Ubuntu 24.04 and Mint 22 - Fail to boot with dual AMD gpu's passed through

[erickgruis]

Hey guys,

I have a bunch of VM's setup and running with dual AMD Radeon VII's passed through. I have each GPU attached to it's own monitor.
Everything works great with: PopOS, Arch, Mint 21.3, Ubuntu 22.04.4, Windows 10/11, MacOS.

But, when I try to pass both GPU's to Ubuntu 24.04 or Linux Mint 22, they both either crash during the boot process (Ubuntu) or give garbled/blank screens (Mint).
Both work fine when passing only one GPU.

Does anyone know what is different about these 2 distros compared to the others?

Arch is running kernel 6.9.7 and it's fine. PopOS is on kernel 6.9.3 and it's fine. Mint 21.3 is on kernel 5.15.0 and it's fine. Ubuntu 24.04 and Mint 22 are on kernel 6.8.0-39-generic. I did try installing a newer kernel to Ubuntu 24.04 but get the same crash on boot.

On the boot for Ubuntu I get errors:

[drm:smu_v11_0_i2c_transmit [amdgpu]] *error* Received I2C_NAK_7B_ADDR_NOACK !!!
[drm:smu_v11_0_i2c_xfer [amdgpu]] *error* WriteI2CData() - I2C error occured :1
[drm:amdgpu_ras_eeprom_init [amdgpu]] *error* Failed to read EEPROM table header, res:-5

I'm running Proxmox VE 8.2.4 on kernel 6.8.8.4-pve.
AMG Ryzen 3900xt
64GB Ram
Radeon VII (2)

 

[leesteken]

Arch is running kernel 6.9.7 and it's fine. PopOS is on kernel 6.9.3 and it's fine. Mint 21.3 is on kernel 5.15.0 and it's fine. Ubuntu 24.04 and Mint 22 are on kernel 6.8.0-39-generic. I did try installing a newer kernel to Ubuntu 24.04 but get the same crash on boot.

The amdgpu driver of the Proxmox, Ubuntu and Zabbly(?) kernel 6.8.x crashes on my RX570 (which won't work until a host reboot even with vendor-reset), so I expect that it's the Ubuntu kernel version (which is used by Ubuntu,.Mint and Proxmox).

 

[erickgruis]

The amdgpu driver of the Proxmox, Ubuntu and Zabbly(?) kernel 6.8.x crashes on my RX570 (which won't work until a host reboot even with vendor-reset), so I expect that it's the Ubuntu kernel version (which is used by Ubuntu,.Mint and Prox

You are talking about the Proxmox kernel? Not the guest VM kernel, correct?

 

[leesteken]

You are talking about the Proxmox kernel? Not the guest VM kernel, correct?

I did have the issue both with the Proxmox kernel 6.8 and with guest kernels 6.8 (when using passing through of the GPU).

 

[erickgruis]

I did have the issue both with the Proxmox kernel 6.8 and with guest kernels 6.8 (when using passing through of the GPU).

I tried kernel 6.5 (Proxmox) and kernel 6.9.9 (Ubuntu) but still same result. The funny thing is, everything works fine with one GPU. It's only when I try to pass both GPU's to those VM's that it crashes. I guess I could try kernel 6.9 on Proxmox.