unRAID 6.8.3 crash on new Ryzen 3950X build

[Bagpuss]

Hi All,

 

Just wondering if anyone can help me?

 

I've recently build a new Ryzen 3950X system based on the Asus Crosshair VIII Hero (WiFi) motherboard (latest BIOS v1302).

System has a Gigabyte RTX2080 OC graphics card, Mellanox ConnectX/2 dual NIC, 2x32GB Corsair Vengeance RBG Pro DIMMS, and 4 x SSD (2 x M.2 and 2 x SATA).

The system has been working fine with a native install of Windows 10 1909. I've run Karhu ram test and OCCT for 24 hours and seen no errors. I've also run multiple different benchmarks (3DMark, Cinebench R20, Blender, AIDA64 stress test and Prime95) with no issues. Maximum CPU temperature during any of these runs was around 71C. I'm not overclocking the CPU and am running the ram as it's rated 3600MHz.

 

Once I was certain that the build was stable under Windows, I wanted to test out performance in unRAID. I'm aiming to replace several Synology units with unRAID and also do some GPU passthrough for gaming.

 

I started off by working my way through SpaceInvader One's tutorial on setting up a Windows VM without passthrough. This worked fine, and I was about to embark on GPU passthrough when unRAID paniced. I've attached a screenshot showing the panic strings from the console. When the system crashed, no VMs were running, and I was just clicking between the tabs in the WebUI.

 

Following the crash, the whole system wouldn't even POST. I was getting a 0d error in the Q-Code readout on the motherboard (which is documented as being for future expansion), and the RAM error LED was lit orange on the motherboard. Googling reveals that others have seen this error on the previous versions of the Crosshair motherboard, but I couldn't find anything specific to the X570 variant that I'm running with.

 

In order to diagnose, I removed the Corsair memory, and installed 2 x 8GB T-Force Xtreem DIMMs. This caused the system to POST again, and I was able to reset the CMOS and get things up and running again. I then ran 24 hours of RAM and stress tests with these DIMMs under WIndows, which didn't show any problems. I then re-installed the Corsair memory, repeated the tests and still didn't see a problem.

 

I'd really appreciate peoples thoughts on what I should do next, as I really need stability in unRAID if I'm going to replace my Synology systems.

I've seen suggestions that some tweaks are needed with Ryzen, but I largely thought that these were no longer required with the 3950X. Should I consider any of the following:

 

- Change "Power Supply Idle Control" (or similar) and set it to "typical current idle" (or similar).

- Add RCU callbacks parameter to syslinux file.

- Run latest beta with 5.x kernel.

 

I've also attached the diagnostic output from the server, in case anyone wants to take a look.

 

Thanks,

 

Andy.

 

tower-diagnostics-20200418-1614-anonymous.zip

Edited April 18, 2020 by Bagpuss

[david279]

Try running the ram at 3200 or below. I have a 2700x it seems like ram overclocks can make the system unstable. Remember anything over 2133 for ram is a overclock. 

[Vr2Io]

Problems seems be CPU_idle, pls try disable C6 or disable some APM ( advance power management ) option for troubleshooting.

 

Pls

- Run latest beta with 5.x kernel.

Edited April 18, 2020 by Benson

[ensnare]

Having the same problem currently.

[JorgeB]

Also see this.

[ensnare]

Do you by any chance have docker containers set up with an IP on the br0 interface? If so, switch them to host or bridge mode. This completely resolved the random crashes w/ Ryzen on my end. Make sure there is nothing on the br0 interface.

[Furby8704]

did you resolve this?? im currently getting crashes