you need to upgrade kernel version
Hi,
To which version please ?
Iām actually with the latest one from backport
Linux home-server.local 5.9.0-0.bpo.5-amd64 #1 SMP Debian 5.9.15-1~bpo10+1 (2020-12-31) x86_64 GNU/Linux
Really really late reply.
Aside from having to re-configure the entire BMC again (setting up user accounts & disabling the bond) again - No.
Hi. I am planning to buy the X470D4U. There is a close-by shop that has one, and I spotted used AMD 3700X at a good price (I notice there is beta BIOS / BMC with support for 5000 series, and AMD 5800 is tempting, but I think I will pass, because I want something cheap) that I plan to use it with it.
The most critical part for me, follows:
Does remote power cycle (hard reset) work with ipmitool from other machine over dedicated IPMI port? I need ability to automatically and programmatically able to restart the system using ipmitool from another computer, even if the CPU or Linux kernel crashes on X470D4U.
The machine will be booting remotely over PXE and use NFS or iSCSI to mount stuff, and test hardware driver (PCIe stuff), so system crashes and non-recoverable kernel hangs are expected.
Yes, it will work, ipmitool is pretty much the only way I control the power/reset switches of my X470D4U.
1 Like
I thought it should work, but SHOULD and WILL (ātestedā) are different things. 
Considering history of the bugs in IPMI on some ASrock boards, I wanted confirm.
Thanks!
Having strange problem with x16 PCIe slot. Upgraded Avago 9200 HBA to Avago 9400 HBA and now the PCIe width is set to x1 instead of x8 on boot. The IPMI log shows ābus degradedā entry when this happens. TrueNAS also shows notification about it. Sometimes it does not happen. If I put the HBA card to different x8 slot, then it works. And if I put x4 Mellanox 10g NIC to the suspect x16 slot, then the NIC also works at correct PCIe width. Upgraded firmware on HBA and the motherboard, does not fix it. Only leaving the card to the x8 slot works. Can it be PCIe slot physical damage after handling the card in the x16 slot? Or motherboard defect? The Avago 9400 HBA is PCIe v3.1, but donāt think thatās the issue, since on rare occasions the bus degraded error does not happen.
Perhaps giving the slot a cleanup can help.
I have some issues getting PCIe passthrough in ESXi 7.0 to work with this board.
I have three PCIe cards that I want to passthrough to the same virtual machine - two LSI SAS2008 HBAs and a Digital Devices Cine CT TV card. All three pieces of hardware show up in the ESXi hardware list, but only the HBAs can be toggled for passthrough.
When I try to toggle passthrough for the Digital Devices card, ESXi fails with this error:
Key
haTask-ha-host-vim.host.PciPassthruSystem.updatePassthruConfig-1852875867
Description
Update PCI passthrough device configuration
State
Failed - An error occurred during host configuration.
Errors
Operation failed, diagnostics report: GetDeviceID failed.
If I refresh the hardware list after this, the passthrough status has changed to āEnabled / Needs rebootā. Rebooting doesnāt help, however.
Has anyone else had this issue? Iām starting to suspect that this issue only happens with this particular card, given that Google doesnāt really have anything on this āGetDeviceId failedā error.
i think some people have report more success with passthru on the later bios versions
Iām on the latest stable BIOS (P3.50). I had to change āSlot 6 PCIe widthā (canāt remember the exact name) to 8x8 in order for all three cards to appear in ESXi at all. Iām wondering if thereās something else that needs changing?
Hi again. I got my X470D4U, and Ryzen 3700X with stock cooler. I also got HyperX HX432C16FB3K2 (Kingston/HyperX Fury, 16GB per module, two modules, DDR4-3200), hoping I can run it at high frequency.
The modules fits nicely, run cool even during stress testing, but of course by default they run slow, at 2400MT/s.
The BIOS is really complex and hard to configure memory OC quickly. I did find some settings for UMC, and frequency is available up to ā3000MHzā. So, I just tried that, with all the other timings set to Auto. The system didnāt boot again. LOL. In fact I was scared, because most desktop mobos will try to boot, and revert back to previous settings if it fails memory training and tests. That didnāt happen. In fact there was zero post codes on the LED display, even after powering off and on few times. It also didnāt help to short the Reset CMOS jumper.
Fortunately, after removing the CMOS backup battery, switching off standby power, waiting a bit and starting again, it reverted to defaults. That was a relief.
But still, I want to find some settings that make the RAM work better than out of the box. Any ideas or guidelines, of how to approach it, i.e. which settings to change first, what conservative values to set, etc.
BTW. I am running BIOS P3.30, should I upgrade?
@baryluk most likely the solution is my previous post: ASRock Rack has created the first AM4 socket server boards, X470D4U, X470D4U2-2T - #1298 by Tenrag
TL:DR: not a gaming board, no XMP, frequencies in the bios are not doubled. (if you want 3000Mhz then you need to set 1500)
@Tenrag Thank you. What frequencies and timings you used on your system to OC memory? (Also which modules and what amount of them).
2 ECC sticks of QVL KSM26ED8/16ME
I have the frequency set to 1333 (2666) and the timings to auto. Never tried any non-ECC kits.
1 Like
Hey all,
Iām hoping someone may be able to assist meā¦
Iām revisiting an old issue I was experiencing with my system that I abandoned and now am trying to again make it work
Iām running the X470D4U, Ryzen 3950x with the following add-in cards
Nvidia Quadro P2000
Mellanox ConnectX-2 dual SFP+ card
LSI SAS9211-8i
Iām running unRAID, just updated to 6.9.0 last night
I also updated my IPMI and BIOS to the latest stable builds from ASRockRack
unRAID is failing to initialize the GPU
I know that it worked in the previous build I had, I also had it steal video output last night following the firmware upgrade. I switched that back in the BIOS but Iām still struggling to make it so unRAID will recognize it
My logs in unRAID have the following chunk of info regarding the device
Mar 3 18:16:50 unRAID kernel: nvidia: loading out-of-tree module taints kernel.
Mar 3 18:16:50 unRAID kernel: nvidia: module license āNVIDIAā taints kernel.
Mar 3 18:16:50 unRAID kernel: Disabling lock debugging due to kernel taint
Mar 3 18:16:50 unRAID kernel: nvidia-nvlink: Nvlink Core is being initialized, major device number 245
Mar 3 18:16:50 unRAID kernel:
Mar 3 18:16:50 unRAID kernel: nvidia 0000:2b:00.0: enabling device (0000 ā 0003)
Mar 3 18:16:50 unRAID kernel: NVRM: This PCI I/O region assigned to your NVIDIA device is invalid:
Mar 3 18:16:50 unRAID kernel: NVRM: BAR1 is 0M @ 0x0 (PCI:0000:2b:00.0)
Mar 3 18:16:50 unRAID kernel: NVRM: The system BIOS may have misconfigured your GPU.
Mar 3 18:16:50 unRAID kernel: nvidia: probe of 0000:2b:00.0 failed with error -1
Mar 3 18:16:50 unRAID kernel: NVRM: The NVIDIA probe routine failed for 1 device(s).
Mar 3 18:16:50 unRAID kernel: NVRM: None of the NVIDIA devices were initialized.
Mar 3 18:16:50 unRAID kernel: nvidia-nvlink: Unregistered the Nvlink Core, major device number 245
Mar 3 18:16:51 unRAID kernel: nvidia-nvlink: Nvlink Core is being initialized, major device number 245
Mar 3 18:16:51 unRAID kernel: NVRM: This PCI I/O region assigned to your NVIDIA device is invalid:
Mar 3 18:16:51 unRAID kernel: NVRM: BAR1 is 0M @ 0x0 (PCI:0000:2b:00.0)
Mar 3 18:16:51 unRAID kernel: NVRM: The system BIOS may have misconfigured your GPU.
Mar 3 18:16:51 unRAID kernel: nvidia: probe of 0000:2b:00.0 failed with error -1
Mar 3 18:16:51 unRAID kernel: NVRM: The NVIDIA probe routine failed for 1 device(s).
Mar 3 18:16:51 unRAID kernel: NVRM: None of the NVIDIA devices were initialized.
Mar 3 18:16:51 unRAID kernel: nvidia-nvlink: Unregistered the Nvlink Core, major device number 245
Looking in Tools > System Devices, I see an IOMMU group for the card
I am, admittedly, not super hip on the Linux side of things and what the IOMMU ties to, if itās anything I need to be concerned with, etc
Itās not a super urgent matter just more of an annoyance that I canāt get this card to pickup.
I feel like, judging on the logs, Iāve managed to bunk up something in the BIOS but for the life of me I canāt sort out what
Any assistance or direction is appreciated 
@fiore00713 I dont know unRAID myself, does it do containers or virtual machines? or what is using the GPU? I am gonna drop here some buzzwords: VFIO, IOMMU, SR-IOV
I think your PCIe switch setup is ok since you can see the card in unRAID, but you may wanna still check that the PCIe is in 2x8 mode in your BIOS ā Advanced ā AMD PBS.
You might want to see what the SR-IOV does [enabled] in the BIOS ā Advanced ā Chipset Configuration
I had to once enable IOMMU on Proxmox (to pass-through GPU to VM). In Proxmox for passing through I needed to configure /etc/default/grub with text-editor and add line there GRUB_CMDLINE_LINUX_DEFAULT="quiet amd_iommu=on"
But! this is on Proxmox so look for the unRAID specific instructions if this is needed.
For you X470D4U Folks here is my setup:
- MOBO: X470D4U2-2t
- CPU: Ryzen 3950X
- Cooler: Inter-Tech Dynatron A-24
- RAM: 4 x 32GB Samsung 2666Mhz ECC UDIMM (M391A4G43MB1-CTD)
- Chassis: Inter-Tech IPC 2U-20255
- PSU: Inter-Tech ASPOWER R2A-DV0550-N (550W redundant 1+1)
- NIC: Intel X520-DA2
- STORAGE: 2 x Micron 7300 PRO m.2 NVME
- PCIE-RISER: Supermicro AOC-SLG3-2M2-O
Mistakes I think I made and some suggestions:
- X470D4U2-2t is missing 4x PCIe slot (replaced with 1x something) => pick X470D4U for the flexibility.
- Cooler is the best available in this form factor, but its for 95W CPU => undervolt CPU
- 1 x Samsung DoA, thats why 3 dimms in picture => RMA, board works with 3 still
- 2 x NVMe was classic dumb move on this board as it forces to run 4x/4x/4x/4x mode, no 8x/4x/4x option => for boot drives use SATA, (2 for Raid1). For more then 1 speedy NVMe drives, use suitable 8x riser that does bifurcation itself.
1 Like
@polpo Iām ultimately trying to pass the GPU to a docker container, the latest update includes the NVIDIA driver plugin that installs (currently) v455.45.01 to the system. From there I would be able to pass the GPU to the container.
PCIe is on 2x8 mode so that should be okay
I will need to check the SR-IOV that you mentioned as Iām not familiar with it but itās worth a shot
Iāll give it a try when I have the opportunity to reboot the system and see if it does me any good
Thanks!
Currently using the X470D4U + Ryzen 5800X with the latest 4.02L UEFI and everything is working fine. Didnāt have to get a 3000 series, just upgraded the UEFI via the IPMI console and that was it.
Only issue Iām having with the board is the fact that the fans keep spinning up for every single little thing. Seems to me that the temps for CPU are way higher than they should be.
1 Like