal to see how other people interpret it, but I think there are some bugs around device resets.

 

[marcoi]

hopefully it just a bios update to get memory stable at 4800 with 4 slots filled. Anyways, ill be watching this thread but holding off on buying a pair as replacements for my 4 other systems running asrock MBs with intel e-2200 cpus.

 

[carbide]

Honestly the memory issues I’ve experienced and that I’ve seen others report about remind me very much of the early days of Zen1/AM4. I was an early adopter then too, and it took quite a few bios/AGESA updates before I was reliably booting 4x8GB at 3200, but we did get there.

 

[VMman]

This could be a nice option to expand the storage on these since it has 4x4x4x4 support.

ASRock Rack RB1U4OCU_G4 Server Riser Card