Asrock X570d4u, x570d4u-2L2T discussion thread

New member here, specifically for this thread!

I picked up a X570D4U last Friday to replace a Asrock B450M Pro4-F that’s been running my core network ESXi server for three nearly flawless years.

Maybe I’m doing something wrong, but it’s been utterly terrible. I just cannot get the board to not hard lock minutes after booting an OS (Win10 for quick and dirty testing, ESXi 8.03 for use).

I’ve been building PC and servers since Intel Xeons were slotted and bigger than modern mini PCs so this is not remotely my first rodeo. Still, entirely possible I’m missing something here.

Reused components

  • AMD Ryzen7 5700G
  • Noctua 120mm tower cooler
  • 2x Corsair CMK64GXM42E3200C16 2x 64Gb kits (128Gb total)
  • Samsung PM991a 1TB NVMe SSD. Runs very hot, disappears from devices after some (literally minutes) usage.
  • IBM/LSI(Avago/Broadcom?) ServeRAID-BR10il card flashed to plain SAS2004-IT, FW v20.00.07.00
  • Mellanox Connect-X4 dual 25GbE card CX4121A, FW updated to whatever was current mid 2024

OG BIOS (1.10) was too old to POST the chip so I updated it to latest beta (1.57) from the website, along with latest beta BMC FW (2.07).

I’m getting random reboots and freezes during installs. When Windows installs, it’s plagued with random hard freeze lockups. The Samsung PM991a SSD simply disappears from the BIOS and the heatsink is roasting hot.

Replacing that with a Samsung 970 EVO Plus 1Tb sorts both the heat and disappearing issues. Doesn’t stop the system hard locks and reboots. ESXi or Windows doesn’t seem to matter, although ESXi reliably locks up 1-3 minutes after booting just sitting there idle.

I’ve halved the Corsair ram kits, swapped it out for 2x Crucial CT16G4DFRA32A.M16FR, played with voltages to the ram (the Corsair wants 1.35v which upsets the BMC logs very much).

  • Pull the SAS2004-IT card - no fix
  • Pull the Mellanox CX4 card - no fix
  • Flash back to BIOS v1.50 - no fix
  • BMC firmware back to 1.39 - no fix

I’m kinda at the end of my road here so looking for hail Mary suggestions as to the secret BIOS setting that says “don’t crash in minutes”.

AsrockRack tech support is 48 hours deep since I reached out and got the automated reply so there’s not much to go on there…

Planning to wipe my Proxmox PVE server clean and start from scratch in the next week or so.
Just wondering if many people here have used the Beta 1.78 BIOS for the X570D4U-2L2T?
I’m currently on 1.70 with BMC 1.35 which is what it was delivered with, but there are security fixes in 1.78 so updating seems like the wise thing to do. Bonus points if it fixes the sensor graphs not updating.

@Aetherone you’ve tried all the obvious things I would. Doesn’t sound good.

Just throwing some ideas out there.
Have you tried running the system for a while from a live iso on a USB? Just to cut storage out of the equation completely.
Also has all your testing been done using the integrated GPU on the 5700G? Have you tried not plugging in any displays and just testing it remotely via the IPMI/KVM? Or maybe adding a GPU if you have a spare laying around trying with that?

Yeah, if it posts it’ll Passmark Memtest for the four full loops.
Windows USB boot is hit or miss, mostly hit.
Not using the integrated GPU, only the aspeed VGA output or IPMI/KVM. As part of the testing I’ve disabled the iGPU entirely.

No point adding a GPU, I don’t have the slots to host it when the machine is doing it’s thing. Hence iGPU equipped CPU for the OG desktop board and IPMI equipped server board.

Hello,
have you a fan at the heatsink of the chipset at the right bottom of the Mainboard. You have many Cards and the chipset heatsink can be very hot.

Can you test without any Card, only one ram, CPU and NVMe (at the SLLOT1 CPU, SLOT2 is Chipset). When this fails then without NVMe.

Yes, I’d read that the chipset runs quite hot and is expecting a rack case worth of airflow so I printed up a mount for a 4020 Noctua fan right over it’s heatsink. It runs cool to the touch.

Have tried removing all my cards and only the boot drive in NVMe #2 as that’s boot, performance sensitive data (AKA all of the VM’s) will go in NVMe #1 if this thing ever works.

Have not come down to one stick, but subbed in 2x Crucial CT16G4DFRA32A.M16FR for the 4x Corsair. Not ideal as 128Gb is the sweet spot.

Please test in NVMe 1 fiirst. I also have this Board (the one without 10GBE) and Have trouble with NVMe in Slot 2 as Boot device (also an 970 EVO Plus 1TB). Then i Switches NVMe 1 to boot and NVMe 2 as VM´s and now since years (02/2021)i have no Problem anymore.

Don’t quite under these version numbers. Per ASRock Rack website, the posted latest ones are BIOS 1.70 (or 1.78 if you at the “Beta Zone”), BMC 01.35.00.

@ZeeGee As per here:

AsrockRack X570D4U vanilla, not the -2L2T

After much fiddling (and ZERO response from either AsrockRack or Newegg’s tech support - seriously I was looking at my CC providers chargeback rules), I do believe I’ve located the “do not hard lock when idling the OS” button.

Different NVMe slot, different ram, diff PSU, empty slots, every BIOS version from 1.20 upwards (1.20 was the first to support my 5700G). C states disabled or not, AMD PBO disabled or auto or tuned to minimal boost, CPB disabled or not, more ram volts, less ram volts None of it made any real difference. Windows or ESXi simply hard locked after a couple of minutes sitting on the desktop.

Then I finally twigged to the commonality of the failure mode - it’s always sitting idle on the desktop. Pass mark memtest, fine for hours and hours and hours. AIDA stress test, also fine for 12+ hours as long as I got it running before it locks up.

Sitting idle is the cause of lockups.

So it turns out that if I head into the manual overclocking and simply lock the chip at 4.4Ghz (leaving PBO & CPB disabled), it stops crashing after a couple of minutes. It sits idle for hours in both Windows and ESXi without further issue.

I’m still bench testing and evaluating but it’s looking remotely solid for now.

I also seem to have gotten the Samsung PM991a to work with disabling both "PCIe Redriver"entries in the BIOS. Unsure of how/what/when they’re supposed to apply (long PCIe cables for rack chassis maybe?)?

I’m assuming the BIOS is just bugged to hell and considering the batpoop insane layout and many many duplications of settings that don’t respond to their doppelgangers, I don’t feel unjustified in thinking this way.

Will really really be trying to never buy AsrockRack again ever.