I have X570D4U and I’m trying to passthrough SATA controller. I have installed proxmox on drive in M2_2 PCH port (in port M2_1 CPU I have nvme drive which later I plan to passthrough to Win11 VM) and trying to passthrough SATA controller with 8x SATA ports (there are two SSD drives atm).
Acc. to my iommu script I have two SATA controllers in different groups. I have checked with udevadm each drive to see if I’m passing right PCI device. M2_2 drive is in 2b:00.0 and both SSD in SATA ports are in 2c:00.0.
Unfortunately, every time I start VM with 2c:00.0 passthrough it completely breaks proxmox because it loosing access to M2_2 SSD drive where it is installed.
Is it possible to passthrough SATA controller with ports and leave M2_2 SSD available to host?
Install hwloc and check with lstopo. If you don’t want to install X11 stuff on your server you can install hwloc-nox and check with lstopo-no-graphics instead.
What you’re seeing is one SATA controller in the PCH and one SATA controller in the I/O die (on the processor). Most X570 boards are wired up such that all the SATA ports connect through the controller in the PCH.
It is possible however unlikely that some of the SATA ports connect through the controller in the I/O die. You can try plugging a SATA drive into all the different ports to see if it pops up on the other controller.
The one exception to what I wrote above is that the primary M.2 slot is usually wired up such an M.2 SATA drive will connect through the controller in the I/O die. So one possible solution for you is to swap the two M.2 drives (i.e. put the NVMe M.2 in M2_2 and the SATA M.2 in M2_1). Unfortunately that will leave your NVMe drive connected through the PCH, which is probably undesirable.
The other alternative is to replace the SATA M.2 in M2_2 with an NVMe M.2.
I just got this board (the X570D4U) and it runs fine with a 5600 and 2 X 16 GB of ECC RAM from the QVL. I currently have Linux Mint 21.2 on it just to benchmark it and test it for stability.
It’s going well so far except for the IPMI.
I can see the IPMI web pages fine, it’s the remote control by HTML5 or Java. Both those come up blank, no display, and there are no remote control functions at all. Then (both) seem to hang the web interface, when I close them I get a “Processing” spinny on any IPMI page until it crashes. A few minutes later the web interface is responsive again.
The video works fine over VGA. I also selected “D-sub” as the primary graphics adapter in the BIOS (Advanced → AMD PBS → Primary Graphics Adaptor).
I’m connected to the BMC LAN port and using the IP address assigned to that port.
I also tried connecting over VNC. Settings - Services shows it’s active but not running on any interface. I set VNC in Settings - Media Redirection - Remote Session. VNC won’t connect.
Is there any setting I’m missing? Any tips to get remote display working?
Good idea. This was on Chrome on Linux, I have tried Chrome on Windows (same result) and Edge on Windows (can’t access IPMI at all). I’ll try other browsers.
I can confirm that the IPMI interface’s remote viewer appears to be very browser-sensitive. Not so much just that it doesn’t like certain browsers (which I’m sure is true), but that it sometimes stops liking browsers that previously worked.
If a browser is giving you trouble, and especially if it worked before but isn’t working now, try opening a private/incognito browser window. That’ll bypass whatever confused cookies or local cache issue is happening.
Didn’t work for me, it’s throwing some errors about resources being blocked
06:48:24.381 viewer.min.js:6 App initialized...
06:48:24.384 viewer.min.js:17 KVM default-screen
06:48:24.385 viewer.min.js:7 doing layout after render
06:48:24.385 viewer.min.js:16 Uncaught TypeError: Cannot read properties of undefined (reading 'indexOf')
at t.initialize (viewer.min.js:16:3934)
at n.View (viewer.min.js:5:13871)
at l.View (viewer.min.js:6:10380)
at new t (viewer.min.js:5:20789)
at t.default (viewer.min.js:17:29695)
at t.execute (viewer.min.js:5:16746)
at Object.callback (viewer.min.js:5:16600)
at viewer.min.js:5:19918
at Function.ae (viewer.min.js:4:18550)
at n.History.loadUrl (viewer.min.js:5:19862)
06:48:24.456 viewer.min.js:7 doing layout after render
06:48:24.505 api/configuration/project:1
Failed to load resource: the server responded with a status of 401 (Unauthorized)
06:48:24.519 api/configuration/runtime:1
Failed to load resource: the server responded with a status of 401 (Unauthorized)
Oof. Make sure you’re allowing pop up windows on that domain, and also check to make sure the admin user has authorized remote viewing (there’s a setting for this in one of the dropdowns).
Your browser also might be restricting javascript from running freely on that domain, which should be adjustable.
Even if you are admin user, you still have to authorize … yourself. While logged in. As you. ¯_(ツ)_/¯
Last week, I bought an entirely new NAS system based on the X570D4U.
Initially had to boot with some other RAM to set up an XMP profile (DDR4-3000), then I switched to my new ECC memory and was able to boot, install TrueNAS etc, all fine. Updated the BIOS to 1.5 and now I can’t see XMP profiles anywhere but I’ve set the RAM speed in the OC Tweaker, as well as manually setting the voltage (but I still get voltage errors occasionally, 1.71V rather than the necessary 1.2V, but this seems to be only at boot?)
Now, my system seems to reboot every few hours. I’ve combed through some logs without any success. TrueNAS doesn’t report an unexpected shutdown at all. I’ve tried changing a bunch of stuff like lowering the memory speed and setting the CPU to eco mode but nothing seems to help. Temperatures seem fine, CPU is at 28°C.
CPU: AMD Ryzen 7 Pro 4750G
RAM: Kingston KSM32ED8/32HC
MB: ASRock Rack X570D4U
PSU: Seasonic Prime TX-650
HDD: 8x Seagate Exos X16 ST16000NM001G 16TB
NVME: 1x Samsung 980 250GB + 1x Samsung 970 EVO Plus 2 TB
Really not sure what this could be, it’s all brand new equipment.
EDIT: Flashing bios back to 1.20 seems to have stabilized my system.
EDIT2: It actually no longer seems stable without any changes.
Has anyone managed to undervolt the x570 chipset on the X570D4U board?
Temps are on average around 55-60 but want to get them as low as possible by undervolting.
It’s supposed to work in 1U/2U chassis with a lot of airflow. AsRock Rack uses the board in several of their barebones and they all have high rpm 40mm/60mm datacenter mode cooling
55-60°C is totally fine. x570 has an operation temperature of up to 85°C. Silicon is not a human body with proteins dissolving once you approach 45°C.
I use a 40mm Noctua because it was going 75-78°C on idle on my server. 55°C is a level where I would never have considered any additional cooling. Even my memory exceeds this at times.
I’ve had the X570D4U-2L2T for a while now. Working great! I’m considering getting a Ryzen 9 5900 for it. I was able to upgrade the BIOS without issue.
Is there any major reason for me to upgrade the BMC? I have the IPMI socflash tool on a FreeDOS img all ready to go… but some of the reports have made me a bit gun shy. What’s the major reason, if any, to upgrade the BMC? Thanks!
While this is usually the way to go, I’ve found myself eagerly awaiting BIOS/BMC updates on this machine more than once.
Parts of the BMC and BIOS weren’t feature complete when I bought it, and still might not be (particularly re: the ability to control BIOS settings from within the IPMI interface). At least twice they’ve had to push updates for new CPU support, but that should be less of an issue now that AM4 is EOL.
That said, at this point I don’t feel like there are any major features missing that are actually necessary, depending on whether the BIOS settings thing works now.
I think there have also been some security mitigations released by AMD that haven’t actually been patched on our side of things? It’s been about a year since the last BIOS update.
An old wisdom in computing that served me well for decades.
My X570D4U-2L2T is still running stock everything (P1.40 BIOS and 1.2 on BMC). IPMI reports 340 power-on days, but server was offline at times during vacation, but BMC just keeps going.
I had my trouble with memory clocks (half of this megathread is about memory after all) and cooling the chipset, but otherwise I got everything I ever wanted for BIOS options, features, IOMMU groups…just a workhorse that keeps on running.
I could use more lanes, but that’s an AM4 thing in general.
